**Long-Term Effect of Pig Slurry and Mineral Fertilizer Additions on Soil Nutrient Content, Field Pea Grain and Straw Yield under Winter Wheat–Spring Barley–Field Pea Crop Rotation on Cambisol and Luvisol**

**Lukáš Hlisnikovský \*, Ladislav Menšík, Pavel Cerm ˇ ák, Kateˇrina Kˇrížová and Eva Kunzová**

Crop Research Institute, 16106 Prague, Czech Republic; ladislav.mensik@vurv.cz (L.M.); pavel.cermak@vurv.cz (P.C.); krizovak@vurv.cz (K.K.); kunzova@vurv.cz (E.K.) ˇ **\*** Correspondence: l.hlisnik@vurv.cz; Tel.: +420-233-022-248

**Abstract:** Different fertilizers have different effects on soil chemistry and crop yields. In this paper, we analyzed how long-term and regular application of mineral fertilizers, pig slurry and their combinations (15 fertilizer treatments totally) affect soil pH, nutrient content and yield of field pea at two sites with different soil (cambisol and luvisol) and climatic conditions. The long-term trials evaluated in this paper were established in 1972 at Pernolec and Kostelec, Czech Republic. Results of the soil analyses (evaluated period) are from the years 2015–2020, covering two sequences of crop rotation (winter wheat–spring barley–field pea). The fertilizer treatments significantly affected the soil reaction; application of mineral fertilizers and their combinations resulted in the lowest pH values. On the other hand, the same treatments provided the highest yields and left the highest pool of nutrients in the soil. Pig slurry can provide the same yields of field pea as mineral NPK fertilizers, without a negative effect on soil reaction. Analyzing the mineral fertilizers only, a reasonable dose of N (according to the linear-plateau model) can range from 73 and 97 kg ha−<sup>1</sup> N in Pernolec, according to the weather conditions.

**Keywords:** *Pisum sativum* L.; organic manure; NPK; pH; SOM; macronutrients; nutrient content

#### **1. Introduction**

The nutrient content of the soil is one of the parameters determining its fertility and quality. It is a parameter influenced by a wide range of natural, anthropogenic and interrelated factors such as soil type [1], farming method (conventional, organic farming) [2], crop rotations and fertilization [3–5], microbial activity in the soil [6], or soil organic matter content [7]. The application of fertilizers represents the main way of supplying nutrients to the soil; for the crops grown, fertilizers thus directly affect soil chemical [8–11], physiological [12,13] and biological [14,15] properties and crop growth.

Fertilizers are divided into three categories, namely mineral and organic fertilizers and organic manures. They differ in origin, composition and nutrient content, speed of nutrient release and availability to farmers. Mineral fertilizers are fast-acting and have a precisely defined composition, which makes it easier to adjust the dose of nutrients delivered. On the other hand, they are costly and, if used unwisely, can pose a significant threat to the environment [16,17] or arable products [18]. In particular, the effect of nitrogen mineral fertilizers on soil pH poses a risk of acidification [19–22] and a risk to elements' availability [23]. Manure fertilizers have a low nutrient content and must be applied in large doses (the classic dose of cow farmyard manure is 40 t ha−<sup>1</sup> in Czech Republic). The nutrients contained in manure are released gradually, depending on the origin [24] and C:N ratio. Manures with a low C:N ratio (slurries) release nutrients to a greater extent already in the first year of application; on the other hand, manures with a high C:N ratio

**Citation:** Hlisnikovský, L.; Menšík, L.; Cermák, P.; Kˇ ˇ rížová, K.; Kunzová, E. Long-Term Effect of Pig Slurry and Mineral Fertilizer Additions on Soil Nutrient Content, Field Pea Grain and Straw Yield under Winter Wheat–Spring Barley–Field Pea Crop Rotation on Cambisol and Luvisol. *Land* **2022**, *11*, 187. https://doi.org/ 10.3390/land11020187

Academic Editors: Guido Wyseure, Julián Cuevas González, Jean Poesen and Angelinus Franke

Received: 3 December 2021 Accepted: 21 January 2022 Published: 25 January 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

(farmyard manure) release nutrients over a longer period in smaller doses [25]. According to [26], approximately 11% of the organic N is mineralized during the first year from the application of composted manure and around 20% for non-composted manure. In the case of slurries, approximately 40% of the organic N is mineralized at the same time [25]. The application of organic manures is usually associated with positive effects on soil properties [7,8,27–29], but one has to be careful about the doses and dry matter content. In the case of slurry, the dry matter and nutrient content is very important information in order to correctly adjust the applied dose. Ignorance of this information can easily lead to overdosing, which can significantly damage the crops grown or adversely affect the environment via leaching and volatilization of nitrogen and salinization [30–33]. In addition to directly supplied nutrients, the unifying factor for the positive effects of organic manures and the nutrient content, pH value, physical and biological properties of soils, is organic matter. Soil organic carbon and nutrient content are usually higher after application of solid organic manures [34–36], while the benefit of liquid organic manures, such as slurries, is mainly to increase the nutrient content and the effect on soil organic carbon can be neutral (no changes) [37] or positive [38,39] as the liquid manures contain a lower amount of organic carbon than the solid. As in the case of mineral fertilizers, organic manures can also pose a threat to the environment if not applied judiciously [40,41] or because of the presence of pharmaceuticals [42].

One of the major problems of agriculture in Czech Republic is the disruption between crop and livestock production, reduction of cultivated crops in crop rotations and the fact that most of the arable land is rented [43]. Disruption of the balance between crop and livestock production is manifested by a lack of organic manure and reduced input of organic carbon into the soil. Together with the significant dependence of crop production on mineral nitrogen (and the low level of phosphate and potassium fertilizers applied), we then experience soil erosion (lack of organic carbon), lower content of macronutrients (doses of P and K mineral fertilizers) and soil acidification (due to nitrogen fertilizers application) [44,45]. One way to reduce the negative impact of mineral nitrogen fertilizers on the soil while ensuring good soil nutrient supply and crop yields is to apply mineral and manure fertilizers together. Multiple scientific papers have indicated that joint application of mineral fertilizers and manure has a positive effect on both crop yields and reduction of negative impacts of mineral fertilizers on the soil properties [46–50]. Another problem of Czech agriculture is the reduction of crops in crop rotation. Over the years, there has been a change in the proportion of crops grown, mainly in favor of winter rape. Soil-improving crops such as root crops, forage crops and legumes are grown to a lesser extent than in the past [51]. While root crops (potatoes and sugar beet) are considered as soil-improving plants mainly due to the manure applied to them, legumes have a unique ability to fix airborne nitrogen in the soil, due to their symbiosis with rhizobacteria. Field pea (FP) is the most cultivated legume in Czech Republic (79% of all legumes), yet its representation in the crop rotations of Czech Republic is low (1.2%) (average values from 2015 to 2019 [52]). From the point of view of human nutrition and soil care, it is a valuable crop. Thanks to their symbiosis with rhizobia bacteria, legumes and FP cover a large part of their nitrogen needs from the symbiosis (depending on the type of legumes, they cover their nitrogen requirement from the soil from approximately 15-30%) and leave nitrogen in the soil for use by subsequent crops [53,54]. Although FP can use nitrogen from symbiosis with rhizobia bacteria, fertilizer application significantly affects its yield and quality. Foliar application of phosphorus can significantly improve yield and quality parameters of FP, especially on soils with low phosphorus content [55], but even on soils rich in P content, it has a positive effect on FP yields [56]. N fertilization can also increase yield and quality. Early application of N fertilizers is important, as the actual fixation of airborne nitrogen takes place only in the later stages of growth. Depending on soil and climate conditions, optimum N rates can range from 40 to 80 kg ha−<sup>1</sup> N and higher doses can provide lower yields as high N doses can reduce bacteria nodule mass [57]. However, under different soil and climate conditions

the response of FP to N fertilization may be different as yields can increase up to the dose of 135 kg ha−<sup>1</sup> N [58].

In 1972, long-term experiments were set up at two sites with different soil and climatic conditions to study the effect of the application of organic manure (pig slurries), mineral fertilizers and their combinations on soil chemistry and yields of wheat, barley and peas. The design of this experiment allows us to analyze the long-term effect of different fertilizer combinations on soil properties, which is currently a hot topic due to the dependence of conventional agriculture on mineral nitrogen, the low rates of applied P and K fertilizers and the limited availability of organic manures (slurries). In other words, our experiment can provide answers on how to take better care of the soil with the help of organic manure and how to avoid undesirable effects of mineral nitrogen applied without organic manure (current situation in Czech Republic). Soil types are represented by Cambisol (about 45% of the soil in Czech Republic) and Luvisol (about 13% of the soil in Czech Republic), representing the two most widespread soil types in Czech Republic. The article includes an analysis of the effect of fertilization on pea yields in 2017 and 2020 in Pernolec and the determination of a reasonable dose (using a linear-plateau model) of mineral nitrogen fertilization.

#### **2. Materials and Methods**

#### *2.1. General Information and Sites Description*

The results come from two long-term field trials located at Pernolec and Kostelec, Czech Republic, Central Europe. Both trials were established in 1972. The long-term trials aim to analyze the effect of mineral fertilizers (mineral nitrogen–N, phosphorus–P and potassium–K), pig slurry (three different doses), and their combination on the yield of arable crops. The crop rotation of both trials consists of winter wheat (*Triticum aestivum* L., WW), spring barley (*Hordeum vulgare* L., SB) and field pea (*Pisum sativum* L., FP). At the same time, the effect of long-term fertilizer application on basic soil properties is monitored (pH, the concentration of soil P, K, magnesium–Mg, calcium–Ca, the content of soil organic carbon–Cox, the content of soil total N–Ntot). In this paper, we assessed the period from 2015 to 2020 (six years) to analyze how long-term regular application of mineral fertilizers and organic manures affects soil properties and FP yields (yields from the years 2017 and 2020; 2015–WW, 2016–SB, 2017–FP, 2018–WW, 2019–SB, 2020–FP).

According to Köpper–Geiger climate classification [59], both sites are located in warm summer humid continental climate (Dfb). The basic site description of both localities is shown in Table 1. Detailed weather information can be found in Section 2.3. It should be noted here that our team was not the team that established the experiments in 1972 and we have not been able to find the results of soil analyses from the period of the trial establishment.


**Table 1.** The description of trial sites—Pernolec and Kostelec.

Note: the long-term average precipitation and temperature for Pernolec are based on the data from the years 1977–2014 (37 years) and for Kostelec from the years 1982–2014 (32 years).

#### *2.2. Field Trials Description*

In both long-term trials, the effect of a total of fifteen fertilization treatments with four replications has been running since 1972. The trial consists of sixty plots (15 × 4) arranged in a completely randomized block design. The plot size is 8 m × 5.5 m (44 m2). The fertilization treatments are identical in both trials, but the fertilization rates differ slightly (Tables 2 and 3 show fertilizer treatments and rates applied to FP in Pernolec and Kostelec. Tables S1 and S2 show the fertilizer treatments and rates applied over the whole three-year crop rotation—the sum of nutrients applied to all three crops over the three years). Mineral N was applied in two forms. Ammonium sulfate (AS) was applied in the spring, before the planting. Calcium ammonium nitrate (CAN) was applied during the beginning of stem elongation (BBCH 30). Mineral P was applied as superphosphate and mineral K as potassium sulfate. Both mineral fertilizers and PS were applied in the autumn, before tillering. Mineral fertilizers were spread by hand at both sites. Pig slurries were applied by manual sprayer. The average content of dry matter (DM) ranged from 0.68% (Pernolec) to 1.8% (Kostelec). This is a very low value: the amount of DM in slurry usually ranges from 0.7% to 24% [61] and is significantly affected by the season of the year [62]. Quality slurry is considered to have a dry matter content between 6% and 8%. The average pH and concentrations of N, P, K, Ca and Mg (% of DM) in Pernolec were 7.75, 1.79%, 0.52%, 16.77%, 1.11%, and 0.73%, respectively. In Kostelec, the average pH value of PS was 7.68 and the concentrations of N, P, K, Ca and Mg (% of DM) were 1.95%, 1.53%, 14.53%, 3.84% and 0.98%, respectively. Pig slurries were obtained from the nearest livestock farms that were able to supply manure in time. The FP (cul. Eso) was sown at the beginning of April (one million germinating seeds per ha, approximately 270 kg) and harvested in the first half of August.


**Table 2.** Forms and doses of mineral fertilizers and pig slurry (PS) according to the fertilizer treatments applied to FP in Pernolec.

Note: AS–ammonium sulphate; CAN–calcium ammonium nitrate; N (kg ha−1) in the PS column represents the content of N applied in PS.

#### *2.3. Weather Information*

Weather data (average monthly temperatures and monthly precipitation) were evaluated according to [63], which describes the World Meteorological Organization's recommendations for describing meteorological and climatological conditions of a defined period (text in Czech, tables in English). The weather analysis was based on long-term records. In Pernolec we compared the years 2017 and 2020 with the period from 1977 to 2014 (37 years). In Kostelec, we based our analysis on the period from the years 1982 to 2014 (32 years).

In Pernolec, the year 2017 was evaluated as warm (+0.9 ◦C in comparison with long– term average) and 2020 as very warm (+1.5 ◦C). In terms of precipitation, 2017 was a very dry year (76% of the long–term average), while 2020 was normal (98%). In Kostelec, the year 2017 was evaluated as warm (+0.8 ◦C) and 2020 as very warm (+1.5 ◦C). In terms of precipitation, 2017 was a normal year (109%), while 2020 was a very wet year (135%, Table 1). Detailed weather information for 2017 and 2020 at both sites, including assessments, is provided in Tables S3 and S4.

**Table 3.** Forms and doses of mineral fertilizers and pig slurry (PS) according to the fertilizer treatments applied to FP in Kostelec.


Note: AS–ammonium sulphate; CAN–calcium ammonium nitrate; N (kg ha−1) in the PS column represents the content of N applied in PS.

#### *2.4. Soil Analyses*

Following the harvest of the crops, soil samples were taken using the stainless-steel soil probe sampler. The soil samples were taken from the topsoil layer (0–20 cm). Four samples from each plot were taken. The samples were then mixed and transported to the laboratory, where they were dried and sieved to get fine and dry soil. The soil pH was analyzed potentiometrically using 0.2 mol KCl (inoLab pH 730, WTW, Xylem Analytics, Weilheim, Germany). The concentration of total N (Ntot) was analyzed using sulfuric acid in the heating block (Tecator, Foss Analytics, Hillerød, Denmark), followed by the Kjeldahl method [64]. The concentrations of P, K, Ca and Mg were analyzed using the Mehlich III solution [65], followed by the ICP-OES analysis (Thermo Scientific iCAP 7400 Duo, Thermo Fisher Scientific, Cambridge, UK). The SOC content was analyzed colorimetrically and via oxidimetric titration according to [66,67].

#### *2.5. Data Analyses*

One-way and multivariate analysis of variance (ANOVA, MANOVA) was used to compare the results of pH and soil element concentrations as affected by fertilization treatments and to analyze the effect of weather and fertilization treatments on FB yields (Pernolec locality only). Due to the occurrence of certain problems, we have only the summed FP yield values (average values without repeats) from the Kostelec site. For this reason, it was not possible to perform statistical analysis as in the case of the Pernolec. However, the average FP yield values from the Kostelec were suitable for PCA. FP yields from the Kostelec site are shown in Table S5. In this article, we have analyzed a total of fifteen fertilization treatments. Such a large set makes the interpretation of the results difficult and ambiguous (the results of the post hoc analysis overlap widely). For this reason, we proceeded to group the treatments (Control, PK, NPK, PS, PS+PK, PS + NPK) and calculate separate ANOVA for soil parameters where significant differences between the fertilizer treatments were recorded previously. If statistically significant differences were found, we used Tukey's HSD post hoc analysis to separate treatments. Statistical analyses were performed in Statistica 13.3. (Tibco Software Inc. Palo Alto, CA, USA). The nitrogen use efficiency (NUE) was calculated as ((GYT-GYC)/N rate) where GYT represents grain yield from the particular fertilizer treatment and GYC represents the grain yield from the Control treatment. The NUE was calculated from seven fertilizer treatments (NPK, PS1, PS1+NPK, PS2, PS2+NPK, PS3, and PS3+NPK. To evaluate the relationships between the yields, fertilizer treatments and soil parameters, principal component analysis (PCA) and factor analysis (FA) were used (Statistica 14.0.). MS Excel 2019 was used for weather analyses (Microsoft Corporation, Washington, DC, USA). The linear-plateau model, analyzing the reasonable N dose for FP (calculated from mineral fertilizer treatments), was calculated using R software (R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria, 2020), together with three R packages [68–70].

#### **3. Results**

#### *3.1. Comparison of Localities*

The two sites are statistically significantly different from each other in all observed soil parameters (results from all fertilizer treatments and for the whole period 2015–2020, Table 6). Compared to Kostelec, the soil in Pernolec is characterized by a higher pH value, and lower mean content of available P, K and Ca. In contrast, the average content of Mg, Cox and Ntot is higher in Pernolec.

#### *3.2. The Effect of Fertilizer Treatments on Soil Chemical Properties*

In the following sections the results of the effect of fertilization on pH, nutrient concentration, Cox and Ntot at each site will be presented. A summary description of the relationships between the fertilization treatments and the individual parameters is given in the last section, in which the PCA results are presented.

#### 3.2.1. Soil Reaction

In Pernolec (Cambisol), the soil pH was statistically significantly affected by the fertilizer treatment (d.f. = 14; F = 10.6; p < 0.001). Comparing all 15 treatments, the lowest mean pH value (4.73) was recorded in the NPK E3 treatment. The highest mean pH value (5.93) was recorded in PK treatment (Table 4). Comparing the groups of fertilizers, the lowest mean pH was recorded in NPK treatments (5.16), followed by PS+NPK (5.48), while the highest pH was recorded in PS+PK (5.81) and PK (5.93) treatments (Figure 1a).

In Kostelec (Luvisol), the value of the pH was also significantly affected by the fertilizer treatment (d.f. = 14; F = 4.2; *p* < 0.001). The lowest mean pH value was recorded in PS2 + NPK (5.04) treatments, while the highest was in Control (5.75) and PK (5.67) treatments (Table 5). Comparing the groups of fertilizer treatments, similarly to Pernolec, the lowest mean pH values were recorded in treatments with mineral N–NPK (5.33) and PS+NPK (5.14), while the highest value was recorded in Control (5.75) treatment (Figure 1b).

The results show that the application of NPK, either alone or in combination with PS, results in the lowest pH values. In Kostelec, the pH values for the NPK and PS+NPK treatments were comparable and significantly different from the other treatments. In Pernolec, the effect of NPK was most significant, while the combined application of PS+NPK was comparable to PS, yet lower. The negative effect of ammonium nitrogen on pH is particularly noticeable when compared to the PK treatment (Figure 1).

#### 3.2.2. Phosphorus

The concentration of P in the soil was not affected by the long-term application of slurry and mineral fertilizers in Pernolec (d.f. = 14; F = 0.6; *p* = 0.84). The lowest mean concentration was recorded in Control (58 mg kg−1), and the highest in PS3+PK treatment (111 mg kg−1) (Table 4). A different situation occurred in Kostelec, where differences between fertilizer treatments were significant (d.f. = 14; F = 16.47; *p* < 0.001). As in Pernolec, the lowest concentration was recorded in Control (124 mg kg−1), and the highest in PS3+NPK (262 mg kg−1) treatment (Table 5). Comparing the groups of fertilizers, ANOVA separated three groups of fertilizers according to their effect on soil P concentration in Kostelec (Figure 2a). The lowest mean concentration was recorded in Control (124 mg kg−1), followed by NPK, PK and PS treatments. The combined application of PS+NPK and PS+PK resulted in the highest mean P concentrations, ranging from 229 to 235 mg kg−<sup>1</sup> (Figure 2a).


**Table 4.** Soil pH value, the concentration of P, K, Ca and Mg (mg kg<sup>−</sup>1), the content of organic carbon (Cox, %) and total nitrogen (Nt, %) as affected by the fertilizer treatments (2015–2020) in Pernolec.

Mean values (±SE) followed by the same letter (a vertical comparison of the effect of fertilizer treatment) are not statistically significantly different. Columns without letters (P, Mg, Cox, Nt) represent values without statistically significant differences, where the effect of fertilizer treatments was insignificant.

**Figure 1.** The effect of fertilizer treatments on pH in (**a**) Pernolec and (**b**) Kostelec (2015–2020). Mean values followed by the same letter are not significantly different. Red triangles represent the raw data. The blue lines represent the mean value of the particular treatment, while the red line represents the mean value calculated from all fertilizer treatments.


**Table 5.** Soil pH value, the concentration of P, K, Ca and Mg (mg kg<sup>−</sup>1), the content of organic carbon (Cox, %) and total nitrogen (Nt, %) as affected by the fertilizer treatments (2015–2020) in Kostelec.

Mean values (±SE) followed by the same letter (a vertical comparison of the effect of fertilizer treatment) are not statistically significantly different. Columns without letters (P, Mg, Cox, Nt) represent values without statistically significant differences, where the effect of fertilizer treatments was insignificant.

**Table 6.** Average values of soil parameters in Pernolec and Kostelec. The values are based on the results of soil analyses of all fertilization treatments and all analyzed years (2015–2020).


Note: F: F statistic; d.f.: degree of freedom; *p:* level of significance. Mean values ± standard error of the mean (SE) followed by the same letter are not significantly different.

#### 3.2.3. Potassium

Of all the analyzed parameters, potassium was the element most affected by the fertilization treatments (d.f. = 14; F = 22.83; *p* < 0.001 for Pernolec and d.f. = 14; F = 19.11; *p* < 0.001 for Kostelec). In Pernolec, the lowest concentration was recorded in NPK E1 treatment (116 mg kg−1), and the highest in PS1+NPK treatment (215 mg kg−1) (Table 4). In Kostelec, the K mean concentration varied from 112 mg kg−<sup>1</sup> (NPK E1) to 252 mg kg−<sup>1</sup> (PS3+PK) (Table 5). If we compare the fertilizer groups, we find that both localities have a comparable pattern. In Pernolec, application of no fertilizers (Control), NPK and PS resulted in lower K soil concentrations without differences between these treatments, while application of PK, PS+NPK and PS+PK resulted in higher K concentrations (Figure 3a). The situation in Kostelec was similar, with one exception, namely for the PK and PS treatments. The differences between these two treatments were not significant, as in Pernolec (Figure 3b).

#### 3.2.4. Calcium

In Pernolec, the mean Ca soil concentrations varied significantly (d.f. = 14; F = 4.83; *p* < 0.001) between the treatments and ranged from 1113 mg kg−<sup>1</sup> (NPK E3) to 1447 mg kg−<sup>1</sup> (PS1) (Table 4). Similarly, in Kostelec, the differences between fertilization treatments were significant (d.f. = 14; F = 4.46; *p* < 0.001), and varied from 1213 mg kg−<sup>1</sup> (PS2+NPK) to 1549 mg kg−<sup>1</sup> (Control) (Table 5). Comparing the fertilizer groups, we find that the

effect of fertilization on soil Ca content is similar in the two sites. The lowest mean Ca concentrations were recorded in NPK and PS+NPK treatments (Figure 4a,b), while application of no fertilizers (Control) and PS resulted in the highest Ca concentrations.

**Figure 2.** The effect of fertilizer treatments on (**a**) soil P and (**b**) Mg concentration in Kostelec (2015–2020). The differences between P and Mg concentrations as affected by fertilizer treatment were insignificant in Pernolec (Table 6). Mean values followed by the same letter are not significantly different. Red triangles represent the raw data. The blue lines represent the mean value of the particular treatment, while the red line represents the mean value calculated from all fertilizer treatments.

**Figure 3.** The effect of fertilizer treatments on soil K concentration in (**a**) Pernolec and (**b**) Kostelec (2015–2020). Mean values followed by the same letter are not significantly different. Red triangles represent the raw data. The blue lines represent the mean value of the particular treatment, while the red line represents the mean value calculated from all fertilizer treatments.

#### 3.2.5. Magnesium

Average soil Mg concentrations in Pernolec were not significantly affected by the fertilization treatments (d.f. = 14; F = 0.57; *p* = 0.88) and ranged from 104 mg kg−<sup>1</sup> (NPK E3) to 131 mg kg−<sup>1</sup> (PS1) (Table 4). In Kostelec, on the other hand, the long-term application of slurry and mineral fertilizers had a significant effect on the Mg concentration (d.f. = 14; F = 7.10; *p* = 0.001), which varied from 58 mg kg−<sup>1</sup> (NPK) to 83 mg kg−<sup>1</sup> (PS3) (Table 5). Comparing the groups of fertilizers, the lowest mean concentrations were recorded in NPK

and PK treatments, while the highest concentrations occurred in PS and PS+PK treatments (Figure 2b).

#### 3.2.6. Soil Organic Carbon Content

Long-term and regular application of slurry, mineral fertilizers and their combinations did not significantly affect the soil organic carbon content in either Pernolec (d.f. = 14; F = 0.91; *p* = 0.56) or Kostelec (d.f. = 14; F = 0.77; *p* = 0.70). In Pernolec, the Cox content in the soil varied from 0.88% (PK, PS1) to 1.02% (PS3+NPK) (Table 4). In Kostelec, the Cox ranged from 0.77% (PK) to 0.87% (PS3+NPK) (Table 5).

#### 3.2.7. Total Nitrogen Content

Similar to soil organic carbon, long-term and regular application of manure, mineral fertilizers and their combinations did not significantly affect total soil nitrogen content at either of the two sites (Pernolec: d.f. = 14; F = 0.52; *p* = 0.91; Kostelec: d.f. = 14; F = 0.64; *p* = 0.83). In Pernolec, the Ntot content ranged from 0.11% to 0.13% (Table 4), in Kostelec from 0.10% to 0.11% (Table 5).

#### 3.2.8. Principal Component Analysis (PCA)

Based on the PCA results (Figure 5a), we can classify the fertilizers in Pernolec (cambisol) into four categories according to their effect on yield and soil properties (Figure 5b). (1) The unfertilized treatment (Control) gives lower crop yields and has low P and K concentrations due to no external supply of nutrients. (2) Pig slurry (PS) applied alone, application of mineral P and K (PK), and combination of PS+PK (generally the fertilizers without mineral N): these fertilizers have a positive relationship with pH and Ca and Mg content, and there is no decrease in pH compared to other treatments. On the other hand, the absence of mineral N puts this group at a disadvantage in terms of low grain and straw yields and the soils have a low organic matter content (no organic matter in the PK treatment and low organic matter in the slurry). (3) The third group is represented by PS+NPK treatments. The joint application of PS and mineral NPK represents a kind of golden mean ensuring relatively high grain and straw yields, nutrient and soil organic matter content. However, the presence of the ammonium form of mineral N negatively affects soil pH. (4) The fourth group consists of separately applied mineral fertilizers (NPK, without manure supplement). Mineral fertilizers are clearly closely and positively associated with yield, followed by soil organic carbon and nitrogen. On the other hand, the presence of the ammonium form of

nitrogen, accompanied by the absence of slurry, accentuates the negative effect on pH even more significantly (compared to PS+NPK combinations).

**Figure 5.** Results of the PCA*—*relationships between soil chemical parameters and grain and straw yields as affected by the fertilization treatments: (**a**,**b**) Pernolec, (**c**,**d**) Kostelec. Grain and straw yields are based on the average WW, SB and FP yield from 2015 to 2020.

With a change in soil type (Kostelec, luvisol), we can see a different response to the long-term application of manure and mineral fertilizers on yield and soil properties (Figure 5c). The separation of fertilizers in Kostelec (Figure 5d) is not as clear-cut as in Pernolec, which means that the differences between fertilizer treatments are not as pronounced. As in Pernolec, unfertilized Control is strongly and positively correlated with pH and soil Ca content. On the other hand, treatment without external nutrient inputs (Control) is associated with low grain and straw yields and also with low concentrations of soil P and K (soil depletion). The PK group (mineral P and K fertilizers) has a completely different status than PS (in Pernolec these two fertilizer groups were together in one cluster). PK has a strong negative relationship with soil organic carbon and total nitrogen. This treatment highlights the need for nitrogen, either supplied in mineral form or the form of manure. In contrast, the application of pig slurry (PS) is strongly and positively associated with soil organic carbon content combined with a neutral relationship to both yield and pH. Mineral fertilizers (NPK) occupy a similar position to PS in terms of yield and pH, with the exception that they are closer to higher yields and lower pH. Quite different (compared to Pernolec) is their relationship to soil organic carbon, with which it is moderately and rather

negatively correlated. Similarly to Pernolec, the PS+NPK fertilizer group is dominant. It is associated with high yields, high soil nutrient content, a relatively neutral relationship to soil organic matter and a significantly (strongly) negative relationship to soil pH (stronger negative relationship to pH than in Pernolec).

#### 3.2.9. FP Grain and Straw Yields

As mentioned in Section 2.5, the results of the FP grain and straw yields from Kostelec cannot be statistically analysed. The average grain and straw yields in 2017 and 2020 are shown in Table S5. In 2017, the grain yields varied from 2.8 t ha−<sup>1</sup> (PS1, PS1+PK) to 3.7 t ha−<sup>1</sup> (PS3+NPK), while in 2020 the grain yields varied from 2.5 t ha−<sup>1</sup> (NPK E2) to 3.3 t ha−<sup>1</sup> (PS1+PK). Straw yields varied from 2.1 t ha−<sup>1</sup> (PS1) to 3.2 t ha−<sup>1</sup> (PS3+NPK) in 2017 and from 3.0 t ha−<sup>1</sup> (PS1+NPK, NPK E1) to 3.6 t ha−<sup>1</sup> (PS2) (Table S5).

According to MANOVA results, the FP grain yields were significantly affected by year (d.f. = 1; F = 71.55; p < 0.001), fertilizer treatment (d.f. = 14; F = 6.76; *p* < 0.001), and their interaction (d.f. = 14; F = 2.33; *p* < 0.01) in Pernolec. The effect of year was dominant (89%), while the effect of fertilizer treatment influenced yields by 8%. If we look at the weather in a particular year, we find that 2017 in Pernolec was marked by drought in May and June. Moreover, 2017 was significantly marked by very high temperatures in June and July (Table S3). These were factors that caused significantly lower yields compared to 2020, which was characterized by both higher precipitation and milder temperatures (Table 1). Straw yields were comparable in 2017 and 2020 as the differences were insignificant in Pernolec (d.f. = 1; F = 0.40; *p* = 0.53), while the effect of the fertilizer treatment was significant (d.f. = 14; F = 4.32; *p* < 0.001). The interaction between the factors of year and treatment was insignificant (d.f. = 14; F = 1.90; *p* < 0.07).

In 2017, the grain yields were significantly affected by the fertilization (d.f. = 14; F = 3.18; *p* < 0.01) and varied from 1.2 t ha−<sup>1</sup> (Control) to 2.3 t ha−<sup>1</sup> (PS3+PK and NPK E3) (Table 7). Significantly different were Control and PS1+PK against PS3+PK, PS3+NPK and NPK E3. Grain yield slightly increased with increasing nitrogen rate (Figure 6a). According to the linear-plateau model, calculated from the mineral fertilizer treatments (NPK, NPK E1, NPK E2, NPK E3), the FP yield response to different rates of mineral N plateaued at 97 kg ha−<sup>1</sup> N, with a corresponding yield of 2.08 t ha−<sup>1</sup> (Figure 7, left). Comparing the nitrogen use efficiency (NUE), the highest NUE was recorded in NPK treatment (23.3 kg per 1 kg of N applied), followed by PS1 (5.9 kg), PS1+NPK (4.3 kg), PS2, PS3 and PS3+NPK (3.5 kg), the lowest NUE was recorded in PS2+NPK treatment (3.0 kg per 1 kg of N applied). This calculation shows that mineral fertilizers, compared to organic manures (slurries), supply nutrients very quickly and, even in small quantities can significantly and efficiently promote growth. On the other hand, their effectiveness is offset by their negative effect on the soil environment.

In 2020, the grain yields varied from 1.8 t ha−<sup>1</sup> (PS1 and PS1+PK) to 2.8 t ha−<sup>1</sup> (NPK E3). As we can see, the response to the fertilization was a little different as the weather conditions changed (Figure 6b, the red line representing a quadratic model). We can see that grain yield slightly increased with increasing N dose, as in 2017. The course of the function indicates the attainment of a local maximum, which, according to the quadratic model, is located at an N rate of 400 kg ha−1. At this rate, the maximal average yield of 2.4 t ha−<sup>1</sup> would be achieved, which is actually lower than the yields already obtained with lower inputs (Figure 6b). According to the linear-plateau model, the response of FP yields to different rates of N doses plateaued at 73 kg ha−1, corresponding with the yield 2.71 t ha−<sup>1</sup> (Figure 7, right), showing better weather conditions for yield development in 2020. Comparing the NUE, the highest efficiency was again recorded in NPK treatment (10.0 kg per 1 kg N applied), followed by PS2+NPK (3 kg), PS1+NPK (2.6 kg), PS3+NPK (2.4 kg), PS3 (0.4 kg) and PS1 and PS2 (−2.4 and −0.5, respectively), where the efficiency was negative as the mean yield was lower than in the Control treatment.


**Table 7.** The effect of the year (2017, 2020) and fertilizer treatment on FP grain and straw yield (t ha<sup>−</sup>1) in Pernolec.

Mean values (±SE) followed by the same letter (a vertical comparison of the effect of fertilizer treatment) are not statistically significantly different.

**Figure 6.** FP grain yield (t ha−1) as affected by N dose in Pernolec in (**a**) 2017 and (**b**) 2020. The average yields (blue points) are interleaved with the quadratic function (red line). The equation of the quadratic model is given above the figure.

Comparing the results from both years (Table 7), we find that the highest average yields were obtained with the NPK E3 treatment (2.5 t ha−1). However, lower, but statistically comparable, yields were obtained with the NPK (30 kg mineral N ha−<sup>1</sup> with an average yield of 2.1 t ha<sup>−</sup>1) and PS3 (51 t ha−<sup>1</sup> with an average yield of 2.1 t ha<sup>−</sup>1) treatments. This is a very important finding as PS applied in higher doses can completely replace mineral fertilizers and a negative effect of mineral fertilizers on soil pH can be partially avoided.

**Figure 7.** The response of FP yields to different doses of mineral N fertilizers (NPK, NPK E1, NPK E2, NPK E3 treatments) in 2017 (**left**) and 2020 (**right**). Yields (black dots) are interleaved with the linear-plateau model (blue line). The equation of the model is given above the figure.

Straw yields were significantly affected by fertilization in 2020 (d.f. = 14; F = 3.26; *p* < 0.05), with insignificant differences in 2017 (d.f. = 14; F = 3.07; *p* < 0.05, Tukey's test did not confirm ANOVA as multiple comparison methods generally have lower test power than analysis of variance*-*ANOVA). Straw yield tended to increase with increasing fertilizer rate. The differences between 2017 and 2020 were insignificant. The highest yields obtained were recorded for the PS3+NPK and NPK E3 treatments; however, the PK, PS1 and PS2 treatments also provided statistically comparable yields (Table 7).

#### **4. Discussion**

Long-term and regular application of mineral fertilizers, pig slurry, and their combinations significantly affected soil properties and the effect of fertilizers depends on soil conditions (type) of the site. One of the most important soil properties is the value of the pH. Soil pH is considered to be the dominant factor directly influencing other soil properties such as elements' availability [10,71,72] and abundance and representation of plant and microbial communities [73] and their activity [74]. All macronutrients are best available in neutral to alkaline soils, while in acid soils their availability decreases and the availability of elements such as Fe, Mn, B, Zn and Al increases. Changes in pH thus directly affect the soil's ability to supply nutrients to plants. In our case, the lowest pH values were recorded for the NPK treatments (applied alone or in combination with PS, but only in the NPK treatments with the highest N doses, Tables 4 and 5). The same result was recorded worldwide [11,13,21,72,75] and has been known for a long time [76]. The primary driver of downward pH changes is mineral nitrogen, in its ammonium form, because the conversion of the ammonium form to nitrate in soils releases hydrogen, directly affecting its concentration in the soil environment. This can be particularly evident in the case of PK treatments. As mentioned above, Czech conventional crop production is primarily dependent on mineral nitrogen. Add to this the fact that most of the cultivated land is rented and its owners have no idea or do not care about acidification. This leaves room for acidification to run freely. An interesting survey was carried out in the USA, which also shows that acidification is taking place there and that about half of the farmers were not even aware of it [77]. One way to reduce the negative effects of mineral fertilizers on soil pH is to combine mineral fertilizers and organic manures [78]. Co-application of mineral fertilizers and organic manures is often cited as a sustainable method of fertilization, providing high and stable yields and a healthy state of the soil. The unifying element of this approach is organic matter (together with nutrients) [3,5,7,8,13,29,79,80] added to the soil, beneficially affecting soil chemical, physical and microbiological properties. From this point of view, we can support these results only partially as the combined application of PS+NPK provided better pH values than NPK only in Pernolec (Figure 1a), in contrast to Kostelec (lower

and comparable to NPK treatment, Figure 1b). This may be due to the overall higher soil organic matter content in Pernolec (Table 1) and the very low organic matter content in the slurry, which seems to be behind the non-significant Cox differences between fertilization treatments in both locations (Tables 4 and 5). The DM of pig slurry usually ranges from 0.7% up to 23% [61] and quality slurry has a dry matter content between 6% and 8% in Czech Republic. In our case, the dry matter content of the available and applied pig slurry was very low, which is probably the reason why the soil organic matter content is slightly higher in the high slurry fertilizer treatments, but not statistically significantly higher compared to the other fertilization treatments.

From the point of view of nutrients, the highest concentrations of macronutrients were always connected with PS+PK, NPK and PS+NPK treatments (Figure 5), while nutrient depletion can be found in Control treatment. PS+PK treatment has a close relationship to nutrient content and a moderate relationship to yields (Figure 5), showing that nitrogen is a limiting element in this treatment and its P and K nutrients are not utilized completely. The combination of mineral fertilizers and organic manures provides high yields while leaving a high micronutrient content in the soil (Figure 5). From the point of view of agriculture in Czech Republic, we can expect that acidification problems will intensify, as mineral nitrogen is important for all agricultural crops and significantly affects yields, which is the most monitored parameter. The application of mineral fertilizers at higher doses (NPK E3, PS2+NPK, PS3+NPK treatments) significantly reduced the soil reaction values at both sites (Kostelec and Pernolec) compared to the Control; a more significant decrease was recorded on the luvisol soil type (Kostelec). Similar findings (decrease in pH in treatments fertilized with mineral fertilizers only) are supported by some other studies [81–84]. The negative effect of acidification on the content of available nutrients (Ca, Mg) in the plough soil horizon is shown in Tables 4 and 5 (in the NPK E3, PS2+NPK, PS3+NPK treatments, low Ca and Mg contents were recorded at both sites). For available nutrients P and K, the acidification effect was predominant in the mineral fertilized treatments (NPK E1-3). This is confirmed by the results of the multicriteria PCA evaluation. These results are in agreement with [85,86], which showed a negative effect of acidification on the regime of available nutrients in the soil. Without the addition of other nutrients (PK treatments), there will be a reduction in the content of these nutrients in the soil (as in the case of Control). The combination of mineral fertilizers and organic manures can partially reduce the negative effect of mineral fertilizers on pH (depending on the location and soil and climate conditions), which is good news, but the lack of organic manures due to reduced livestock production in the country plays against the solution to the current problems.

In terms of pea yields, we can clearly see the dependence of yields on nitrogen, with pea yields increasing with increasing nutrient rates, although the differences are not statistically significant between higher doses of fertilizers. The yields are strongly affected by fertilization and by weather conditions. While nutrient utilization is lower in years with poorer weather conditions, nutrient utilization increases in years with normal conditions. This can be seen in the results of the linear-plateau model, which compared nutrient and yield dependence in 2017 and 2020. Based on this model, we can say that under normal weather conditions the optimum nitrogen rate in Pernolec is around 70 kg ha<sup>−</sup>1. As the variation from normal conditions increases, the nutrient requirement increases as the optimal dose of N raised to 97 kg ha−<sup>1</sup> N in 2017. Another important finding is that mineral fertilizers can be completely replaced by PS applied in higher doses (51 t ha−<sup>1</sup> in our case). PS has a low C:N ratio, and the mineralization of slurries is rapid, providing a huge amount of available nutrients at the beginning of the season before symbiosis with mycorrhizal bacteria fully develops. Replacing mineral fertilizers with PS can provide comparable yields without a negative effect on soil pH value.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10.3 390/land11020187/s1, Table S1: Forms and doses of mineral fertilizers and pig slurry (PS) according to the fertilizer treatments applied in Pernolec. Cumulative doses for the entire three-year crop rotation. Table S2: Forms and doses of mineral fertilizers and pig slurry (PS) according to the fertilizer treatments

applied in Kostelec. Cumulative doses for the entire three-year crop rotation. Table S3: The longterm mean precipitation (Mean; 1977–2016 for Pernolec; 1982–2016 for Kostelec; mm) and the sum of precipitation (mm) in individual months in 2017 and 2020 in Pernolec and Kostelec. The comparison between long-term mean and actual (2017, 2020) precipitation was done according to [63]. Table S4: The long-term mean temperature (Mean; 1977–2016 for Pernolec; 1982–2016 for Kostelec; ◦C) and the average temperature (◦C) in individual months in 2017 and 2020 in Pernolec and Kostelec. The comparison between the long-term mean and actual (2017, 2020) temperature was done according to [63]. Table S5: The effect of the year (2017, 2020) and fertilizer treatment on FP grain and straw yield (t ha<sup>−</sup>1) in Kostelec.

**Author Contributions:** Conceptualization, L.H., L.M. and E.K.; methodology, E.K.; software, L.H, L.M. and K.K.; formal analysis, L.H., L.M. and K.K.; investigation, L.H., L.M. and K.K.; resources, E.K. and P.C. writing—original draft preparation, L.H.; writing—review and editing, L.H., L.M. and ˇ E.K.; visualization, L.H. and K.K.; supervision, E.K.; project administration, E.K.; funding acquisition, E.K. and P.C. All authors have read and agreed to the published version of the manuscript. ˇ

**Funding:** This research was funded by the Ministry of Agriculture of the Czech Republic, grant number RO0418, Czech National Agency for Agricultural Research, grant number QK1810010, QK21020155 and QK21010124, and EU project H2020 677407—Soil care for profitable and sustainable crop production in Europe.

**Institutional Review Board Statement:** Ethical review and approval were waived for this study, due to the non-use of experimental animals or human subjects.

**Informed Consent Statement:** No human trials were conducted during the study.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** We would like to thank the staff at the experimental station in Pernolec, Pavel Beneš and Vera Folejtárová, and the staff of the experiment in Kostelec nad Orlicí, Petr Iviˇcic, Vlatimil Basaˇr and Jaroslav Málek.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **Perceived Causes and Solutions to Soil Degradation in the UK and Norway**

**Niki Rust 1, Ole Erik Lunder 2, Sara Iversen 3, Steven Vella 4, Elizabeth A. Oughton 1, Tor Arvid Breland 2, Jayne H. Glass 5, Carly M. Maynard 5, Rob McMorran <sup>5</sup> and Mark S. Reed 6,\***


**Abstract:** Soil quality is declining in many parts of the world, with implications for the productivity, resilience and sustainability of agri-food systems. Research suggests multiple causes of soil degradation with no single solution and a divided stakeholder opinion on how to manage this problem. However, creating socially acceptable and effective policies to halt soil degradation requires engagement with a diverse range of stakeholders who possess different and complementary knowledge, experiences and perspectives. To understand how British and Norwegian agricultural stakeholders perceived the causes of and solutions to soil degradation, we used Q-methodology with 114 respondents, including farmers, scientists and agricultural advisers. For the UK, respondents thought the causes were due to loss of soil structure, soil erosion, compaction and loss of organic matter; the perceived solutions were to develop more collaborative research between researchers and farmers, invest in training, improve trust between farmers and regulatory agencies, and reduce soil compaction. In Norway, respondents thought soils were degrading due to soil erosion, monocultures and loss of soil structure; they believed the solutions were to reduce compaction, increase rotation and invest in agricultural training. There was an overarching theme related to industrialised agriculture being responsible for declining soil quality in both countries. We highlight potential areas for land use policy development in Norway and the UK, including multi-actor approaches that may improve the social acceptance of these policies. This study also illustrates how Q-methodology may be used to co-produce stakeholder-driven policy options to address land degradation.

**Keywords:** conservation agriculture; deliberative democracy; q-methodology; regenerative agriculture; soil conservation; sustainable land management

### **1. Introduction**

"Countries can withstand coups d'état, wars and conflict, even leaving the EU, but no country can withstand the loss of its soil and fertility". (Rt Hon Michael Gove, former Secretary of State for the Environment, speaking at the British parliamentary launch of the 'Sustainable Soils Alliance', October 2017).

The ground beneath our feet is not only a substrate upon which we traverse this earth but is also a vital component of our natural capital. Soils are the foundation of terrestrial

**Citation:** Rust, N.; Lunder, O.E.; Iversen, S.; Vella, S.; Oughton, E.A.; Breland, T.A.; Glass, J.H.; Maynard, C.M.; McMorran, R.; Reed, M.S. Perceived Causes and Solutions to Soil Degradation in the UK and Norway. *Land* **2022**, *11*, 131. https:// doi.org/10.3390/land11010131

Academic Editors: Guido Wyseure, Julián Cuevas González, Jean Poesen and Christine Fürst

Received: 24 November 2021 Accepted: 13 January 2022 Published: 14 January 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

food production, supporting directly or indirectly 95% of our food production [1]. Along with providing a substrate to grow our food, soils also confer other essential ecosystem services, such as water storage and filtration, nutrient cycling, biodiversity and carbon storage [2]. However, demand for food, increasing human populations and the effects of climate change are placing unprecedented pressures on soil. Over the last 70 years, the supply of global per capita food calories increased by about one-third, with the use of irrigation water roughly doubling and use of inorganic nitrogen fertiliser increasing nearly nine-fold [3]. At the same time, climate change has led to faster rates of warming on land than the global mean and altered precipitation patterns, which have contributed to altered growing seasons and regional crop yield reductions [4]. With rising human populations, coupled with increased individual wealth, it is expected that food demand will grow by as much as 70% by 2050; an estimated 46% of that demand needs to come from increasing food production [5]. This increase in food productivity must be achieved whilst significantly reducing greenhouse gas emissions from agriculture, if warming is to be restricted "well below" 2 ◦C, as proposed in the Paris Agreement [6]. How this is achieved without negatively impacting soils any further remains a challenge.

Soil quality <sup>1</sup> in many parts of the world is declining due to a combination of physical, chemical and biological degradation coupled with socio-economic drivers, reducing the soil's ability to undertake these important ecosystem functions [7]. Globally, 20–30 gigatons of soil are lost each year due to water erosion [7] and climate change is projected to increase erosion from water and reduce levels of soil organic carbon, especially in drylands [4]. There is thus an urgent need to develop and encourage widespread adoption of effective and profitable sustainable soil management practices [8,9]. This is articulated in Sustainable Development Goal 15, which aims to "protect, restore and promote sustainable use of terrestrial ecosystems, sustainably manage forests, combat desertification, and halt and reverse land degradation and halt biodiversity loss" [10], and its Land Degradation Neutrality target which aims to counterbalance expected losses with measures to achieve equivalent gains within the same type of land [11].

There are many competing methods to deal with agricultural soil degradation at different governance scales: from multilateral policies such as the United Nations Convention to Combat Desertification (UNCCD) and the proposed EU Soils Directive, to national and sub-national policies and measures designed to incentivise and regulate the management of soils. These policies typically seek changes in management at farm and field scales, for example through the adoption of soil-improving cropping systems and other sustainable land management technologies and approaches (e.g., WOCAT [12]). The lack of scalable policy options was cited in the UNCCD's Global Land Outlook [9] as a key barrier to more sustainable land management, but there are no easy solutions given the different social, cultural, economic, environmental and technological contexts in which policies and practices need to operate [8]. Again, the attractiveness and appropriateness of different options for policy and practices differ based on the subjective experience and contrasting knowledge and values of the people the policies are meant to serve.

Policies and practices that can tackle the multiple causes of declining soil quality are urgently needed and stakeholder engagement in the policy formulation process is crucial for this complex issue, given the subjective and value-laden nature of both the causes of and solutions to the challenge. Effectively representing diverse stakeholder perspectives in decision-making processes can lead to better informed, more durable, and flexible outcomes across a wide range of contexts (De Vente et al., 2016). Policies created through deliberative democracy can align better with social and cultural norms, resulting in increased trust and ownership of problems and solutions; together, this can lead to decisions that are more likely to be accepted and implemented, helping to achieve environmental goals more effectively [13,14].

As interest has grown in participatory approaches to policy making and other forms of deliberative democracy, methods have been sought to represent and integrate the range of perspectives, values and beliefs held by citizens in the policy-making process [15]. The application of Q-methodology to the co-production of policy options with stakeholders has been used by Rust [16] and Addams and Proops [17] as a form of deliberative democracy. These studies had the normative goal of representing more diverse perspectives in the policy-making process. They also had a pragmatic goal of improving the quality of decisions or range of policy options based on more comprehensive information inputs and/or improving the acceptability of policies based on deeper insights into the way publics conceptualise environmental issues.

In this study, we used Q-methodology to understand a wide range of stakeholder perspectives that could inform the design of socially acceptable options in agricultural soil management policy and practice. To address the lack of scalable policy options noted above by the UNCCD, this was done in the UK and Norway which are two countries experiencing similar types of soil degradation that are broadly representative of soil quality issues and climatic variation across northern temperate regions of Europe. The countries have contrasting agricultural policy environments, with Norway not being a part of the EU and the UK in the process of leaving the EU when the research was conducted. The countries also have quite different models of social democracy in land governance [18], which provides an opportunity to consider how the different land tenure regimes influence policy formulation and application. Both countries are interested in the co-development of policies to increase food production whilst reducing the environmental impacts of agriculture. By understanding where agricultural stakeholders agreed and disagreed over causes and solutions to declining soil quality in each country, we sought to highlight potential scalable options for land use policy development. This study also illustrates how Q-methodology may be used to co-produce stakeholder-driven policy options to address land degradation.

#### **2. Materials and Methods**

*2.1. Study Sites*

2.1.1. UK

Like much of the rest of Europe, the UK has a long history of unsustainable soil management practices, leading to a loss of soil structure and fertility. About 25% of the total land area of the UK is suitable for arable cropping, with an average farm size of 81 hectares [19]. Currently, soil erosion exceeds the rate of soil formation in many areas in the UK, with around 17% of arable land showing signs of erosion, although as much as 40% may be at risk of further degradation [20]. The cost of soil erosion to the UK has been estimated at £45 million a year, including £9 million in lost production [21]. UK soil is being lost at a rate ten times that which it is created [2], with dramatic economic implications.

A comparison of soil nutrient balances from the year 2000 to 2019 shows a 24% decrease for nitrogen and a 46% decrease for phosphate (in kg per hectare) [19]. Soil erosion, compaction and loss of organic matter are thought to cost arable farmers an average of £5584 per year [22] and English water companies spend £21 million a year on addressing soil erosion [23]. Improving soil management in the UK is therefore not only an environmental but also an economic imperative. Soil quality decline in the UK is more pronounced in arable regions due to the highly intensive practices used, such as monocropping, use of heavy machinery, overuse of chemical inputs and a lack of integration of organic material [21,24].

#### 2.1.2. Norway

Only 3.1% of the total land area in Norway is suitable for arable cropping, with an average farm size of 23.9 hectares in 2016; cereals can only be grown on one third of this area due to limiting natural conditions [25]. Although agricultural policies in Norway advocate multifunctional agriculture [26], regional agricultural specialisation, known as "kanaliseringspolitikken", was introduced in the after-war period, which led to increased agricultural production by incentivising cereal production in lower-lying areas [27]. In the last two decades, the total Norwegian cereal yield has declined due to a reduction in the area used for cropping [28]. Despite this decline, the Norwegian government has set a target of increasing food production by 20% by 2030 from 2010 levels, to meet projected population growth in Norway [29]. Three counties (Akershus, Østfold, and Hedmark) in southeastern Norway produce 60% of the country's cereal; however, soil organic matter (SOM) content has declined in the region, with an average loss of 1% of SOM a year from 1991 to 2001, which is not sustainable [30]. The underpinning governance and institutions (both formal and informal) are strongly communal in character [26]. The long history of collective land management, the regulation of the Norwegian land market and the selfimposed limits to farm scale are in contrast to the generally unregulated land market and existence of larger-scale farms in the UK.

#### *2.2. Research Design*

Q-methodology is a mixed-methods approach using interviews to explore participants' subjective understanding of a topic using Q sorts where respondents rate the extent to which they agree with statements, which are then analysed using by-person factor analysis, correlating people with others who hold similar opinions based on their Q-sorts. Q-methodology was chosen due to its capacity to shed light on complex, subjective phenomena where individuals hold differing views and values [30]. It allows for exploration of tensions in knowledge and perspectives between stakeholders that may affect the effectiveness and acceptability of a land use policy. The results can show areas of statistical agreement and disagreement, whilst also revealing distinct narratives emerging from groups of respondents [31,32]. When applied to situations with conflicted stakeholder dynamics, Q-methodology can be useful in identifying common ground among diverse stakeholders in situations where conservation or resource management is contested [16,33]. This makes the method particularly useful for this study due to the above benefits.

#### *2.3. Data Collection*

The research undertaken in the UK took place in late 2018 (when the UK was still part of the EU) and in Norway in mid-2019. Q-methodology studies commonly begin by using a qualitative approach, where interviews are undertaken with a range of stakeholders on a study's topic to gather the diversity of opinions on the phenomenon in question. This data collection can be enhanced or replaced with a literature review. This qualitative step is used to develop the "concourse", which is the range of views (listed as statements) held on a topic, followed by a structured, quantitative interview where participants rank the concourse statements, usually based on the extent to which they agree/disagree. During these interviews, qualitative information is gathered from participants on their decisionmaking processes and preferences. Because the concourse is designed to cover as closely as possible all perspectives on a topic, and participants are chosen to cover the range of views, then random sampling from the wider population is not necessary. Because of non-random sampling and smaller sample sizes, conclusions cannot be generalised but the aim is to understand the range rather than the frequency of the views, and to find points of convergence or divergence of opinion.

The concourse for this study was developed by interviewing 18 European agricultural stakeholders on causes of declining soil quality and corresponding solutions. Interviewees were purposefully chosen to represent researchers, land managers and other stakeholders from ten European countries participating in the wider project, SoilCare, on which this study is based. Ten researchers and eight other stakeholders (representing agricultural unions, farmers and other landowners) were interviewed. An interview guide was used, which was piloted on a subset of the sample population and amended due to feedback. Interview themes and prompts are shown in Table 1. Interviews were undertaken by telephone or Skype and lasted an average of an hour. Free, prior informed consent was obtained from all interviewees and ethical approval was gained from Newcastle University. Interviews were recorded with permission from the participants and later transcribed. Interviews were conducted in English, apart from one which took place in Italian, which was later translated to English for analysis.

**Table 1.** Interview guide used to develop the concourse.


A narrative review was undertaken, based on a broad-based search for relevant material, to provide further evidence to supplement the interviews and further expand the concourse. This review was to ensure that the topic was sufficiently covered by the statements developed from the interview data. Data were then analysed using a thematic analysis focusing on reasons for soil quality decline and solutions for how to fix this. A total of 142 statements was obtained from across all interviews and the literature review, which included statements both for the problem Q-set and the solution Q-set.

Similar statements for each set were merged, whilst trying to retain as far as possible the original wording of the interviews to capture the intent of the source. For both studies, some statements arising from the literature were amended subtly to match the country's context, e.g., changing the statement "EU agricultural policy" to "Norwegian agricultural policy", and adding local problems such as drainage. For the UK study, this resulted in 41 statements for the "problems" Q-set and 34 statements for the "solutions" Q-set, and in Norway, this resulted in 42 problem statements and 36 solution statements (see Tables A1–A4, Appendix A).

A "Q-sort" is the ranking of the Q-set by participants. Data collection for the Q-sort was undertaken via an online survey using Google Forms. The Q-sort survey was first piloted on a subset of the target population and subsequently adapted following feedback to improve question clarity and to include additional statements that were not captured via the interviews or literature review. Participations then ranked the statements on a scale of −2 (strongly disagree) to +2 (strongly agree). The UK survey was distributed via soilspecific newsgroups, British agricultural union members and by sharing on agricultural social media channels. A total of 61 UK respondents undertook the survey: 19 scientists, 19 farmers, 16 agricultural advisers, three water company employees who work in agriculture, two nature conservationists, one agricultural union representative and one civil servant. For the Norwegian study, a link to the survey was distributed in "Plantenytt", a newsletter from the government extension service Norsk Landbruksrådgivning Øst and to a local "soil education group". Forty-two Norwegian farmers took part in the survey, as well as six agricultural advisers and five scientists, totaling 53 respondents. The substantial weighting towards farmers in the Norwegian study was deemed acceptable due to the smaller average farm size in Norway and the historical legacy of communal land management which is embedded in national agricultural institutions [26]. However, the findings of our analysis may need to be interpreted in light of the greater diversity of stakeholders in the UK study.

At the end of the survey, participants were asked what they thought was the leading cause of declining soil quality and the most important solution to solve this problem. Participants could choose a statement from the Q-sort or add a new statement. These open-ended questions were used to find out what, subjectively, respondents thought were the most important drivers for causing declining soil quality and how to fix these. Data from these open-ended questions were analysed via thematic analysis. Quotes in the results section are used to highlight common sentiments as well as responses that stood apart from the rest. Quotes from the Norwegian study were translated into English.

#### *2.4. Analysis*

Data from the Q-sorts were analysed using KenQ (https://shawnbanasick.github. io/ken-q-analysis, accessed on 10 January 2022). First, a principal component analysis (PCA) was used to identify the groups of participants who ranked their Q-sorts similarly, also known as a "loaded factor". Flags were automatically added to respondents that significantly loaded onto these factors at *p* < 0.05.

For the UK study, the PCA for the problem Q-sort revealed eight factors with Eigenvalues >1 (which together explained 67% of the variance) but most loaded onto factors 1–4 (which together explained 53% of the variance). Large datasets, such as in this study, run the risk of inflating the Eigenvalues [34]. Because of this, we focused on the first four factors for the problem set as this explained over half the variance. A Varimax rotation was then applied to the four factors, which calculated the highest variability between factors. A *z*-score was calculated based on the average ranking participants gave to the statement within each factor group. Respondents that significantly loaded onto more than one factor were excluded from subsequent analysis because their inclusion gives little information about the clustering of opinions. Statistical disagreement (and agreement) between participants was set where *p* > 0.01, which meant that the groups of participants did (not) rank the statements differently at the 99% confidence level. The PCA for the solutions Q-sort revealed eight factors with Eigenvalues > 1 (which together explained 79% of the variance) but most loaded onto factors 1–3 (which together explained 65% of the variance). The rest of the solutions analysis followed the same process as with the problem Q-sort.

For the Norwegian study, the analysis followed the same procedure as the UK study. For the problem sort, eight principal components with Eigenvalue above 1 were extracted through the PCA, which explained 69% of the variance. Most of the participants loaded onto the first three problem factors, which together explained 51% of the variation, and these three factors were carried forward for further analysis. For the solution sort, eight factors with Eigenvalues above 1 were extracted, explaining 78% of the variance, though as respondents loaded onto factors 1–3, explaining 63% of the variance, these three factors were used in further analysis.

#### **3. Results**

This section describes results from the problems Q-sorts (Table A1, Appendix A: UK; Table A2, Appendix A: Norway) and solutions Q-sorts (Table A3, Appendix A: UK; Table A4, Appendix A: Norway). The number of respondents loading onto each factor (i.e., who ranked statements similarly) is shown in Figure 1 (UK problem Q-sort), Figure 2 (Norway problem Q-sort), Figure 3 (UK solution Q-sort) and Figure 4 (Norway solution Q-sort). Results are grouped under the key defining factors that emerged from each Q-sort, which are summarised in short, narrative phrases based on the main defining traits of each factor. Key areas of consensus and disagreement that emerged across these different groupings are then highlighted.

**Figure 1.** Professions of UK respondents loading onto the four problem factor groups.

**Figure 2.** Professions of Norwegian respondents loading onto the three problem factor groups.

**Figure 3.** Professions of UK respondents loading onto the three solution factor groups.

**Figure 4.** Professions of Norwegian respondents loading onto the three solution factor groups.

*3.1. Perceived Problems Causing Declining Soil Quality*

3.1.1. UK Study

Factor 1: "Intensive Agriculture to Blame"

This factor was defined by respondents who were significantly more likely to think the problems causing declining soil quality were due to "intensive use of soil without time to recover" and "overuse of inputs", more strongly agreeing with these statements than other factors. In contrast, they strongly disagreed that the problem was caused by the fact that "soil has become too saline", ranking this statement more negatively than other factors.

#### Factor 2: "Farmers Need to Change"

This factor was defined by respondents more strongly agreeing than other factors that "lack of knowledge of soils amongst farmers" and "some traditions of farmers are damaging" were causing problems in soil quality. Conversely, they more strongly disagreed with statements regarding "declining level of nutrient status", "loss of number of wild species" and "I do not believe there is a problem with soil quality" than other respondents.

#### Factor 3: "It's the EU, Not Farmers That Are to Blame"

Respondents here were defined by more strongly agreeing than other factors with the idea that "EU agricultural policy" was the cause of declining soil quality, along with "lack of knowledge of soils amongst farmers" and "natural local climate constraints". Conversely, they strongly disagreed that the problems were "use of contractors" and "loss of numbers of wild species" compared with other respondents.

#### Factor 4: "Weather and Farm Management to Blame"

Respondents here more strongly agreed than other factors that "pressure on farmers to produce at a low cost", "choice of cropping system" and "flooding or drought" were causing problems with soil. Conversely, they more strongly disagreed with statements regarding "lack of knowledge of soils amongst farmers", "overuse of inputs" and "distrust of scientists by farmers" were causing problems, when compared with other respondents.

#### Areas of Agreement and Disagreement

Respondents in all factors strongly agreed that soil quality was declining due to loss of soil structure, and agreed/strongly agreed that compaction, soil erosion, loss of organic matter and insufficient knowledge exchange were other causes (Table 1). In contrast, they did not think that the causes were due to farmers having little control over their land or due to a distrust of scientists by farmers. Conversely, the only area of statistical dissensus between each factor was "lack of knowledge of soils amongst farmers" (Table 1).

#### Leading Causes of Declining Soil Quality

There were two themes that were frequently mentioned by UK respondents as the leading causes of declining soil quality when answering this open-ended question. The first group blamed market pressures for pushing farmers into intensifying farming, with a sentiment that an ever-increasing drive to produce more food at cheaper costs was a fundamental driver of unsustainable land management, including soil quality decline. This could relate to the Q-set statement 21, "pressure on farmers to produce at low cost", which respondents in Factors 1, 3 and 4 strongly agreed was a cause for declining soil quality. This sentiment is captured by an agricultural adviser (UK2) who said:

"There is an increasing demand to produce cheaper food for a larger population using the same/declining land area. Pressure is put on producers by supermarkets and the general public to provide food to contracts, often unknowingly, which results in poor management choices".

Conversely, the second group blamed farmers and thought that intensive agricultural practices, such as ploughing and insufficient crop rotation, were the leading causes of soil quality decline. Many respondents felt this was due to a lack of understanding by the farmer of better soil management practices. One nature conservationist (UK4) summarised this theme by saying the problems causing declining soil quality were due to:

"Traditional' farming practices and cropping, which means too many farmers not being innovative/open to new methods. Time to start re-thinking about how we measure what makes a successful farm-it's not all about productivity".

This sentiment was not reflected in the answers to the problem Q-sort. One of the reasons for this could be that it encapsulates many of the problem statements related to farm management, as it is a multi-faceted and complex problem.

#### 3.1.2. Norway Study

Factor 5: "Disconnection between Farmer and Soil"

Respondents in this factor were more likely to rank "poor management of the soil" as one of the main reasons for the decline in soil quality. This group disagreed more strongly with the statement "I do not think there is a problem with soil quality". Instead, they ranked statements on agricultural practices and farmers' knowledge as leading causes for declining soil quality, such as "farmer has lost the finer touch with his land", "overuse of input like fertiliser and chemicals", and "lack of knowledge of soil amongst farmers".

#### Factor 6: "There Is No Problem with the Soil Quality"

Respondents in this factor disagreed that agricultural practices are reasons leading to a decline in soil quality such as "intensive agriculture to blame" and "overuse of input like fertilisers and chemicals". They also strongly disagreed with the statements "use of contractors", "too much leased land" or "farmers have lost the finer touch with their land". They agreed more strongly than others with the statements "too little advice on soil-improving practices" and "lack of knowledge-sharing between scientists, advisors, and farmers" as problems for soil quality.

#### Factor 7: "Industrialised/Intensive Agriculture to Blame"

Respondents in this factor thought the problems were often outside of the farmer's actions and responsibility compared to factor 5, being significantly more likely to agree on "pressure on the farmer to produce at low cost", and "intensive agriculture" than the other factors. This group also more strongly agreed about structural characteristics like "too large farms" and "high share of leased land" as problems of declining soil quality compared with other factor groups.

#### Areas of Agreement and Disagreement

Respondents in the three factors agreed/strongly agreed that soil quality was declining because of "soil erosion", "repetition of the same crop, year after year", and "loss of soil structure". There were also numerous areas of statistical disagreement between the factor groups (Table A2, Appendix A), such as knowledge/education, environmental conditions and management of the farm.

#### Leading Causes of Declining Soil Quality

There were two common perceived causes of declining soil quality. The first was an increase in use of large machinery causing soil compaction, captured by the statement (N5):

"Larger farms stimulate heavier machinery leading to more compaction" and " . . . modern machinery can drive in unfavourable conditions".

The second aspect was lack of crop rotation, which respondents felt contributed to declining levels of SOM, while some connected monocultures to the regional specialisation policy.

#### *3.2. Perceived Solutions to Address Declining Soil Quality*

#### 3.2.1. UK Study

#### Factor A: "Anti-Innovation"

This factor was defined by respondents more strongly agreeing that there is not much we can do to improve soil quality and that the problems were due to natural climatic constraints. They also disagreed with innovations and increasing early adoption of new techniques to solve the issue, ranking these statements more negatively than other factors.

#### Factor B: "Yes to Financial Incentives but No to Regulation"

Respondents here more strongly agreed that financial incentives could be a solution but more strongly disagreed that restrictive policies, such as more regulation (including for fertilisers and to reduce water usage) and creating a Soil Directive, would improve soil quality.

#### Factor C: "Early Adoption of New Techniques"

This factor was defined by respondents more significantly agreeing to increasing early adoption of new techniques as a solution to declining soil quality. They also more strongly disagreed that solutions were maintaining small farms, giving more freedom for farmers to manage their land as they would like, and that farmers have already tried lots of things.

#### Areas of Agreement and Disagreement

Respondents loading onto the three factors strongly agreed that more research should be done in collaboration with farmers, and all agreed in investing in education and training (Table A3, Appendix A). They also agreed that we should work towards improving trust between farmers and regulatory agencies and initiatives to reduce compaction. Respondents did not think changing the timing of tillage would improve soil quality. There was disagreement on numerous solutions, particularly around maintaining small farms, increasing the early adoption of new techniques and giving more freedom to farmers to manage their land.

#### The Most Important Perceived Solutions to Addressing Soil Quality Decline

There were two main themes that emerged in the responses to the open-ended question, with the first (and most common) requesting improved knowledge exchange between agricultural stakeholders. This links to the Q-set statements on "more research should be done in collaboration with farmers" and "investing in education and training", to which all factors agreed. This theme can be best encapsulated by a quote from a researcher (UK5) who said the solution lay with:

"Two-way communication between farmers, researchers and policy makers. Even the best solutions will not work if they can't be shown as favourable or acceptable to the farmer".

The second theme was around suggestions of using soil-improving cropping systems, or derivatives thereof, such as diverse crop rotations, direct drilling and reduced tillage. This related to many of the solution Q-set statements, such as on cover crops, rotation and less use of heavy machinery.

#### 3.2.2. Norway Study

Factor D: "Farmer-Led Demonstration and Innovation"

Respondents in this factor were more likely to rank "setting examples to follow; if one farmer succeeds others will follow", "more innovation" and "more targeted mapping of soil threats" as solutions to declining soil quality than others. This group disagreed more strongly than others on "more small farms" and "reduction of leased land" as solutions to increase soil quality and was the only group that was neutral on the statement "reduce use of heavy machinery".

#### Factor E: "No More Regulation or Financial Incentives"

Respondents in this group agreed more strongly on "farmers have already tried many measures to improve soil quality" compared with other factors. They disagreed on "more use of cover crops", "financial incentives", "creation of a soil directive", "more regulation of fertiliser use", and "more regulation" as solutions.

#### Factor F: "Society Needs to Change"

The respondents in this factor distinguished themselves from the others by strongly agreeing on "society needs to change focus on what farmers produce". This group also agreed that the solutions could be to "reduce use of heavy machinery" and "more use of cover crops", though not at the *p* < 0.01 level.

#### Areas of Agreement and Disagreement

Respondents agreed on "less soil compaction" and "more variation in crop rotation" as ways to improve soil quality (Table A4, Appendix A), as well as on statements related to education, such as "investment in education and training" and "more farmer demonstration days". Respondents did not think "there is not much we can do with the cropping system to improve soil quality" or that the "problems are due to natural, climatic variations". Further, respondents were strongly against "more use of financial penalties" and were neutral or disagreed with "financial incentives" as a solution.

#### The Most Important Perceived Solution to Declining Soil Quality

More than half of the respondents mentioned "soil organic matter", "cover crops" or "crop rotation" in the open-ended section as solutions to improve soil quality. In addition, more drainage was mentioned by eight respondents as the most critical measure to increase soil quality in the open-answer section, a factor not discussed at all in the UK survey.

#### **4. Discussion**

Understanding the range of stakeholder perceptions of the causes of, and solutions to, declining soil quality is useful as it can highlight potential tensions and agreement that might affect the acceptability of land management policies and measures. In our work in the UK and Norway, whilst there were disagreements between respondents on the perceived causes of soil degradation, there was consensus on numerous soil-specific factors, e.g., compaction, soil erosion and loss of organic matter. Both groups agreed that the underlying drivers of declining soil quality were related to wider issues around industrialised agriculture and demand for cheap food, which many farmers felt were out of their control. When it came to solutions, some stakeholders felt that society needs to change in order to address these underlying drivers. Knowledge exchange between agricultural stakeholders was also seen as key. However, many respondents were against further regulation or financial mechanisms including both incentives and penalties.

When focusing on the causes of declining soil quality, studies show that UK soils are threatened by soil erosion, compaction and organic matter decline [21], which reflected the main problems that UK respondents believed were causing declining soil quality. Respondents in Norway also thought soil degradation was due to soil erosion and loss of soil structure, reflecting findings in southeast Norway, where erosion and loss of soil structure have been linked to increased soil compaction [35]. However, Norwegian respondents considered lack of crop rotation as a problem causing soil decline, which was not noted in the UK study, perhaps reflecting the fact that crop rotations were at the time incentivised in the UK via the EU Common Agricultural Policy's three crop rule [36].

Reducing compaction was agreed to be key to improving soil quality for UK and Norwegian respondents. There has been significant interest in the effects of compaction over the last few years in both Norway and the UK, with numerous research projects, training events, innovations and industry-led technology to help address this problem (e.g., [35,37,38]). This suggests compaction is a salient issue for respondents. However, some of the ways for dealing with compaction, such as reducing usage of heavy machinery, were not highly rated by respondents in this survey. More research would be needed to understand why this is.

Industrialisation of the agri-food sector was thought to be a driver of soil degradation. This perception reflects the significant structural changes in southeastern Norway, described by Bjørlo and Rognstad [39] in their report "Barely recognisable" (translated from Norwegian). When analysing the answers to the open-ended question about the main problem causing declining soil quality, both British and Norwegian respondents often highlighted the complex nature of soil degradation, related to external pressures

along the food supply chain such as consumers demanding cheap food and agricultural policy and supermarkets dictating farm management. One UK farmer summarised this sentiment succinctly by stating: "give a farmer the right tools and he can put things right but remember he is only a puppet in a political system". Competing demands were thought to be placed on farmers, pulling them in different directions and this was thought to have a negative effect on the soil. To illustrate this tension, one UK adviser stated that "the machinery industry wants to sell big heavy machinery and the agronomy industry wants to sell more chemistry and soil health is the loser". The pressure for farmers to produce more food as cheaply as possible appeared to be part of the symptom of industrialised agriculture where farmers felt trapped and unable to improve their soil quality due to these powerful external market forces.

Whilst respondents in the UK study felt there had been a decline in soil quality, many respondents in factor 6 of the Norwegian study ("I do not think there is a problem with soil quality") did not agree. There can be several reasons to why this may be, such as respondents in factor 6 conceptualising "soil quality" in a different way to others, thereby not considering there to be a decline. For instance, a crop consultant wrote, "I do not think that soil quality has gone down, farmers are harvesting higher and higher yields". This might reflect a more historic definition of the term "soil quality" that focused on productivity rather than wider ecosystem services [40], where continued application of fertiliser and pesticides can mask underlying soil quality issues [41]. In addition, the larger proportion of non-farmers in the UK study group may have resulted in a greater emphasis on declining soil quality, with the scientists in the UK group most commonly identifying industrial agriculture as a causal factor in declining soil quality.

When it came to improving soil quality, many respondents in both studies were neutral towards or disagreed with financial measures, including penalties and incentives. This is a finding also established in other studies [16,42] whereby farmers felt financial incentives in particular were bribes to coerce farmers into doing what others wanted them to do. Whilst there may be some reluctance to agree to using financial incentives to change farmer behaviour, when implemented effectively they can change farmer practices and produce environmental benefits [43,44]. Leaving the EU presents the UK with an opportunity to revolutionise its agricultural policy and there is increasing interest in paying farmers for providing essential ecosystem services, such as soil conservation [45]. However, given that numerous respondents in this study did not think financial incentives could reduce soil degradation, it remains to be seen whether this approach will result in widespread uptake or improved soil quality, especially if fundamental drivers of soil degradation are not also addressed. For instance, if supermarkets and consumers continue to demand and purchase high quantities of cheap food, it is possible that market forces may undermine Government incentive schemes if the profits that farmers make from selling cheap, industrially produced food is more than what the Government can offer. Equally, supermarkets often tie farmers into contracts, with strict requirements on yield and quality of food but limited attention towards how the food is produced. To transform the agri-food system, supermarkets should also start requiring food to be produced in more sustainable ways [46].

Many UK and Norwegian farmers in this study did not believe EU or national intervention could improve soils, such as by creating a Soils Directive, and were also more negative with regards to any form of regulation. In the UK, this may partly be as a result of longstanding political opposition on the issue; in 2012, UK Ministers, together with Germany, France, The Netherlands and Austria, played a key role in blocking an EU Soils Directive [47]. This finding may also reflect the fact that the survey ran during the EU "Brexit" negotiations, where trust in the EU by many UK citizens was at a low point, suggesting a lack of faith in the UK's national application of the Common Agricultural Policy. It remains to be seen whether trust can be rebuilt between British farmers and policymakers as the UK leaves the EU and devises its own agricultural policies.

Investing in education and training were additional solutions that respondents agreed upon. Research has shown that education and training can be effective at spreading awareness and encouraging uptake of more sustainable agricultural practices [48,49]. This may work best with farmers who are more open to learning about new topics and trying novel approaches. However, conservative farmers that are more risk-averse and less willing to change might be less likely to attend training events or try new practices [50] and it could be these farmers that are undertaking the most soil-damaging practices. Targeting these hard-to-reach farmers has continued to prove challenging though one way of addressing this could be to frame knowledge exchange events in ways that attract these farmers by focusing on aspects they are passionate about, and where the event is run by someone whom they respect and relate to. In particular, local peer-to-peer knowledge exchange events have been identified as offering scope for strengthening land manager networks and facilitating behavioural change through exchange of information and experience [51,52]. Opportunities also exist to use online information and integration to influence farmers to change their practice, although more understanding of the effectiveness of this type of approach is needed [53]. Another solution agreed upon by many respondents was to undertake more research in collaboration with farmers. The EU Horizon 2020 funding stream promotes a "multi-actor approach" for agricultural research projects, encouraging a diverse group of stakeholders to work together rather than research solely (or primarily) being conducted by researchers [54]. This approach has many potential benefits as it can help promote greater understanding of different perspectives, building empathy, making research more robust, allowing quicker uptake of results, and grounding research in nonacademic stakeholder experiences and knowledge, as well as others [55]. This collaborative approach may help stakeholders understand their epistemological differences and build trust to work together more effectively and respect each other's perspectives. Given that one of the suggested solutions was to build trust between regulators and farmers, future work should encourage participation of regulators in multi-actor projects. In the Norwegian study, "farmer demonstration days" were considered as an agreed solution to improve soil quality, where researchers, extension services and farmers meet to discuss both theoretical and practical aspects of agriculture. These demonstration days could provide a valuable opportunity for knowledge exchange between researchers, farmers and other stakeholders. Similar events are held in the UK and have been met with great success from the farmers attending.

#### **5. Conclusions**

To be well-informed, equitable and transparent, public decision-making needs to take account of the views of the diversity of stakeholders they may affect. Taking these perspectives into account in the policy-making process has the potential to deliver more robust decisions that are more beneficial for the environment and more likely to be implemented. Participatory processes can elicit a more inclusive range of perspectives than conventional consultative processes, revealing areas of consensus and disagreement that can inform policy development. They are also able to capture the likely social impacts of proposed policies, which are often neglected in favour of more straightforward environmental and economic appraisals [16,56]. Using Q-methodology to analyse the diversity of stakeholder perspectives, this research has shown that there is a diversity of perspectives on the problems of and solutions to declining soil quality across different professions within the agricultural sector in Norway and the UK.

Respondents in both countries found it easy to agree on the physical processes causing declining soil quality (in both countries, respondents pointed to a loss of soil structure and soil erosion). It was harder to find agreement on social and political drivers from the Qsorts, other than a lack of knowledge exchange in the UK. However, analysis of qualitative data suggested that respondents primarily blamed industrial agricultural methods, which in turn, they blamed on market drivers, pushing down farm-gate prices (in the UK) and regional specialisation policies (in Norway). Although these drivers of declining soil quality are difficult to address in the short term, and market drivers are outside the control of policymakers, the proposed solutions were pragmatic, focusing primarily on capacity

building measures. Respondents in both countries agreed that more investment was needed in training for farmers to use soil-improving cropping systems as an important way of improving soil quality. Respondents in Norway were strongly against the use of financial penalties to encourage the use of these cropping systems, and UK respondents believed that trust needed to be built between farmers and regulators, and that more research needed to be done in collaboration with farmers. It may be possible to engage farmers in action research, including the development of evidence-based training tailored to their needs, drawing on both existing evidence and findings from new collaborative research.

Linked to this, future research might integrate Q methodology with Delphi or other structured elicitation techniques to further triangulate and increase the robustness of findings. Notwithstanding the sample sizes in this research, it is important to note that Q methodology alone should not be used to generalise findings to wider populations, and so these findings should be seen as indicative of the views of some stakeholders in each country, rather than as an authoritative representation of the perspectives of these groups in general.

Although the limited sample makes generalisations inadvisable at national scales, areas of consensus are important for policy makers to understand, as they could indicate areas where policy changes might be more acceptable to a range of stakeholders, addressing the challenge of creating scalable policy options noted by the UNCCD. It is also useful to highlight areas of disagreement among stakeholders, so that further consultation can be carried out to understand the basis of dissensus and its likely impact on policy implementation. For example, in this study we highlighted the diverging view of perspectives on the underlying causes of declining soil quality, which variously blamed farmers (who "have lost touch with their land and are afraid of doing something new"), policy-makers ("it's EU policy, not farmers that are to blame"), the industrial system ("intensive agriculture to blame") and external forces from society ("pressure to produce at low cost").

As policymakers in the UK, Norway and other countries grapple with the challenge of feeding growing populations whilst mitigating climate change, there is a greater need than ever before to develop policies that are acceptable, implementable and sustainable. In this context, policies are needed that address the widest possible range of real and perceived causes of declining soil quality, harnessing the adaptability and ingenuity of farmers and other stakeholders as part of wider attempts to address systemic market and policy failures across the agri-food system. Dealing with soil degradation requires tackling underlying drivers and this study has highlighted numerous solutions for addressing this challenge that are acceptable to a range of agricultural stakeholders.

**Author Contributions:** Conceptualization: M.S.R., E.A.O.; Data curation: N.R., O.E.L., N.R., S.I., S.V., E.A.O., T.A.B.; Formal analysis: N.R., O.E.L.; Funding acquisition: M.S.R.; Investigation: N.R., O.E.L., S.I., S.V., E.A.O., T.A.B., M.S.R.; Methodology: N.R., O.E.L., S.I., E.A.O., T.A.B., M.S.R.; Supervision: M.S.R., N.R., T.A.B.; Project administration: M.S.R., N.R.; Writing—original draft: N.R., O.E.L., N.R., S.I., S.V., E.A.O., T.A.B., M.S.R.; Writing—review & editing: J.H.G., C.M.M., R.M., M.S.R. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by EU Horizon 2020 project 'SoilCare' (Soil Care for profitable and sustainable crop production in Europe, H2020-SFS-2015-2). This funder had no role to play in the design, data collection, analysis or interpretation of results.

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Newcastle University SAGE Faculty Ethics Committee.

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** The data that support the findings of this study are available from the corresponding author upon reasonable request and approval from the study sites representative co-authors.

**Conflicts of Interest:** MR is CEO of Fast Track Impact Ltd. and Research Lead for IUCN UK Peatland Programme. The other authors declare no competing interests.

#### **Appendix A**

**Table A1.** Average Q-sort scores for each of the four factors identified as causing problems by UK respondents. Bold text indicates distinguishing statement at *p* < 0.01, underlined text indicates consensus statements; scale from −2 (strongly disagree) to 2 (strongly agree).


**\*** Factor 1—Intensive use of agriculture to blame; Factor 2—Farmers need to change; Factor 3—It's the EU, not farmers that are to blame; Factor 4—Weather and farm management are to blame.

**Table A2.** Average Q-sort scores for each of the four factors identified as causing problems by Norwegian respondents. Bold text indicates distinguishing statement at *p* < 0.01, underlined text indicates consensus statements; scale from −2 (strongly disagree) to 2 (strongly agree).


\* Factor 5—Disconnection between farmer and soil; Factor 6—There is no problem with the soil quality; Factor 7—Industrialised agriculture to blame.

**Table A3.** Average Q-sort scores for each of the four factors identified as solutions to improve soil by UK respondents. Bold text indicates distinguishing statement at *p* < 0.01, underlined text indicates consensus statements; scale from −2 (strongly disagree) to 2 (strongly agree).


\* Factor A—Anti-innovation; Factor B—Yes to financial incentives but no to regulation; Factor C—Early adoption of new techniques.

**Table A4.** Average Q-sort scores for each of the four factors identified as solutions to improve soil by Norwegian respondents. Bold text indicates distinguishing statement at *p* < 0.01, underlined text indicates consensus statements; scale from −2 (strongly disagree) to 2 (strongly agree).


\* Factor D-Farmer-led demonstration and innovation; Factor E—No more regulation or financial incentives; Factor F—Society needs to change.

#### **Note**

<sup>1</sup> Definitions of "soil quality" vary and have progressed from focusing solely on agricultural production to a broader focus on the complex and diverse functions that soil confers to humans and our environment [7]. Here, we define soil quality as "the capacity of a specific kind of soil to function, within natural or managed ecosystem boundaries, to sustain plant and animal productivity, maintain or enhance water and air quality, and support human health and habitation. In short, the capacity of the soil to function" [57].

#### **References**


**Felice Sartori, Ilaria Piccoli \*, Riccardo Polese and Antonio Berti**

Department of Agronomy, Food, Natural Resources, Animals and Environment, University of Padova, Viale dell'Università 16, 35020 Legnaro, Italy; felice.sartori@phd.unipd.it (F.S.); riccardo.polese@unipd.it (R.P.); antonio.berti@unipd.it (A.B.)

**\*** Correspondence: ilaria.piccoli@unipd.it; Tel.: +39-049-827-2841

**Abstract:** The evaluation of the effects of conservation agriculture during the transition from conventional tillage to no-tillage requires numerous indicators to be considered. For this purpose, we monitored changes in a multi-parameter dataset during a three-year experiment that combined three tillage intensities (conventional tillage—CT; minimum tillage—MT; and no tillage—NT) with three soil covering managements (tillage radish cover crop, winter wheat cover crop and bare soil). Using a multivariate analysis, we developed a Relative Sustainability Index (*RSI*) based on 11 physical (e.g., bulk density and penetration resistance), chemical (e.g., soil organic carbon and pH) and biological soil properties (e.g., earthworm density) to evaluate cropping systems sustainability. The *RSI* was most affected by tillage intensity showing higher *RSI* values (i.e., better performances) in reduced tillage systems. Specifically, the *RSI* under NT was 42% greater than that of CT and 13% greater than that of MT. Soil covering had little impact on the *RSI*. Among the tested parameters, the *RSI* was increased most by saturated hydraulic conductivity (+193%) and earthworm density (+339%) across CT and NT treatments. Our results suggest that conservation agriculture and, particularly, reduced tillage systems, have the potential to increase farm environmental and agronomic sustainability.

**Keywords:** conservation agriculture; no tillage; minimum tillage; principal component analysis; soil quality index; scoring function

#### **1. Introduction**

Conservation agriculture (CA) is defined as the combination of three principles: minimum soil disturbance, permanent soil organic cover and species diversification [1]. In addition to reduced management costs, CA is considered to enhance several ecosystem services (soil physical and chemical properties, soil organic carbon (SOC) and biodiversity) [2,3] and prevent some soil threats, such as soil erosion [4–6]. Although some of these benefits remain less clear, the worldwide adoption of CA grew to 12.5% of 2016 global cropland. At odds with this growth trend, there is Europe, where only 5% of total cropland is managed with CA. One country that has shown particularly limited adoption is Italy—less than 300,000 ha (a mere 2% of agricultural land) [7].

Adoption of CA has suffered slow adoption in Europe primarily due to the long transition time that follows conversion from conventional agriculture to CA before the positive effects are realized. During this crucial period, farmers face reduced crop yield and new equipment expenditures. Conversion to CA also requires a permanent soil covering, yet another cost that would benefit from economic support [8]. A key reason behind the very limited use of CA in Italian agrosystems is the long conversion time (more than five years) required before SOC, fertility and nutrient use efficiency benefits are observed [9,10]. Most studies have considered transition time only as a function of a single parameter, such as soil physical properties [11,12], yield [13–16], net SOC stock [17,18], soil aggregate stability, biodiversity, SOC content [19,20], earthworm density, or CO2 emission reduction [21]. However, each of these exerts an effect on CA. As such, we suggest that a

**Citation:** Sartori, F.; Piccoli, I.; Polese, R.; Berti, A. A Multivariate Approach to Evaluate Reduced Tillage Systems and Cover Crop Sustainability. *Land* **2022**, *11*, 55. https://doi.org/ 10.3390/land11010055

Academic Editors: Amrakh I. Mamedov, Guido Wyseure, Julián Cuevas González and Jean Poesen

Received: 14 December 2021 Accepted: 30 December 2021 Published: 31 December 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

holistic approach capable of considering multiple parameters may provide a better means by which to evaluate the effects of CA.

The ecosystem services delivered by CA justify the need for environmentally conscious policymakers to consider economic support of the practice through a program such as the "green payments" program already established in the EU Common Agricultural Policy [22]. Alternatively, yield losses and/or other negative effects could be limited or compensated in some fashion. In either case, programs such as these are effective only when the protocol created is adaptable to local area specifics but is assessable by a single, consistent set of criteria. As is often described in the literature, an index represents one way to determine and compare the impact of different management strategies [23,24]. Similarly, the literature has already identified potential soil quality indicators to comprise an index: physical soil property measures (soil hydraulics, penetration resistance and bulk density) [25,26] plus soil aggregate stability [27], soil C and N content and earthworm density [28]. Masto et al. [29] previously adopted a statistical methodology to determine the impact of different management strategies on soil quality and sustainability using a dataset with several soil characteristics, as reported above. The method involved the application of a principal component analysis (PCA) to derive the weight of the different soil parameters in promoting the sustainability. The derived index showed to be a reliable tool to assess the performance and impacts of alternative land uses and soil management options [23].

In this work, three tillage systems (no tillage, minimum tillage and conventional tillage) were combined with three different soil coverings (tillage radish cover crop, winter wheat cover crop and bare soil) to compare the effects of the main factors influencing CA. A multivariate approach was applied to a dataset of soil quality measures taken during a north Italy field experiment. A sustainability index was calculated to compare different treatment combinations as a function of the selected indicator variability. This study aims to determine the short-term effects of reduced tillage and cover crops on the studied parameters. Our starting hypothesis was that a reduced tillage system combined with tillage radish could minimize conversion time side effects and improve soil properties.

#### **2. Materials and Methods**

#### *2.1. Experimental Design*

The experiment took place at the Lucio Toniolo Experimental Farm, located in Legnaro, PD (NE Italy, 45◦21 N; 11◦58 E; 6 m a.s.l.). The climate is sub-humid with average temperatures between −1.5 ◦C in January and 27.2 ◦C in July. Rainfalls reaches 850 mm annually. The reference evapotranspiration of 945 mm exceeds rainfalls from April to September. The highest rainfalls occur in June (100 m) and October (90 mm), while winter is the driest season, with average rainfalls of 55 mm. The shallow water table ranges in depth from 0.5 to 2 m, with the lowest values recorded in summer.

This three-year study began in spring 2018 and it was designed as a split plot, with two replicates located in a flat area of the Po valley with a maximum slope < 1%. An area of 2 ha was divided into 18 plots, each of 1111 m2. The soil at the site was Fluvi-Calcaric Cambisol [30] with a silt loam texture (25% clay, 50% silt and 25% sand), pH 7.8, 27.1% total carbonate content, <1% soil organic carbon and <0.1% total nitrogen. The main factor was tillage intensity; conventional tillage (CT) was ploughed to a depth of 30 cm and then harrowed to 15 cm; minimum tillage (MT) was tilled with a harrow to a depth of 15 cm; no tillage (NT) was sod seeded with a zero-tillage seeder that included double disks for furrow openers and press wheels for soil firming. Within each main plot, three winter soil coverings were randomized: tillage radish (TR—*Raphanus sativus* L.), winter wheat (WW—*Triticum aestivum* L.) and bare soil (BS), where only residues from the previous year crop were present. Cover crops were seeded on residues from the main crop (always maize, *Zea mays* L.) in autumn 2018 and 2019.

#### *2.2. Field Surveys*

According to what already reported by other authors [25–28], we chose 11 parameters to monitor changes in the condition of the soil: (1) aggregate stability (Agg), (2) bulk density (BD), (3) soil organic carbon (C org), (4) total nitrogen (N tot), (5) gravimetric water content (GWC), (6) penetration resistance (PR), (7) saturated hydraulic conductivity (Ks), (8) earthworm density (EW), (9) mineral nitrogen (N min), (10) pH and (11) cash crop yield (Y) (Table 1). Each parameter was measured at two times. The first measurement was taken immediately after treatment combination adoption (T0) and the second measurement was taken at the end of the three-year period (T1). The method for determination of the measure of each parameter is fully described below.

**Table 1.** Soil parameters used for building the sustainability index.


A continuous value of Agg was determined. The Slakes application [27,31] was employed to soil aggregates in the 0.2–2 cm fraction sampled from the 0–20 cm soil layer. Three randomly selected aggregates from each sample were analyzed to produce a dimensionless slaking index (*SI*) with a value ≥ 0. A low *SI* (<3) represents high aggregate stability, an *SI* between 3 and 7 indicates moderate stability and an *SI* above 7 indicates that the aggregates have low stability. The *SI* was calculated as the difference between the wet aggregate area (*At*) after 10 min of water saturation and the dry aggregate area (*At*0), divided by *At*0, as shown in Equation (1).

$$SI = \frac{A\_{\rm f} - A\_{\rm t0}}{A\_{\rm t0}} \tag{1}$$

The BD was measured in the 0–30 cm soil profile with the core method as described in Grossman and Reinsch [32]. In the studied soil, a BD value of 1.55 g cm−<sup>3</sup> was considered a limiting condition to the growth of plant roots [33].

The C org and N tot contents were determined from shallow layer (0–30 cm) samples. The soil was air-dried and sieved at 0.5 mm and the inorganic carbon was removed with an acid pre-treatment. Subsequently, SOC and N tot were determined with flash combustion using a CNS Elemental analyzer (Vario Max; Analysensysteme GmbH, Langenselbold, Germany).

Four sampling areas were selected in each plot for GWC and PR measurements. For the PR, the measures were taken from the 0–20 cm layer and an average PR value was calculated. In each sampling area, a disturbed soil core was collected, weighted and ovendried at 105 ◦C to determine the GWC. For the PR, the measures were taken in each plot with the Penetrologger (Eijkelkamp, Giesbeek, The Netherland). A PR value above 2.5 MPa was considered a limiting factor to plant root growth [34].

The *Ks* [35] was determined using the double-ring infiltrometer method [36]. An inner ring of 60 cm in diameter was used to measure both the row and inter-row areas in the tillage radish plots. The water within the inner ring was maintained at two levels. As one operator measured the time for the water to reach the lower level from the upper level, another added more water to reach the upper level again. This operation was replicated

until the infiltration rate was constant. Meanwhile, the water in the external ring was suspended at an average value between the two levels of the inner ring. Then, the data were analyzed by fitting Philip's equations [37] with the Solver Add-in for Microsoft Excel.

$$i(t) = S \times t^{1/2} + At \tag{2}$$

$$v(t) = \frac{\mathcal{S} \times t}{2}^{-1/2} + A \tag{3}$$

where *i(t)* is the water infiltration (m) and *v(t)* is the infiltration rate (m s<sup>−</sup>1) expressed as a function of time. Parameters S and A were calculated with the Solver add-in by minimizing the square difference between predicted and observed *i(t)* and *v(t)*. The *Ks* was calculated as below and m is a constant equal to 2/3.

$$\text{Ks} = \frac{A}{m} \tag{4}$$

The EW was measured with a mustard extraction as described by Valckx et al. [38]. The measure was performed by taking an earthworm extraction from the soil surface using a water-suspended mustard in a 25 × 25 cm<sup>2</sup> frame [38]. First, we used the number of extracted earthworms to score soil quality [39]. A density of <4 was the lowest score or of "poor" soil condition, a density of 4–8 was "moderate" soil condition and the highest density (>8) was "good" soil condition. Then, the earthworm count was compared amongst the different treatment combinations.

We estimated N min based on samples of the 0–20 cm soil layer. Concentrations of ammonium, nitrite and nitrate were measured using a KCl extraction followed by photometry, as described by García-Robledo et al. [40].

Soil pH was determined from air dried, mixed and sieved (0.5 mm) samples taken from the 0–20 cm soil layer. The pH was measured in a 1 M KCl solution (1:2.5 solid–liquid ratio) [41].

At the end of the cropping season, four biomass samples were collected from each subplot to determine maize grain Y at 27% grain moisture. After the harvest, a grain sample was oven dried at 105 ◦C until it maintained a constant weight to determine the dry mass weight. The Y was expressed in kilogram of dry grain per hectare.

#### *2.3. Data Analyses and Statistics*

First, a mixed-effects model was constructed using tillage, covering and their interactions in each monitored year. These effects were treated as fixed effects and the block effects as random. Post hoc pairwise comparisons of least-squares means were performed, using Tukey's method to adjust for multiple comparisons, with a *p* < 0.05.

To calculate a soil quality index, we relied on the method of Masto et al. [23,29]. The procedure requires that indicators be selected once they have been surveyed and normalized with linear or non-linear scoring functions, so that higher scores represent better-performing observations. The indicators and their weights were determined using the multivariate analysis method of Andrews et al. [42,43], which has been adapted and applied to many studies evaluating long-term practices [23,29], combinations of various crop rotations under different residue managements [44,45] and different tillage practices [46].

The sampled data were normalized with a linear scoring function [23] by applying Equations (5)–(7).

$$S = \frac{\mathbf{x}\_{ij} - \mathbf{x}\_{i\text{ min}}}{\mathbf{x}\_{i\text{ max}} - \mathbf{x}\_{i\text{ min}}} \tag{5}$$

$$S = -\frac{\mathbf{x}\_{ij} - \mathbf{x}\_{i\max}}{\mathbf{x}\_{i\max} - \mathbf{x}\_{i\min}}\tag{6}$$

$$S = \frac{|\mathbf{x}\_{ij} - \mathbf{7}|}{|\mathbf{x}\_i - \mathbf{7}|\_{\max} - |\mathbf{x}\_i - \mathbf{7}|\_{\min}}\tag{7}$$

where *xi max* is the maximum value measured during the *i* parameter survey and *xi min* is the smallest. The *S* value ranges between 0 and 1, which corresponds to the minimum and maximum values, respectively, observed in the *i* parameter. Equation (5) was used as a "more is better" scoring function for C org, GWC, Ks, EW, N min, N tot and Y. Alternatively, the parameters Agg, BD and PR were scored with Equation (6), according to a "less is better" approach. Finally, Equation (7) was used for pH scoring. In this way, treatment combinations that most favorably impacted the parameters scored highest.

The Relative Sustainability Index (*RSI*) was calculated as the sum of the observed parameter score, weighted with principal component analysis weighting factors (*PWs*). These factors were calculated according to Masto et al. [23], by selecting principal components (PCs) explaining at least 10% of the variability. Within each of these PCs, loaded factors (values > |0.2|) were selected and their correlations were measured [43]. In cases in which r > |0.8|, only the factor with the highest load was used for *RSI* calculation, together with all the other uncorrelated highly loaded factors. The percentage of variation explained by each PC provided the *PW*. The *RSI* was calculated with Equation (8).

$$RSI = \sum\_{i=1}^{n} PW\_i \times S\_i \tag{8}$$

To normalize the *RSI*, this was divided by the highest *RSI* value obtained. A total of 36 *RSIs* were calculated, one per treatment combination replication in survey T0 and another in T1.

*RSI* differences amongst tillage, soil covering and their interaction were tested with mixed models and the model with the smallest Akaike's Information Criterion (AIC) was selected [47]. Post hoc pairwise comparisons of least-squares means were performed using Tukey's method to adjust for multiple comparisons, with a *p* < 0.05. The statistical analyses were performed using Microsoft Excel 2016, ClustVis [48] and SAS (SAS Institute Inc., Cary, NC, USA), version 5.1.

#### **3. Results**

Below is a description of the mixed model results comparing changes in the 11 indicators of soil quality under the tested treatments over time (Table 2). Table 3 reports the average 2019 and 2020 values used to calculate RSIs.

**Table 2.** Comparison of *p*-values among the linear mixed-effect model analysis of observed parameters (Agg—aggregate stability; BD—bulk density; C org—soil organic carbon; N tot—soil total nitrogen; GWC—gravimetric water content; PR—penetration resistance; *Ks*—saturated hydraulic conductivity; EW—earthworm density; N min—mineral nitrogen; Y—yield, CC—cover crop).


\*, \*\* and \*\*\* mean *p* < 0.05, <0.01 and <0.001, respectively.


**Table 3.** Descriptive statistics of studied parameters for T0 and T1 surveys (Agg—aggregate stability; BD—bulk density; C org—soil organic carbon; N tot—soil total nitrogen; GWC—gravimetric water content; PR—penetration resistance; *Ks*—saturated hydraulic conductivity; EW—earthworm density; N min—mineral nitrogen; St. Dev.—standard deviation; Var. Coef.—coefficient of variation).

All values of Agg, GWC and pH were significantly higher at T0 than at T1. Overall, average Agg was higher at T0 (4.50) than at T1 (3.19) and was characterized as of high-tomoderate stability, according to the *SI* range (0.3–6.1). At T0, all aggregate samples, except one, were of moderate stability (>3). The exception sample value, collected from treatment combination MT–BS, was 2.9. During the T1 survey, 44% of the observations were <3 (high aggregate stability) and the lowest values found in the reduced tillage systems (NT and MT). The measures of the GWC were strictly related to the pedoclimatic conditions on the sampling dates, with the GWC ranging from 20% to 25% in 2019 and from 12% to 22% in 2020. Cover crop treatments showed significant effects on GWC, as demonstrated by values of 18.3% in TR and 20.3% in WW, while BS had an intermediate value. Despite the significantly lower pH values at T0 versus T1, the pH values maintained non-critical averages (7.36 in T0 and 7.05 in T1).

Between survey T0 and T1, the N tot, PR and Ks all increased significantly. The N tot rose from 0.88‰ at T0 to 1.01‰ at T1. During each survey, the N tot maintained a modest variability, as indicated by the coefficients of variation at T0 (0.26) and T1 (0.13). The PR test values differed from an average of 0.70 MPa in T0 to an average of 1.34 MPa in T1. In the second survey, the PR was not only significantly higher, but also more variable than it was in T0; all of the PR observations across both surveys registered below the 2.5 MPa threshold. The PR differences occurred among the differing tillage systems. Specifically, CT reported a PR of 0.88 MPa, which proved to be significantly higher than the 1.18 MPa observed under NT. The PR result under MT was intermediate. Last, the *Ks* increased by 158% between T0 (3.4 × <sup>10</sup>−<sup>5</sup> m s−1) and T1 (8.7 × <sup>10</sup>−<sup>5</sup> m s−1). This parameter showed it was also significantly impacted by different tillage intensities, as shown by the average *Ks* values of 1.05 × <sup>10</sup>−<sup>4</sup> m s−<sup>1</sup> in NT, 3.58 × <sup>10</sup>−<sup>5</sup> m s−<sup>1</sup> in CT and an intermediate value in MT.

The parameters BD and EW were affected by the time × tillage interaction. Despite a generally limited effect on the BD across the various treatments, the average BD under CT was lower during the first survey (1.39 g cm<sup>−</sup>3) versus the second survey (1.45 g cm−3). All measures of BD were less than its 1.55 g cm−<sup>3</sup> threshold. In the case of EW, variability was higher; it ranged between 0 and 20 (Table 3). Among the treatments, during T1, the EW differences were, on average, significantly higher (13.17) under NT than under CT (3.00).

The C org, Y and N min parameters resulted as unaffected by all factors tested. On average, the C org was 0.83% and displayed only a modest variability within and between the surveys. Similarly, Y (10.00 Mg ha<sup>−</sup>1, on average) and N min (24.54 mg kg−1, on average) showed no significance among the treatment combinations in the different surveys.

The values presented in Table 3 were normalized. The average of each treatment combination is presented in Figure 1 (biochemical parameters) and Figure 2 (physical parameters). Normalization allows higher values to be associated with parameter improvement and wider areas to represent an overall sustainability increment.

**Figure 1.** Biochemical parameter scores with average values in treatment combinations at T0 and T1 surveys (C org—soil organic carbon; N tot—soil total nitrogen; EW—earthworm density; N min—mineral nitrogen; Y—yield; BS—bare soil; TR—tillage radish; WW—winter wheat; CT—conventional tillage; MT—minimum tillage; NT—no tillage).

**Figure 2.** Physical parameter scores with average values in treatment combinations in T0 and T1 surveys (Agg—aggregate stability; BD—bulk density; GWC—gravimetric water content; PR—penetration resistance; *Ks*—saturated hydraulic conductivity; BS—bare soil; TR—tillage radish; WW—winter wheat; CT—conventional tillage; MT—minimum tillage; NT—no tillage).

Figures 1 and 2 show sizeable differences between the treatment combinations and two years. The correlation matrix between each parameter pair is shown in Table 4. As expected, the highest correlation resulted between the C org and N tot (r = 0.924). To identify which of these highly-correlated parameters could best explain treatment variation—and warrant inclusion in the *RSI*—we performed a principal component analysis (PCA).



Table 5 presents the PCA results. Each parameter was weighted according to the treatment variation it explained based on the PC selected.

**Table 5.** Results of principal component analysis under different treatment combination in different years. Bolded factor loads were considered as high. Bolded and underlined factor loads determined for each variable were those the PC considered in the *RSI* calculation. The weighting factor (*PW*) for each variable was equal to the variation explained by the PC selected (Agg—aggregate stability; BD—bulk density; C org—soil organic carbon; N tot—soil total nitrogen; GWC—gravimetric water content; PR—penetration resistance; *Ks*—saturated hydraulic conductivity; EW—earthworm density; N min—mineral nitrogen).


The parameters selected in PC-1 were N tot, GWC, Ks and EW. It showed that the N tot should be included in the *RSI* because it had the highest weight and it was highly correlated to the C org. In PC-2, the highly weighted parameters BD, PR and Y were all included in the *RSI* as they showed limited correlation amongst them. In PC-3, the Agg and N min were selected and, in PC-4, the pH was chosen. The *PW* of each parameter equals the variability explained by the PC selected for that specific factor (0.241 for PC-1, 0.149 for PC-2, 0.135 for PC-3 and 0.117 for PC-4). To normalize the RSI, the sum of the weighted parameters was divided by the highest sum of the weighted parameters reported across all observations (1.247). The value was reported under NT–WW (block 1) during the survey T0. The lowest value (0.358) was under CT–TR (block 2) in T1. Then, the resulting *RSI* was expressed by Equation (9).

#### *RSI* <sup>=</sup> 0.135*Agg* <sup>+</sup> 0.149*BD* <sup>+</sup> 0.241*EW* <sup>+</sup> 0.241*GWC* <sup>+</sup> 0.241*Ks* <sup>+</sup> 0.135*N*min <sup>+</sup> 0.241*N tot* <sup>+</sup> 0.117*pH* <sup>+</sup> 0.149*PR* <sup>+</sup> 0.149*<sup>Y</sup>* 1.247 (9)

Then, mixed models were calculated on *RSI* values, considering the combination of tillage and CC effects. The smallest AIC for the *RSI* linear mixed model was obtained when intercept, tillage and covering were tested as fixed factors and block was a random factor. Table 6 summarizes the *p*-values for the selected mixed model.


**Table 6.** Linear mixed model analysis of *RSI* output.

\* and \*\*\* mean *p* < 0.05 and <0.001, respectively.

Figure 3 displays the average RSIs and corresponding contribution from each parameter to it under each treatment. On average, the GWC (0.09) and N tot (0.13) impacted the *RSI* the most. During T1, their highest scores were in NT (GWC = 0.10 and N tot = 0.13.) Observations of the *Ks* and EW were notable in that they contributed little to the RSI, yet they were high variable across treatments. During T1, the *Ks* averaged 0.08 under NT, which was three-fold the value observed under MT (0.03) or CT (0.02). Similarly, the EW averaged 0.13 in NT, which was double that in MT (0.06) and four-fold the value observed in CT (0.03).

**Figure 3.** Average *RSIs* and the contribution of each parameter under different tillage systems in different years. Agg—aggregate stability; BD—bulk density; GWC—gravimetric water content; *Ks*—saturated hydraulic conductivity; PR—penetration resistance; EW—earthworm density; N min—mineral nitrogen; N tot—soil total nitrogen; Y—yield; CT—conventional tillage; MT—minimum tillage; NT—no tillage. Different letters represent significant differences of the global treatment *RSI* at *p* < 0.05.

No clear effect was observed for the soil covering treatment in either year and no statistical difference was found. During both years under TR, the minimum *RSI* was always reached (0.604 in 2019 and 0.583 in 2020). Higher values were recorded for coverings WW in 2019 and BS in 2020 (Figure 4).

**Figure 4.** Average *RSI* values for different soil coverings in different years. BS—bare soil; TR—tillage radish; WW—winter wheat. Agg—aggregate stability; BD—bulk density; GWC—gravimetric water content; *Ks*—saturated hydraulic conductivity; PR—penetration resistance; EW—earthworm density; N min—mineral nitrogen; N tot—soil total nitrogen; Y—yield; CT—conventional tillage; MT—minimum tillage; NT—no tillage. Different letters represent significant differences of the global treatment *RSI* at *p* < 0.05.

#### **4. Discussion**

In total, 8 of the 11 parameters revealed significant differences between T0 and T1, which suggests that the soil system was changing regardless of the agronomic management applied. One potential cause of these results may be attributed to differences in the environmental conditions at T0 and T1. The GWC was found to be affected by CC. For those who have considered the effects of CC on water cycle, the results have been contradictory. Some have found CC to improve water balance and water availability [49], while others have reported soil water reduction in the subsequent crop after CC termination [50]. Our results, where under WW, the GWC was high and, under TR, the GWC was at its lowest, are also mixed. In both instances, the results can be equally attributed to either better maintenance of soil water content by WW, or higher soil evaporation under TR, due to a lesser soil covering.

The different tillage systems seemed to have a stronger impact, especially on some parameters (BD, PR, *Ks* and EW), if compared to the CC effect. For example, under CT, the BD and PR values aligned with previous evidence that reduced tillage systems can increase soil strength and bulk density, especially in the first years [51]. The reduced BD and PR values were expected tillage effects under CT, given that that they were measured in the 0–20 cm soil layer. This result may also relate to instrument resolution; the PR can be negatively impacted by the high spatial variability in reduced tillage systems [52]. However, in this instance, the BD almost always remained below its threshold (1.55 g cm<sup>−</sup>3) for which it is known to limit plant root growth in silty loam soils [33]. Similarly, the PR values (0.46–1.96 MPa) fell well below the growth-limiting threshold usually set at 2.5 MPa [34]. Finally, the soil at the experimental site was characterized as having structural inertia in response to management changes [53–55].

Although the BD and PR values worsened slightly (i.e., soil strength increased) under reduced tillage systems, soil function was improved in NT, as evidenced by an increase of 193% in the *Ks* under NT relative to CT during T1. The highest EW value observed in this study may relate to the significant contributions made by earthworm bio-macropores to soil function and, in particular, air and water permeability, even in compacted soils. Earthworms can improve soil structure [56] and hydraulic properties [57] by burrowing and casting. The positive effects of NT on the EW confirm previous studies evidence [9,21,28].

The computation of the *RSI* highlighted the strong effect of the EW as it carried a high relative weight (11%, on average) within the index. It also showed a high variability among the different treatments. Additional parameters that averaged high impact on the *RSI* were the N tot (17%), GWC, (14%) and PR (11%), which, together, accounted for more than 50% of the *RSI*. In addition to the EW, *RSI* variability was driven by the *Ks*, N min and Y. In absolute terms, the *Ks*, N min and Y each had impacts of less than 10% on the *RSI*, but their variation coefficients ranged the highest (from 0.67 for the *Ks* to 0.31 for the Y). These two conditions suggest that this set of measures should be considered as the best to indicate soil quality changes during the conversion from conventional tillage to CA. The *RSI* results also suggest that the *Ks* and EW are two sustainability indicators that were positively affected by NT.

The final *RSI* score evaluates the combination of tillage intensity and soil covering with an holistic approach [58]. It showed the positive effect of NT relative to conventional tillage, even in the short term. Midway between the effects of CT and NT lay the MT system. It mitigated the negative effects on some physical parameters but lessened the improvements of biological parameters (EW). According to Issaka et al. [59], both the minimum and no-tillage systems resulted as sustainable techniques, considering the nutrient cycles. As opposed to other studies [9,10,54], clear negative effects during the transition time were not detected during this three-year experiment.

The limited differences reported for the various soil coverings may be evidence that a CC effect was masked by the strong effects of reduced tillage systems combined with the sampling methods used. It may be that longer conversion times or different sampling methods are required for CC effects to be revealed [60]. Even in the case of BS, a partial and spontaneous covering (weeds) may impact soil properties in a way not unlike that expected with CCs. Indeed, "spontaneous CCs" have provided ecosystem services [61–65]. In the presence of plant residues, microbial diversity [66] could improve to the point where it should even be considered an environmental sustainability indicator [67].

From another perspective, the modest TR effect could relate to sample timing. Most TR-related benefits (improved porosity and pore connectivity) occur only when taproots are degraded. At the same time, reports of short-term tillage radish benefits exist [68,69], although it seems that longer timespans are necessary to exploit the benefits of TR on soil properties [10]. The bio-tillage effect, which was expected from TR, as suggested by Zhang et al. [70], could be masked by earthworm activity in NT treatments, irrespectively from the presence of TR. The high EW values observed under NT could have performed this bio-tillage effect, which, according to the authors, could replace conventional tillage.

Then, even if the WW fibrous root apparatus had a limited impact on soil structure, many Poaceae CC improved overall system sustainability [71,72] and aggregate

stability [49,65,73], or nutrient cycles [74–76]. The combination of grass CC and reduced tillage systems proved to positively affect environmental sustainability, fostering biodiversity [77] and soil organic carbon [78].

In conclusion, to correctly evaluate the CA effect, especially on the soil system, a holistic approach should be preferred to consider both the effects on crop production and on soil physics, considering different soil function at different scales.

#### **5. Conclusions**

A multivariate analysis of selected sustainability indicators revealed a positive effect of reduced tillage systems management and in particular NT, despite the limited variation in the observed parameters.

Despite the short-term nature of the experiment, this positive result could be the effect of an increase in soil fauna activity, which could have contributed to soil structure improvement. As a consequence, NT seemed to impact soil physics and soil habitability, resulting in a significantly higher *RSI* value. The effect of CC was limited, but WW reported the best results in the short term, with a tendency to have higher RSI values.

Collectively, the combination of NT and WW can be considered the most promising in terms of sustainability improvement. In this study, only the short-term effect of different tillage and soil cover management results were reported. Therefore, longer-term experiments could better evaluate the effects of these management systems on some parameters, such as soil organic carbon, which have a wide impact on sustainability, yet vary little in the short term.

In conclusion, to correctly evaluate the CA effect, especially on the soil system, a holistic approach should be preferred to consider both the effects on crop production and on soil physics, considering different soil functions at different scales.

**Author Contributions:** Conceptualization, F.S.; formal analysis, F.S. and A.B.; investigation, F.S. and R.P.; data curation, F.S. and R.P.; writing—original draft preparation, F.S.; writing—review and editing, F.S., I.P., R.P. and A.B.; visualization, F.S.; supervision, I.P. and A.B.; project administration, A.B.; funding acquisition, A.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** The research leading to these results received funding from the European Union HORI-ZON2020 Programme for Research, Technological Development and Demonstration under Grant Agreement No. 677407 (SOILCARE Project).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data that support the findings of this study are available from the corresponding author upon reasonable request.

**Acknowledgments:** We thank Rudi Hessel (project coordinator) and all the work package leaders for coordinating the project activities.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **Comparison of Compaction Alleviation Methods on Soil Health and Greenhouse Gas Emissions**

**Jennifer Bussell 1, Felicity Crotty <sup>2</sup> and Chris Stoate 1,\***


**Abstract:** Soil compaction can occur due to trafficking by heavy equipment and be exacerbated by unfavourable conditions such as wet weather. Compaction can restrict crop growth and increase waterlogging, which can increase the production of the greenhouse gas nitrous oxide. Cultivation can be used to alleviate compaction, but this can have negative impacts on earthworm abundance and increase the production of the greenhouse gas carbon dioxide. In this study, a field was purposefully compacted using trafficking, then in a replicated plot experiment, ploughing, low disturbance subsoiling and the application of a mycorrhizal inoculant were compared as methods of compaction alleviation, over two years of cropping. These methods were compared in terms of bulk density, penetration resistance, crop yield, greenhouse gas emissions and earthworm abundance. Ploughing alleviated topsoil compaction, as measured by bulk density and penetrometer resistance, and increased the crop biomass in one year of the study, although no yield differences were seen. Earthworm abundance was reduced in both years in the cultivated plots, and carbon dioxide flux increased significantly, although this was not significant in summer months. Outside of the summer months, nitrous oxide production increased in the non-cultivated treatments, which was attributed to increased denitrifying activity under compacted conditions.

**Keywords:** nitrous oxide; N2O; carbon dioxide; CO2; greenhouse gas; compaction; earthworms; direct drilling; bulk density

#### **1. Introduction**

Soil compaction is a form of soil degradation, which is an issue worldwide, due to the detrimental effects it has on agricultural productivity, through reduced crop growth, increased soil erosion and nutrient depletion [1]. Within England and Wales, almost 4 million hectares of soil are at risk of compaction [2]. Compaction was identified by DEFRA as one of the three key threats to the agricultural and environmental productivity of soils [3] and one of the ten soil threats identified in Europe [4]. Although soil compaction is not a recent phenomenon, some modern farming techniques can exacerbate the risks, including increasing field size and weight of farm equipment [5,6]. In this study, we specifically looked at the impact of topsoil compaction exerted by trafficking, which can occur when soils are trafficked by heavy equipment, especially in wet conditions [7,8]. The susceptibility of soils to compaction depends on the interaction between soil physical properties and climate; often soils are workable when soil moisture is lower than field capacity, making the window of opportunity for poorly draining soils particularly narrow [5,9]. Heavy clay soils, such as those found at this experimental site, are therefore often prone to compaction when necessary field operations, such as harvest, coincide with wet weather. This may be exacerbated with the impact of climate change making weather patterns more extreme, with warmer wetter winters and increased occurrences of intense storms, potentially reducing machinery working days [10].

**Citation:** Bussell, J.; Crotty, F.; Stoate, C. Comparison of Compaction Alleviation Methods on Soil Health and Greenhouse Gas Emissions. *Land* **2021**, *10*, 1397. https://doi.org/ 10.3390/land10121397

Academic Editors: Guido Wyseure, Julián Cuevas González, Jean Poesen and Amrakh I. Mamedov

Received: 22 November 2021 Accepted: 15 December 2021 Published: 17 December 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Compacted soils have less pore space, and increased bonding between particles, which leads to several problems. They will take more energy to cultivate, and aggregates will be harder to separate [11]. As pore spaces shrink, less space is available for water, and capillary attraction holding water within the soil increases, reducing water availability and plant uptake. Lower pore space and reduced infiltration also reduces soil aeration; this combined with restricted root growth impairs nutrient and water availability, reducing crop growth [9,12]. Compacted soils can also have detrimental effects on soil fauna, most notably, earthworms are often cited as being negatively affected by compaction, due to physical crushing and disruption of their burrow network [13]. Earthworms are also considered one of the key biological engineers needed to improve soil structure after compaction [14].

Due to the poor structure of compacted soils, they can become progressively poorer at absorbing rainfall, becoming more anaerobic over time without ameliorative action [14], which can affect microbial activity, subsequent nutrient cycling and greenhouse gas emissions. Carbon dioxide (CO2) is produced through many microbial processes [15] and can spike immediately after ploughing due to the flush of CO2 released from the mixing of the microbial community with decomposable substrates and aerated voids produced through tillage [16]. Pore space and pore connectivity allow for oxygen exchange within the soil, and when these are reduced, oxygen will deplete more rapidly leading to anoxic conditions [15] changing microbial activity. The microbial process of denitrification produces nitrous oxide (N2O) and is greatest in wet conditions, so less plant available nitrogen can be found in the soil and more nitrogen is lost to the atmosphere as N2 and N2O [17]. As N2O has a global warming potential 298 time higher than that of CO2 [18], compaction has implications for global warming emissions as well as soil health and productivity.

The efficacy of three methods for mitigating compaction damage was compared with the direct drilled control to see, not only the impact of these methods on crop production, but also their impact on soil health and greenhouse gas emissions. Ploughing was used as the conventional cultivation method for alleviating topsoil compaction. As ploughing aerates the soil profile, it can accelerate the loss of soil organic carbon (SOC) to the atmosphere as CO2, and destroy soil aggregates, exposing organic carbon for mineralization [19]. The physical process of running a plough through a soil can also have a detrimental effect on earthworm populations [13,20]. Both SOC and earthworm numbers have beneficial impacts on aggregate stability and soil structure [21], improving infiltration and resilience to future compaction. Low disturbance subsoiling (LDS) can be used as an alternative method of compaction alleviation, particularly in the subsoil layer, as the topsoil remains undisturbed. LDS theoretically has lower impact on CO2 emissions due to the non-inversion nature of the cultivation, reducing the mixing and oxygenation of SOC, and potentially reducing damage to earthworms in the topsoil layers.

Due to poor root exploration in compacted soils and microbial processes occurring in waterlogged soils, there can be lower access to nutrients for plants [7,22]. Mycorrhizal association has been suggested to benefit plants in these conditions, as the excess hyphae network can scavenge nutrients from a larger volume of soil [23,24]. As a final compaction alleviation method, a mycorrhizal inoculant was introduced to help plants overcome the detrimental effects of compaction on nutrient acquisition. The overall aim of the study was to identify the detrimental impacts of topsoil compaction, and to compare methods of alleviating this compaction in terms of their impact on soil compaction, earthworm populations, plant productivity and greenhouse gas emissions.

#### **2. Materials and Methods**

The experimental area was set up in October 2017 at the Allerton Project—a 300 hectare mixed arable and livestock research, demonstration and education farm (Game & Wildlife Conservation Trust, Fordingbridge, Hampshire, UK), at Loddington, Leicestershire, UK (N 052◦36 53 W 00◦50 31; 186 m a.s.l). Soils are predominantly a heavy clay loam, UK soil series: Denchworth, texture 47% clay, 31% silt, and 22% sand, soil organic matter 4.2%. To create compaction in the field, a tractor (Massey Ferguson 7720, approximate weight

8 tonnes) was driven across part of the field (100 m × 50 m), so that every area of the plots had been passed over by a tractor wheel twice. The compaction was checked using a cone penetrometer (SC 900, Field Scout, Aurora, IL, USA), taking an average of 10 measurements per plot, and showing an average of 15% higher compaction measured across 45 cm depth, that peaked at an increase of 32% at 7.5 cm depth. Penetration resistance measurements were repeated 4 times across the year.

Plots were arranged across the compacted area in randomized blocks (with tramlines excluded from the experimental treatments), measuring 6 m wide and 40 m long, giving an area of 240 m2. The effectiveness of cultivation at alleviating the compaction was tested using four treatments: plough, low disturbance subsoiler (LDS), mycorrhizal inoculant (AMF), and a no cultivation direct drilled control. Cultivations took place each year in autumn. Plough plots were ploughed to a depth of 25 cm, then disked to a depth of 10 cm (Väderstad carrier); LDS plots were subsoiled to a depth of 30 cm; AMF plots received a granular application of inoculant SR1:Cereals (Plantworks Ltd., Sittingbourne, UK) drilled with the crop at a rate of 10 kg/ha; while direct drill plots only received a straw rake before drilling. All crops were established using a direct drill (Eco M, Dale Drills, Market Rasen, UK) and standard farm practice was used for the application of manufactured fertiliser and plant protection products, which was consistent across all plots. Following cultivations in October 2017, *Hordeum vulgare* was planted across all plots and harvested in July 2018. The compaction and cultivation treatments were repeated in October 2018 keeping the same plot structure and *Vicia faba* was planted across all plots and harvested in September 2019.

Topsoil bulk density 0–10 cm was measured yearly in spring using a bulk density ring (10 cm depth, 5 cm diameter); three measurements were averaged per plot. Yield was taken from the combine as each plot was harvested. Plant biomass was also taken before combine harvest, by cutting three 0.25 m2 quadrats per plot, and drying the biomass in an oven at 70 ◦C until a stable weight was achieved. Earthworm abundance was measured using three replicates of 20 × 20 × 25 cm soil blocks per plot that were removed by spade. Soil was sorted by hand and all worms were counted and weighed.

Greenhouse gas measurements were taken monthly across the two cropping seasons using an FT-IR gas analyser (DX4040, Gasmet, Helsinki, Finland), set to measure CO2 and N2O simultaneously, with a 20 cm soil survey chamber attached (Li-cor). Plastic rings (20 cm diameter) were placed in the soil to a depth of 10 cm, allowing a 15 cm lip above the soil, at least 48 h before the first measurement. The chamber formed an airtight seal when placed on top of the rings. Gas flux was measured over 10 min, with the machine set to average measurements over 60 s. The initial 4 min were discarded to allow for gas equilibration in the system, and gas flux was calculated from the increase in gas concentration measured over the remaining 6 min. N2O was multiplied by 298 to give an equivalent global warming potential to CO2 to make comparisons between these two greenhouse gasses [18].

Statistical analysis used the Genstat software package [25]. A one-way ANOVA was used for all statistics, with the exception of the penetration resistance analysis. Where multiple measurements were taken, a repeated measures ANOVA was used. For penetration resistance, a principal component analysis (PCA) reduced the dimensionality of data, so comparisons between treatments at all depths could be made. PC1 (containing 75.53% of the overall variation) was used in a repeated measures ANOVA to test between treatments over the multiple measurement times.

#### **3. Results**

Penetration resistance showed significantly higher compaction in the uncultivated (AMF and control) plots (*p* = 0.002), which was mostly due to differences within the 7.5–2 5 cm depth range (Figure 1). There was also a significant impact of measurement time (*p* = 0.008) due to variation in soil condition over the year.

**Figure 1.** Penetration resistance (kPa) measured 0–45 cm depth through the soil profile. Graph shows average ± SE of all readings taken across the two years of measurements.

Bulk density (0–10 cm) measurements only showed significant results in the first year. Bulk density was lower in the ploughed plots, but surprisingly, significantly higher in the LDS plots (*p* < 0.001) (Figure 2). Bulk density measurements taken in the second year followed the same trend, with plough the lowest and LDS treatment as the highest, but this was not significant.

**Figure 2.** Soil bulk density (g cm<sup>−</sup>3) measured in the topsoil 0–10 cm in 2018. Bars show mean <sup>±</sup> SE. Letters denote significant differences at *p* < 0.05.

Despite the measurable compaction, it was not strong enough to influence yield, with no difference seen in crop yield seen in the two years. For the 2018 barley (*Hordeum vulgare*) crop, overall plant biomass was significantly (*p* < 0.001) higher in the two cultivated plots (Figure 3), but no biomass differences were seen in the subsequent bean crop (*Vicia faba*).

**Figure 3.** Barley crop (*Hordeum vulgare*) plant biomass measured in May 2018. Bars show mean ± SE. Letters denote significant differences at *p* < 0.05.

Earthworm numbers were higher in the non-cultivated plots (AMF and control) in both years (*p* = 0.046) (Figure 4). There was a highly significant difference between years (*p* > 0.001), with 2019 having less than half the number of worms counted in 2018 (average 411 ± 65 in 2018, 172 ± 28 in 2019).

**Figure 4.** Average earthworm number (per m2) measured in 2018 and 2019, to a depth of 25 cm. Bars show mean ± SE. Letters denote significant differences between cultivation treatments for both years at *p* < 0.05.

N2O and CO2 were measured monthly during cropping. Initially, repeated measures ANOVA showed no significant treatment effects for CO2 (*p* = 0.076), however, splitting the results by season resulted in significant treatment effects for CO2 flux in the winter months (*p* = 0.034), with ploughed plots having significantly higher CO2 emissions (Figure 5). Initial N2O emissions analysis showed significant treatment differences (*p* = 0.046), with significant differences between sampling times (*p* < 0.001) and a significant interaction between treatment and time (*p* = 0.033). Further investigation showed the interaction was due to much lower N2O emissions during the warmer drier summer months. Breaking the analysis down into summer months (June, July and August) and all the other months (referred to as winter for simplicity), gave significant treatment effects for winter months (*p* = 0.037), with the AMF and the control plots showing much higher N2O emissions

(Figure 6), but no significant effects were seen in summer months due to the overall lower emissions. The CO2 and N2O results were combined to give total green-house gas emissions for winter. The combined gasses showed no significant treatment differences in total gas fluxes recorded in winter (*p* = 0.595), although the composition of the gas fluxes changes between the plots.

**Figure 5.** CO2 flux measured monthly during 2018–2019 cropping and averaged over summer (June, July, August) and winter (all other months). Bars show mean ± SE. Letters denote significant differences between winter treatments at *p* < 0.05.

**Figure 6.** N2O flux measured monthly during 2018–2019 cropping and averaged over summer (June, July, August) and winter (all other months). N2O flux is displayed as CO2 equivalent by 298. Bars show mean ± SE. Letters denote significant differences at *p* < 0.05.

#### **4. Discussion**

Direct drilling is established as a management practice that can improve aspects of soil health by leaving the soil undisturbed, which helps build soil organic matter and soil biology, such as earthworm populations. However, it does require soil to be in a fit state for conversion to direct drilling; compaction is a common problem across agricultural land, and can lead to issues including reduced root growth, reduced water and nutrient uptake and overall reduced productivity if not resolved with some form of compaction alleviation. This study compared the effectiveness of ploughing, low disturbance subsoiling and mycorrhizal inoculation as compaction alleviation methods to direct drilling over a compacted area.

Uncultivated plots, AMF and the control plots had significantly higher compaction in the 7.5–20 cm range (Figure 1), when measured using a penetrometer. This is unsurprising, as this is the depth of soil that would have been influenced by the plough and the LDS cultivations. High penetration resistance scores of 1000–2000 kPa are linked with slower root elongation rates [26] and thicker roots, due to the increased pressure needed to penetrate the soil [27], which can be detrimental to plant growth as they require extra energy to explore the soil. A lower proliferation of roots can also result in reduced nutrient and water access for the crop. However, in this experiment compaction only reached above 1000 kPa at a depth of approximately 20 cm (Figure 1), suggesting that compaction in the topsoil where cultivation was used was not large enough to elicit a yield response. In 2018, a plant biomass response was seen in the barley crop, with smaller plants in the uncultivated plots (Figure 3). Previous studies have suggested that monocot crops are more capable of tolerating compaction than dicot crops [28], which could explain the difference between the response in the barley and the bean crop. Bulk density measurements taken in the top 10 cm of soil showed that ploughing resulted in the least densely packed topsoil, while LDS, surprisingly, resulted in the highest compacted topsoil (Figure 2). Subsoilers are designed to alleviate compaction at lower levels, leaving the topsoil relatively undisturbed. However, in some cases, subsoilers have been recorded to increase the compaction at the soil surface [29].

Earthworm numbers have been linked to improved infiltration [30], plant rooting depth [31], aggregate stability [21] and overall plant production [32], making them an excellent indicator of soil biological health [33]. Previous studies have suggested that earthworm populations diminish under cropping compared to pastureland, and under tillage compared to untilled cropped systems [34], which has been attributed to the mechanical damage and destruction of the earthworm habitat [20]. However, there is evidence that soil conditions such as high bulk density and low soil pore space caused by compaction can have adverse effects on earthworm populations, sometimes reducing numbers in uncultivated systems [35]. In the present study, earthworm numbers were significantly reduced under the cultivated treatments LDS and plough, across both years measured (Figure 4), suggesting mechanical damage had reduced the earthworm population, with potential detrimental effect on soil health and plant productivity in these plots. This highlights the trade-off between cultivation to alleviate the damaging effects of compaction, with the disturbance this causes on soil fauna needed for healthy soil processes. A long period of drought in 2018 is likely to have been a factor in the reduction in earthworm abundance between years. As the climate changes and the likelihood of prolonged droughts increases, these deleterious effects on earthworm populations will continue [36].

Direct drilling can reduce CO2 emissions and lead to an overall accumulation of SOC, due to an increase in aggregate stability and a change in chemical composition of carbon to more recalcitrant forms [37]. N2O is also a greenhouse gas emitted from soils, but with a far higher global warming potential than CO2 [18]. The production of N2O is primarily through denitrification, which increases when water-filled pore spaces within soils are around 65–75% [38]. As direct drilled soils generally have a greater bulk density, particularly if newly converted or previously compacted as in the present experiment, water-filled pore space is often higher favouring denitrification [39]. Bulk density measurements taken in this experiment were significantly lower in ploughed plots, suggesting that water-filled pore spaces would be similarly lower under the ploughed treatment (Figure 3). Higher bulk density was seen in the LDS plots, but this was only measured in the top 10 cm, as the LDS is designed not to disturb the topsoil, but to alleviate compaction at lower depths; there may still have been higher pore space lower down in the soil profile, which the penetration resistance measurement confirms (Figure 1).

Greenhouse gas flux measurements showed overall higher CO2 emissions in the summer months (June, July and August), when soil activity is at its highest due to warmer

temperatures. No significant treatment differences were seen in CO2 emissions between treatments in these warmer months, but when all other months were analysed together, significantly higher CO2 emissions were seen from the two cultivated treatments, LDS and plough (Figure 5), due to the mechanical stimulation of organic matter breakdown in the soil [16]. In contrast, N2O was produced at a much higher rate during the winter months than the summer months. During the winter months, there was significantly higher production of N2O under the two non-cultivated treatments, AMF and control (Figure 6). These results are similar to those seen by Gregorich et al. [40], who found an increase in N2O production in compacted soils, which corresponded to precipitation and high soil water content and was not seen in uncompacted treatments under the same conditions. Similarly, additional experiments at the site of our experiment, which used direct drill treatments without prior compaction have not shown this increase in N2O flux (data not shown). Therefore, the increase in N2O flux seen within this experiment was attributed to the reduced pore space, and subsequent increased water-filled pores in the compacted soil during the wetter winter months, causing an increase in denitrification activity and N2O emissions. This has implications for compacted soils exacerbating N2O emissions under future climate predictions of warmer wetter winters [10].

#### **5. Conclusions**

Overall, the efficacy of three methods of mitigating compaction damage to soil health and greenhouse gas emissions were compared with a direct drilled control. Two methods tried to improve soil structure and reduce compaction mechanically in situ, whilst the third, a biological method, attempted to reduce the impact of compaction on plant growth and nutrient acquisition. This study highlighted that compaction alleviation techniques differ in their efficacy as well as differing in their impact on soil health and greenhouse gas emissions. Earthworm abundance, a key indicator of soil health, was significantly reduced in the mechanical alleviation treatments, whilst emissions of CO2 also increased. However, the link between compaction and increased N2O emissions during wetter months is concerning, as seen in the AMF treatment and the direct drill control. Considering around 30% of soils in Europe are at risk (or susceptible) to compaction [4] and that winter rainfall is expected to increase due to climate change [18], greenhouse gas emissions may increase, dependent on agricultural (mis)management. This study highlights the importance of understanding how to alleviate compaction if we want to reach our climate emission goals and become net-zero within agriculture by 2040.

**Author Contributions:** Conceptualization, C.S. and F.C.; methodology, F.C. and J.B.; formal analysis, J.B.; data curation, J.B.; writing—original draft preparation, J.B.; writing—review and editing, C.S. and F.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This project has received funding from the European Union HORIZON2020 Programme for research, technological development and demonstration under grant agreement no. 677407.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data that support the findings of this study are available from the corresponding author upon reasonable request and approval from the study site representatives.

**Acknowledgments:** Thanks to Plantworks Ltd., Kent, UK, for providing the SR1 Cereals inoculant. Thanks to Gemma Fox for technical support in taking measurements and Phil Jarvis and Oliver Carrick for technical support in creating and maintaining plots.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


## *Article* **Soil Water Retention as Affected by Management Induced Changes of Soil Organic Carbon: Analysis of Long-Term Experiments in Europe**

**Ioanna S. Panagea 1,\*, Antonio Berti 2, Pavel Cermak ˇ 3, Jan Diels 1, Annemie Elsen 4, Helena Kusá 3, Ilaria Piccoli 2, Jean Poesen 1,5, Chris Stoate 6, Mia Tits 4, Zoltan Toth <sup>7</sup> and Guido Wyseure <sup>1</sup>**


**Abstract:** Soil water retention (SWR) is an important soil property related to soil structure, texture, and organic matter (SOM), among other properties. Agricultural management practices affect some of these properties in an interdependent way. In this study, the impact of management-induced changes of soil organic carbon (SOC) on SWR is evaluated in five long-term experiments in Europe (running from 8 up to 54 years when samples were taken). Topsoil samples (0–15 cm) were collected and analysed to evaluate the effects of three different management categories, i.e., soil tillage, the addition of exogenous organic materials, the incorporation of crop residues affecting SOC and water content under a range of matric potentials. Changes in the total SOC up to 10 g C kg−<sup>1</sup> soil (1%) observed for the different management practices, do not cause statistically significant differences in the SWR characteristics as expected. The direct impact of the SOC on SWR is consistent but negligible, whereas the indirect impact of SOC in the higher matric potentials, which are mainly affected by soil structure and aggregate composition, prevails. The different water content responses under the various matric potentials to SOC changes for each management group implies that one conservation measure alone has a limited effect on SWR and only a combination of several practices that lead to better soil structure, such as reduced soil disturbances combined with increased SOM inputs can lead to better water holding capacity of the soil.

**Keywords:** soil organic carbon; soil-water content; no-till; reduced tillage; manure; compost; soil care

#### **1. Introduction**

Soil water retention (SWR) is a measure of how much water a particular type of soil can retain. It is an important soil property related to the distribution of pore space and, thus, is highly dependent on soil structure and texture, as well as on other related properties such as soil organic matter (SOM) [1]. SWR is critical for crop growth with a profound influence on crop yield and crop failure and acts as the main source of moisture for the soil's biota, which contributes to land productivity and biological soil health.

The relationship between the volumetric soil water content (θ) and the pressure head (or matric potential head, h) is described by the soil water retention curve (WRC), also

**Citation:** Panagea, I.S.; Berti, A.; Cermak, P.; Diels, J.; Elsen, A.; Kusá, ˇ H.; Piccoli, I.; Poesen, J.; Stoate, C.; Tits, M.; et al. Soil Water Retention as Affected by Management Induced Changes of Soil Organic Carbon: Analysis of Long-Term Experiments in Europe. *Land* **2021**, *10*, 1362. https://doi.org/10.3390/land10121362

Academic Editor: Manuel López-Vicente

Received: 25 October 2021 Accepted: 8 December 2021 Published: 9 December 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

known as the soil moisture characteristic curve or pF curve [2]. This curve is characteristic for different soils and is used to predict soil water storage for applications in agronomy, ecology, hydrology and many other soil-related sectors [3–6], as well as in earth systems models [7,8].

For the determination of the WRC, different field and laboratory methods exist [9–11]. Analytical models [12] or regression equations—empirical formulas called pedo-transfer functions (PTF)—are used to predict the WRC values from easily measured or already available soil properties [13–17]. The majority of the PTFs for estimating the WRC use soil texture, bulk density and SOM content as predictors [1,13,17], although the necessity of the latter has been questioned [18] or shown to improve the estimations only for specific soil water potentials [19]. For modelling purposes, analytical functions are fit to a set of observed h-θ values used to represent the continuous WRC. The most common retention functions have been presented by Brooks and Corey [20] and by van Genuchten [21].

The dependence of the SWR on texture and structure has been widely researched and demonstrated [22]. The dependence of the SWR on SOM content has also been proven [23,24], but the results on the quantitative influence of SOM are contradictory and vary with texture, pressure head and soil organic carbon (SOC) content as such [24–26] and therefore need to be further evaluated [27]. Analysing the effect and relationship of SOC content on SWR taking into account different soil textures has shown that the sensitivity of the SWR to SOC changes depending on the soil textural classes and on the SOM content itself [24,26]. For the same SOC increase, soils with coarser textures and low SOC contents present a larger increase in water retention than the finer soils [26], which may also present a decrease [24]. In contrast, for soils with high SOC contents the water retention increases for all textural classes, especially for sandy and silty soils [24]. Nevertheless, as pointed out in a review by Minasny and McBratney [26], a 1% absolute mass increase of SOC (10 g C kg−<sup>1</sup> soil) has a limited effect on the SWR and can increase the available water capacity by up to 1.16% volumetrically. They also found that the effect is relatively larger for sandy soils. A change in the SOC content also influences the water content at the different pressure points in a different way [26], with field capacity (FC at −33 kPa or pF 2.5) to present higher sensitivity than the wilting point (WP at −1500 kPa or pF 4.2) [24]. Nevertheless, the use of SOM as an auxiliary predictor for the SWR through PTFs has been proven to be redundant when bulk density is also used as a predictor [18].

The different management practices applied in cropping systems affect the soil structure and soil composition, and consequently the SWR and other physical soil properties. Organic and conservation farming (defined as a farming system that promotes practices about maintenance of a permanent soil cover, minimum soil disturbances and crop diversification [28]) can increase the soil water storage through better soil aggregation and improved soil structure [29], but in some cases, the conventional systems yield higher water contents as a result of higher microporosity [30]. The SOC decreases when SOC losses due to erosion or/and mineralization, which can be stimulated also through soil tillage, exceed the organic carbon inputs coming from the addition of exogenous organic inputs (compost or manure) and organic inputs from crop residues (shoots and roots) [31].

Adding more exogenous organic materials such as compost or farmyard manure and the incorporation of crop residues into the soil above the SOC mineralization rate causes an increase in the total SOC in most cases [32–34]. However, the quality and stage of decomposition of exogenous organic materials affect how much of this added carbon remains as stable organic carbon in the soil [31,35]. The degree of maturation of manure and the composition of the compost greatly affects retention rates of organic carbon in the soil [36]. The addition of exogenous organic material increases the volumetric water content at most pressure heads, mainly because of the increase in total porosity [37] and the increase in total SOC. Mulching with or incorporating the crop residues has been proven to significantly impact the SWR in the wet range (pF < 2), but not in the dryer range (pF > 3) [34,38,39].

Reduced or no-tillage is often advocated to increase the SOC in the topsoil, which is important for a good structure, increased infiltration and reduced soil erosion rates when compared to conventional ploughing, but the results are controversial when considering the whole soil profile [31,40,41]. Conservation tillage has been proven to improve chemical soil properties such as SOM or physical soil properties such as aggregate stability of the topsoil, but the effects on soil water content are, in many cases not, significant [42] or controversial [43]. Sometimes the results are not solely dependent on soil tillage type but vary with matric potential [44–46]. The water content tends to be larger in the higher pressures (pF < 1 or wetter part) for conventional tillage when compared with conservation or no-tillage but in the smaller potentials the water content is larger for the conservation or no-tillage practices [44,46]. There are also cases where significant differences in the water content are present only in the more negative (pF > 3 or dryer) matric potentials [47]. López et al. [48] and Kargas and Londra [47] found that reduced and conventional tillage result in similar water content values, whereas no-tillage leads to lower values of water storage. On the other hand, Bescansa et al. [44] found that soil water content was higher in the no-tillage fields when compared to conventional tillage, especially in the drier condition because of the higher available water content caused by increased SOC content and changes in the pore distribution of the untilled soils.

Although previous studies have investigated the effects of management on the soil chemical and physical properties, less attention has been given to the link between combined management practices and SWR. In addition, most studies include a limited number of management practices and intensities and are not replicated in multiple agroecosystems and/or study regions that cover broad environmental gradients (i.e., climatic conditions and soil properties), possibly due to the logistical constraints associated with extensive field work. Finally, a comparison between published data is frequently hindered by methodological discrepancies between studies. To this end, studies that investigate broad management practices and intensities in multiple agroecosystems and regions with distinct environmental conditions are well needed to understand the interactions between soil structure, organic carbon, and water retention.

In our study, we compared seven long-term (8–54 years) experimental setups by sampling the topsoil with identical methods and analysing all samples in the same laboratory. The field experiments have been set up with specific and different objectives, but all together they cover a broad range of tillage practices, fertilization, additions of organic materials and management of crop residues. The objective of this study was to evaluate and quantify comprehensively the effect of different management practices on SOC content and their impact on the water-holding capacity of the soils.

#### **2. Materials and Methods**

#### *2.1. Experiments' Descriptions*

Topsoil samples were collected from seven different long-term agricultural experiments with different treatments in 5 European countries (the towns, countries, coordinates, start year of the experiment and main soil type of the sites are given in Table 1). In each country, the experiments were setup with different objectives and under different environmental conditions. Although the diversity of the experiments makes it challenging to combine them, they offer a wide range of representative management practices and pedo-climatological conditions. As the original experiments attempted to answer different scientific questions, they include several management treatments. For this research, a subset of treatments was selected from each experiment to include treatments from three main categories. The first category includes different soil tillage treatments (CZ, HU\_2, UK), the second category comprises the addition of different types of exogenous organic materials (BE, IT\_1c, IT\_1p), and the third category deals with the incorporation of crop residues in the topsoil (HU\_1, IT\_2c, IT\_2l). The experiments in Italy are conducted on two different soil types each and, in this study, are analysed as separate experiments: a clay and an initially peaty soil for IT\_1 (i.e., IT\_1c and IT\_1p) and a clay and a loamy soil

for experiment IT\_2 (i.e., IT\_2c and IT\_2l) resulting in nine experiments in our study. The selected treatments per experiment are presented in Table 2. At the five study sites, an identical sampling procedure was performed for determining the WRC and SOC.


**Table 1.** Description of the study sites.

**Table 2.** Details of the soil treatments in the various experiments ‡ Randomized complete block design (RCBD); ϕ Split plot-randomized complete block design (Split Plot-RCBD).



**Table 2.** *Cont.*

#### *2.2. WRC Points Determination*

To estimate the water content at the different points of the WRC, three undisturbed topsoil samples (positioned in the middle of 0–15 cm layer) were collected from each experimental plot (apart from Italy where the plots are too small and only one ring sample per plot could be collected) with the use of a Kopecky ring, of a known volume (100 cm3). The 177 soil samples taken at different dates do not represent an equal number for each experiment and experimental plot (details about the number of samples per experimental plot are shown in Table 3). The top organic layer was first removed and with the use of suitable equipment (i.e., hammering holders and plastic hammer) to minimize soil disturbances the rings were pushed into the soil to collect the samples, which were stored afterwards at room temperature until analysis.

**Table 3.** Sampling details per experiment.


In this paper, we use pF to indicate the soil water potential. The pF is defined as the decimal logarithm of the absolute value of pressure head expressed in cm (pF = log10|h|).

The drainage or drying cycle was used for the determination of the volumetric water content at moisture tensions (suction) from pF 0 to pF 4.2. Sandboxes were used for the determination of the water content at the lower suction values (pF 0.0, pF 1.0, pF 1.8 and pF 2.0) and pressure plates cells (Soilmoisture Equipment Corp.) were used for the determination of the sample water content at pF 2.7, pF3.4 and pF 4.2 [55].

#### *2.3. OC Determination*

The largest component and easiest indicator of SOM status to measure is the SOC content and it is used in this report both to present the results and when we refer to content

changes in the existing literature. SOC content was determined by dry combustion and mass spectrometry elemental analysis (Carlo-Erba EA 1110, Thermo Scientific, Waltham, MA, USA). Fresh disturbed field topsoil samples (0–15 cm) were taken with a sharp shovel from various spots within each experimental plot, mixed and directly broken to pass a <8 mm sieve. Soil samples were then stored in plastic containers to avoid compaction and disturbance during transportation and then stored in the refrigerator until air-drying could be carried out. All samples were air-dried at 40 ◦C until a constant mass was achieved and stored in a dark and dry place at room temperature. A subsample of the bulk soil was taken with a soil sample splitter to allow for a random representative sample, crushed manually to a homogenized powder and weighted into an Ag capsule. To determine only the carbon present in organic form carbonates were removed by adding HCl (35%). After drying at 40 ◦C the soil samples were loaded into the autosampler for combustion with oxygen at 800 ◦C with the presence of a catalyst. The organic carbon (OC) reacts to carbon dioxide (CO2) which is quantified by infrared absorption spectroscopy and the mass percentage is determined

#### *2.4. Statistical Analysis and Visualization Tools*

The statistical data analysis was performed using R-Studio, R version 3.6.1 [56]. Oneway Analysis of Variance (ANOVA) was carried out with the R software [57] to test for differences between treatments. Estimated marginal means (also known as leastsquares means) by factors were computed by the least square method using the package "emmeans" [58]. Graphs were produced with the package "ggplot2" [59]. In the present work, statistical significance is assumed at *p* < 0.05. The assumptions of normality and homoscedasticity of the residuals were assessed by visual inspection of Q-Q plots and by plotting the normalized residuals against the fitted values.

#### **3. Results and Discussion**

#### *3.1. Soil Organic Carbon*

The total SOC as determined in the bulk soil is presented in Figure 1. Generally, SOC is relatively sensitive to the different management treatments and statistically significant differences are present among most of them. In the field experiments sampled, values of total organic carbon vary between 6 and 56 g C kg−<sup>1</sup> soil (0.6–5.6%). The highest values observed is in the IT\_1p experiment, in which an initially peaty soil was treated with different levels of manure. The lowest values were observed in the IT\_2l experiment, where removal of residues took place in a loamy soil.

In the organic input experiments, as was expected [33], the higher amount of manure or compost resulted in a higher SOC. Nevertheless, statistically significant SOC differences are observed only between the higher levels of additions and the controls, which in all cases are those with only mineral inputs apart from the IT\_1p and IT\_1c in which the control is unfertilized treatment. In the IT\_1p experiment, no statistically significant differences are observed but despite this, treatments with organic inputs show a trend that follows that same assumption i.e., higher input of organic materials leads to higher SOC. In this experiment, where agricultural management was established in initially peaty soil, there is a reduction in the total SOC over the years because of cultivation, but the decline is lower when manure is added, highlighting the importance of organic fertilizers in maintaining soil fertility in the long term [60].

In the tillage experiments, the treatments with the minimum soil disturbance present statistically significantly higher values of SOC content in the 0–15 cm layer than the treatments where conventional tillage took place, since fresh organic material will be kept concentrated at the topsoil [61,62]. As a result of the no- or reduced- tillage practices, apart from the increased organic inputs from the crop residues which are concentrated in the top layer, and the roots that remain intact in the soil, the carbon outputs by mineralization are reduced in the no-tillage systems. Reduced tillage systems present similar levels of mineralization with the conventional ploughing [61,63] but, according to recent evidence, reduced

tillage presents lower CO2 emissions [64] or higher levels of mineralizable carbon [65] when compared to conventional ploughing, indicating reduced mineralization because of minimal disruption of the stable aggregates. A more stable soil structure and improved aggregation also lead to reduced losses of fertile carbon-rich topsoil because of reduced water erosion. Nevertheless, it should always be kept in mind that only the topsoil 0–15 cm was sampled. Carbon stock changes may be observed in the no- or reduced-tillage experiments within the top layer sampled but if the whole soil profile is considered, conclusions cannot be drawn from our analysis. There is recent evidence to show that when the time since the adoption of the no-tillage system is considered (i.e., at least 6 years application of no-till), the increase of the carbon content in the top layers (0–20 cm) and the no change of the SOC in the deeper layers of the soil profile lead to an overall increase in the carbon stocks under no-tillage [41]. On the other hand, there is also evidence that when the entire soil profile is analysed, soil cultivation methods do not affect the SOC quantity but rather redistribute it in the profile [31,61,66–68].

**Figure 1.** SOC content in the topsoil (0–15 cm) for each study site (see also Table 1 for the codes and Table 2 for a description of the soil improving treatments). Error bars represent the standard error (n -number of treatments replications- is denoted in Table 3 for each experiment).

In the HU\_1 experiment, the incorporation of the straw and stalks causes a small but statistically significant difference in the SOC content when compared with the only mineral fertilized treatment and also significantly increased soil aggregate stability [69], supporting this way the physical soil condition. In contrast, in the IT\_2c and IT\_2l experiments, small or no statistically significant changes are noticed between the treatments. This, on the one hand, confirms the little potential of crop residues for soil improvement [70] and, on the other hand, raises questions if only the incorporation of crop residues without any kind of pre-processing like composting, conversion to biochar or the parallel use of other conservation measures such as reducing tillage, can contribute to the build-up of SOC. Indeed, in the same Italian experiment Dal Ferro et al. [71] recently found that

residues incorporation seems to be effective in SOC storage only when coupled with minimum-tillage practice.

The results indicate that the SOC is a sensitive and good indicator to monitor changes caused in the soil quality by management practices in the long-term. The results should always consider the sampling depth, especially when tillage practices with different tillage depths are compared.

#### *3.2. WRC*

In Figure 2, the soil water content as a function of the matrix pressures (expressed in pF) is shown for all the study sites as measured in the laboratory conditions which may be different from field conditions. Although all trends looked consistent, whereby more SOC consistently meant a slightly higher water content for a given pF, we detected only a few statistically significant differences (in only 3 out of 9 experiments and in a limited number of pressure points).

**Figure 2.** Soil water content at the different pressure points for each study site (see also Table 1 for the codes and Table 2 for a description of the soil improving treatments). Error bars represent the standard error (n -number of treatments replicationsis denoted in Table 3 for each experiment).

> In the soil tillage experiments, there are no statistically significant changes in the soil water content among the different levels of tillage, but a pattern is observed. In the CZ and UK experiments, we observe that in the higher matric potential range the treatments with the higher levels of disturbances present higher water content, whereas in the lower pressures the pattern is opposite: here, the higher the soil disturbance the lower is the water content. This happens because in the higher pressures (less negative) the macropores and capillary forces play an important role in the total water content, whereas in the lower matric potential the adsorption on the soil particles and the SOM work as water storage pools [72]. These results follow McVay et al. [42], who also observed that tillage methods

do not significantly affect the WHC of most of the soils analysed and reported changes only in the higher matric potentials [44,46,47], indicating that tillage mainly influences the volume of larger pore sizes, dominantly influencing water infiltration and aeration of the soil.

Higher organic matter input always led to modestly higher water content at all pressures. Only in the BE experiment, there are statistically significant differences among the treatments in the higher pressures (less negative or wetter part) as a result of an annual addition of 45 tonnes of compost per hectare. It is important to point out that a yearly dose of 45 tonnes of compost is excessive as compared to normal practice. In the IT\_1c experiment, in which different levels of farmyard manure are applied in clay soil, the high variability in the water content among the replicates did not allow to detect statistically significant differences, but the same trend is observed i.e., higher organic matter input leads to higher water content at all pressures. In the same experiment with the peaty soil with initially 18% organic carbon, this trend is not noted. The results are consistent with Eusufzai and Fujii [37] who found that organic amended soils present increased water content, especially in the higher pressure points.

In the last experiment group in which the crop residues are incorporated into the soil after the cropping season, our analysis does not reveal consistent changes in the water storage capacity, or at least consistent trends to justify the reported findings in the literature that incorporation of residues impacts the water retention in the higher matric potentials [34,38,39].

An analysis of the effect of different practices in the whole WRC as calculated from the Rosseta version 3 model [17] is presented in Figure S2 of the Supplementary Materials; changes are observed only in the wetter part of the retention curve, whereas the dryer part is not affected by the different applied cropping systems.

#### *3.3. Water Retention as Affected by Carbon Changes and Management Practices*

In Figure 3, the percentage differences of water content in relation to the percentage differences of SOC content among the different plots of each experiment are presented as differences with the corresponding control treatment of the relevant block. When all experiments are analysed together, it is observed that the increase of total SOC over the period that each experiment is running, generally causes an increase in water content at all moisture tensions, especially for the higher matric potentials (wetter conditions) at which the regression relationship is also statistically significant (Table 4). This trend is less pronounced at the permanent wilting point (pF 4.2), where the impact is almost negligible. As a result of the negligible increase in the wilting point and the bigger increase in the water content at field capacity, the plant available water increases even with an 1% increase of the total SOC.


**Table 4.** F-statistic of the linear regression analysis. Significance codes of *p*-value: \*\* 0.05, \* 0.1.

**Figure 3.** Relationship of the percentage change in soil water content and percentage change in SOC content. The water content and SOC values represent the percentage differences between the different treatments' plots and the control plot of each block of the experiment. The different colours represent the water content change in the different pressure points of the WRC. (**a**): All the experiments plotted together. (**b**–**d**): Each experimental group plotted separately (**b**): addition of compost or manure, (**c**): tillage experiments, (**d**): residue management).

Comparing the results of the percentage differences of water content in relation with the SOC percentage changes in each experiments group, it is observed that the impact of increased carbon content on the water content does not only depend on the pressure point, texture, and organic carbon content, but also on the applied management practices. Management practices that increase bulk density (no- tillage, reduced OM inputs etc.) decrease the volume fraction of macro-pores, but at the same time increase the volume fraction of both micro- and meso-pores, resulting in an increase of the water content at lower matric potentials and a decrease under wetter conditions [73]. In the soil tillage experiments where the maximum carbon increase observed is about 0.65% as a result of practices that minimize topsoil disturbances, an increase in the soil water content is observed in the lower matric potentials (drier conditions) as a result of increased SOC and surface adsorption [72,74], as also observed from Bescansa et al. [44] and a smaller increase of the water content in the wetter conditions (saturation and pF 1) as a result of changes in the pore distribution and capillary forces [72,74]. These two conditions may lead to a decrease in the plant available water content (AWC) when cultivation practices with less soil disturbance are followed as also mentioned by Hill [75]. Indeed, as shown in Figure S1 in the Supplementary Materials, the AWC in the CZ experiment is statistically significantly lower in the zero-tillage treatment when compared with the minimum tillage, and in the UK experiment it is lower, but not statistically significant in the direct drilling treatment

in comparison with the conventional ploughing. In the experiments where exogenous organic material is added, a cca. 1% increase in SOC increases the soil water content under all applied pore water pressures but the increase is lower in the drier conditions and sometimes also a decrease in the soil water content is presented when compared to no addition of organic materials. The addition of organic material leads to increased macro aggregation and therefore increased meso- and macroporosity [37,73] and increased water content in these pores, resulting in less water available for storing in the micropores. A negligible increase in the dryer conditions denotes also that the increase of SOC does not increase the sorbed water or that it is counteracted by the increased macroporosity. Nevertheless, the negligible increase in the water content in the wilting point and the big increase in the water content at field capacity will lead to higher plant available water content (Figure S1—Supplementary Materials), something that impacts positively the plant growth. In the experiments where the residues have been incorporated in the topsoil, the soil water content decreases or remains similar, even if an increase of 0.3% of the total SOC is observed. Despite the long-term application, there were no large SOC changes following the incorporation of the residues. Building up SOC and simultaneously a stable soil structure might be more important than a large increase in SOC content, especially during wetter soil conditions.

There is a strong belief and impression by practitioners and advocates of conservation agriculture and organic farming that an increase in the total SOC increases water retention directly and substantially [76–78]. In this research, the water retention characteristics present the expected but modest trends. However, it is remarkable that even after 54 years of practices that increase SOC, the observed differences in the water content, especially in the lower pressures (drier conditions), are negligible from a practical point of view, and almost all not statistically significant. It was expected that in the lower water pressures (pF > 2.5) where the macropores and capillarity do not have an important impact and surface adsorption and SOC content seems to play the most important role in the soil's water content, the differences in SWR among the different treatments would be noticeable. The statistically significant linear relationship of the carbon change and the water retention mainly in the wetter conditions (Table 4) and not in the drier conditions suggest that the direct impact of the increased SOC on water retention is limited and the indirect impact stronger. The fact that the change of SOC affects the water content under the different matric potential in a different way and is statistically significant only in the higher matric potentials implies that the impact of SOC is indirect and is more linked to the changes in other soil parameters and most probably in soil structure and aggregation status.

#### **4. Conclusions**

We analysed different groups of management practices for improving soil quality as applied in long-term experiments in five European countries. We investigated their effect on SOC and the link with the capacity of the soil to retain water at different matric potentials. Our findings suggest that practices that minimize soil disturbances cause an increase in SOC in the topsoil but may lead to decreased plant available water content as a result of the increased water content at wilting point and a less profound increase in water content at field capacity, jeopardizing the crop yield. On the other hand, the different soil-improving management practices that increase the organic materials in the soil (both exogenous and incorporation of residues) contribute to an increase in the soil water availability for the crops, but not because of increased water holding capacity as a result of increased SOC. The addition of organic materials affects the soil structure, and it is more likely that the soil structure—as improved by the SOM—affects the water availability because of more macro and mesopores, rather than because of larger waterholding capacity per soil volume caused by a SOC increase. The better structure formed by higher amounts and more stable SOC and the increase in SWR are important factors leading to increased water infiltration, even under long-term rainy conditions, and promoting several soil functions such as less soil erosion minimised effects of extreme rainfalls and

droughts deeper rooting of the crops end enhanced crop productivity. The negligible effect of increased SOC under different management practices during drier conditions, and the increased effect in wetter conditions, implies that the indirect effects of SOC increase in the soil structure are more important and should be considered in future research.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/land10121362/s1, Figure S1: Plant Available Water Content as calculated from the difference between water content at field capacity (pF 2) and water content at wilting point (pF 4.2). Figure S2: WRC for each treatment. The lines represent the WRC as calculated from the Rosetta version 3 model with input the average silt clay and sand percentages, the average bulk density of the treatments as measured and the water content at field capacity and wilting point.

**Author Contributions:** Conceptualization, I.S.P. and G.W.; methodology, I.S.P.; formal analysis, I.S.P.; data curation, I.S.P.; writing—original draft preparation, I.S.P.; writing—review and editing, G.W., J.P., J.D., M.T., A.E., I.P., A.B., C.S., Z.T., H.K. and P.C.; visualization, I.S.P.; supervision, G.W., J.D. and ˇ J.P.; project coordination, G.W.; funding acquisition, G.W. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the project "SoilCare: Soil Care for profitable and sustainable crop production in Europe" from the H2020 Programme under grant agreement n◦ 677407.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data that support the findings of this study are available from the corresponding author upon reasonable request and approval from the study sites representative co-authors.

**Acknowledgments:** We thank all the study sites owners for permitting the use of the experimental fields where this research was done and providing all types of help during the sampling campaign. We thank the laboratory assistant of the Department of Earth and Environmental Sciences of the KU Leuven Lore Fondu for providing valuable guidance and help during the soil samples analysis. ISP thanks Antonios Apostolakis for the extensive discussions and guidance on the statistical analyses and concepts.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **Quantifying Cereal Productivity on Sandy Soil in Response to Some Soil-Improving Cropping Systems**

**Jerzy Lipiec 1,\* and Bogusław Usowicz 1,2**


**Abstract:** Little information is available on the effect of soil-improving cropping systems (SICS) on crop productivity on low fertility sandy soils although they are increasingly being used in agriculture in many regions of the world due to the growing demand for food. The study aimed at quantifying the effect of four soil-improving cropping systems applied on sandy soil on cereal productivity (yield of grain and straw and plant height) in a 4-year field experiment conducted in Poland with spring cereal crops: oat (2017), wheat (2018), wheat (2019), and oat (2020). The experiment included the control (C) and the following SICS: liming (L), leguminous catch crops for green manure (LU), farmyard manure (M), and farmyard manure + liming + leguminous catch crops for green manure together (M+L+ LU). To quantify the effect of the SICS, classic statistics and the Bland–Altman method were used. It was shown that all yield trait components significantly increased in the last study year (2020) under SICS with M and M + L + LU. All yield trait components were significantly lower in the dry years (2018–2019) than in the wet years (2017 and 2020). The relatively large rainfall quantity in May during intensive growth at shooting and the scarce precipitation during later growth in the dry year 2019 resulted in a significantly greater straw yield compared to the other dry year 2018. The values of Bland–Altman bias (mean difference between the particular SICS and the control) varied (in kg m<sup>−</sup>2) from <sup>−</sup>0.002 for LU in 2019 to 0.128 for M and 0.132 for M + L + LU in 2020. The highest limits of agreement (LoA) were in general noted for all yield trait components (the least even yield) in the most productive SICS including M and M + L + LU in the wet year 2020. The Bland–Altman ratio (BAR) values indicate that quantification of the effects of all soil-improving practices was most uncertain in the dry year 2018 for the grain yield and in the wet year 2020 for the straw yield and much less uncertain for the plant height in all SICS and study years. The results of this study provide helpful information about the effect of the SICS on the different yield trait components depending on the period of their application and weather conditions prevailing during the growing season.

**Keywords:** soil improving practices; crop response; weather conditions; Podzol soil; Bland–Altman statistics

## **1. Introduction**

Sandy soils cover globally approximately 900 million ha [1]. They occur in different regions across the world [2–4], particularly in arid or semi-arid regions [5]. In Poland, around 50 percent of soils developed from sands [6,7].

Sandy soils are characterized by low crop productivity. This is mostly attributed to a weakly developed aggregated structure [8], high saturated hydraulic conductivity and permeability and low water-retention capacity due to the high contribution of large pores between sand particles [9–11], low nutrient levels, and poor ability to store and exchange nutrients [1]. Furthermore, after rapid dewatering, the large pores become air-filled first and act as a barrier (discontinuity) to water flow through the smaller pores towards the plant roots in unsaturated soil conditions [12,13].

**Citation:** Lipiec, J.; Usowicz, B. Quantifying Cereal Productivity on Sandy Soil in Response to Some Soil-Improving Cropping Systems. *Land* **2021**, *10*, 1199. https://doi.org/ 10.3390/land10111199

Academic Editors: Guido Wyseure, Julián Cuevas González and Jean Poesen

Received: 19 October 2021 Accepted: 4 November 2021 Published: 5 November 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Another threat limiting crop production on sandy soils is their acidity due to the presence of acid from the post-glacial acidified parent material, leaching exchangeable base cations [14,15], and chemical N fertilization [16]. Soil acidity limits crop productivity by increasing Al3 + toxicity leading to production of short, thick, and shallow plant roots and deficiency of some nutrients in the soil solution [12]. Instead, sandy soils require rather low energy inputs for tillage [17] and warm up rapidly in the spring prior to the growing season to achieve the minimum soil temperature for plant growth [12].

Despite low fertility and quality, sandy soils are increasingly being used for crop production due to the shortage of agricultural land resources [1,18,19] as well as the growing population and demand for food [20,21]. However, arable farming on these soils require large amounts of irrigation and nutrient inputs [22,23] in many areas, which reduces the profitability of agricultural products.

There is a broad agreement that water and nutrient supply for plant growth in sandy soils can be improved by increasing organic matter content [1,24–28]. This is related to the fact that soil organic matter increases plant available water capacity [21,29] by reducing pore diameter [30] and improves the capability of soils to retain and exchange nutrient cations and hold hydrogen ions, thereby neutralizing soil acidity [31]. Furthermore, increase in soil organic carbon (SOC) content in sandy soils is responsible for variation in cation exchange capacity [1].

There are many soil-improving cropping systems to maintain or increase the SOM content. They include application of organic amendments and diversified crop rotation favoring formation of stable soil aggregates, which protect soil organic carbon (SOC) from mineralization [27,32]. Inclusion of legumes fixing nitrogen from the atmosphere in crop rotation reduces the need for mineral nitrogen fertilization, thereby increasing profitability in crop production [33–35], and is one of the ways to meet greening requirements [36]. Furthermore, these practices are important in terms of increasing cereal-based crop rotations that along with conventional treatments (plough) disintegrate soil organic matter by physical disturbance of the soil structure and stability [27]. Increasing the soil organic matter content is part of the global strategy to enhance carbon sequestration stocks, reduce chemical leaching [1,32,37,38], and create drought resilient soils to mitigate global warming effects [21,39].

The aim of the work was to quantify the effect of different soil-improving practices, including application of farmyard manure, liming, and catch crops, on cereal productivity of sandy soil in a 4-year experiment with the use of the statistical Bland–Altman method [40,41]. Plotting the yield differences between a given treatment and the control against their averages and determining the average difference (bias), limits of agreement, and confidence intervals in this method allow quantifying the direct effect of the examined soil-improving cropping systems on crop yields. The Bland–Altman method is widely used in medicine (e.g., [42–44]) and in some satellite research [45,46]. More recent studies showed usefulness of this approach to quantify pure effects of agricultural practices on crop yield and soil physical properties [47,48] and the agreement between methods for determining the Atterberg plastic and liquid limits of soils [49]. This study was inspired by recent literature reviews indicating that, despite their importance, sandy soils have received less research attention compared to other soils [1,5].

#### **2. Materials and Method**

#### *2.1. Study Area and Field Experiment*

The field experiment (350 × 35 m) was localized in a private farm in Szaniawy, Podlasie region, Poland (51◦58 56.5 N 22◦32 22.1 E) on Podzol soil [50] derived from sandy material of glacial origin. The soil contains 62.9% of sand (2–0.05 mm), 34.8% of silt (0.05–0.002 mm), 2.2% of clay (<0.002 mm), and 0.8% of organic carbon and has pH 4.0 (in H20) and cation exchange capacity 12.3 cmol kg−1. Such and similar soils predominate in the region and in Poland. A randomized field-experiment was established in autumn 2016 and conducted for four years with the following spring cereal crops: oats (*Avena*

*sativa* L.) (2017), wheat (*Triticum aestivum* L.) (2018), wheat (2019), and oats (2020), which predominate in the crop rotation of the region.

The experiment included the following 5 treatments: (C) control, (L) liming with 5.6 t ha−<sup>1</sup> CaCO3 (applied once in autumn 2016), (LU) catch crops for green manure including yellow lupin (*Lupinus luteus* L.) with seeding rates in brackets (130 kg ha−1), serradella (*Ornithopus sativus*) (30 kg ha<sup>−</sup>1), and phacelia (*Phacelia tanacetifolia* Benth.) (3 kg ha−1) sown every year, (M) farmyard manure applied at 30 t ha−<sup>1</sup> every year in autumn), and (M + L + LU) manure (10 t ha−1) + liming with 5.6 t ha−<sup>1</sup> CaCO3 + catch crops for green manure including lupines with seeding rates in brackets (130 kg ha−1), serradella (30 kg ha<sup>−</sup>1), and phacelia (3 kg ha−1) (applied every year), except liming only in autumn 2016. Yellow lupine, serradella, and phacelia are common plants used for green manures in Poland. Each treatment had three replicate plots (35 × 20 m) separated by a 1.0-m margin between the plots. The grain yield, straw yield, and plant height were measured in nine one-square-meter sub-plots in each of the five treatments (three sub-plots × three replicate plots).

Stubble tillage (10 cm) using a cultivator plus tooth harrows was done after harvesting in all treatments (first half of August) and then catch crops were sown in treatments LU and M + L + LU. Next, mouldboard ploughing (20–25 cm) in late autumn and disking (10 cm) and tooth harrowing (6 cm) in spring (2nd half of March) were applied in all treatments (1st half of April) to prepare the seedbed for spring cereals. The autumn ploughing in M and M + L + LU also ploughed down the catch crops for green manure. Weed control and crop protection were carried out by herbicides, insecticides, and fungicides used in the farm where the experiment was conducted in the same manner in all treatments. All management practices were done using light wheel tractors (2.5 to 3.5 Mg mass) to minimize soil compaction effects on crop yield.

#### *2.2. Descriptive Statistics*

Descriptive statistics including the mean, standard deviation, coefficient of variation, minimum and maximum values, skewness, and kurtosis were calculated for each yield trait. Pearson correlation coefficients between the yield trait components within the particular years and between the years were determined using STATISTICA 12 PL (StatSoft 2019).

#### *2.3. Bland–Altman Method*

The Bland–Altman statistics was adopted to determine the separate effect of the different SICS vs. control plots on the cereal yield trait components. In this method, the differences in the cereal yield trait components (grain, straw, and plant height) between the plots with different SICS and the control plots against the average yield with SICS and the control were graphically presented for each study year. The agreement between the yield in the plots with SICS and the control plots was assessed using bias (average of differences between the yield from the plots with SICS and control plots), the limit of agreement (LoA) defined as bias ±1.96 × standard deviation (SD), confidence intervals (CI) for the bias and LoA defined as ± standard error × the value of t distribution with n–1 degrees of freedom, and the Bland–Altman ratio (BAR) defined as the ratio of half the range of LoA to the mean of the pair including the yield from plots with SICS and control plots, the and regression line from the equation y = ax + b, where y—differences between the plots with different SICS and the control plots, x—average yield from plots with SICS and the control, a—regression coefficient, b—intercept. The BAR values were graded as good, moderate, and insufficient for values (BAR < 0), (0.2 ≤ BAR < 0.4), or (BAR ≥ 0.4), respectively [48].

Root mean square residuals (RMSR) and maximum relative residuals (MRR), which are the differences in the yield between the plots with SICS and the control plots, were determined for all yield trait components and each study year.

#### **3. Results**

#### *3.1. Weather Conditions*

Figure 1 illustrates the course of monthly mean air temperatures and rainfall sums for 2017, 2018, 2019, and 2020 in the study site. The average temperatures during the growing season (April–September) and the annual temperatures in the successive years were 14.8, 17.1, 15.9, and 15.1 ◦C and 8.7, 9.3, 10.0, and 9.7 ◦C, respectively. The respective sums of the growing season and annual precipitation were 424.1, 308.1, 306.2, and 439.8 mm and 670.1, 509.1, 475.9, and 666.2 mm, respectively. The growing season precipitation rates in 2017–2020 were below the long–term average (567 mm).

**Figure 1.** Monthly average air temperatures and sums of precipitation during the period of 2017–2020.

#### *3.2. Descriptive Statistics*

Basic statistical parameters of the grain and straw yields and plant height are given in Table 1. The ranges of the mean, minimum, and maximum values of the grain yield (in kg m<sup>−</sup>2) in the growing seasons 2017–2020 were 0.123–0.361, 0.070–0.240, and 0.180–0.530, respectively. The corresponding ranges for the straw yield (in kg m−2) were 0.205–0.432, 0.140–0.230, and 0.310–0.770 and plant height (in cm): 53.5–90.1, 48–75, and 63–110. The minimum values of all yield trait components were noted in 2018 or 2019 and the maximum values were determined in 2017 or 2020. The largest and similar coefficient variations (CVs) were recorded for the grain yield (17.5–28.1%) and the straw yield (19.7–28.5%), whereas lower values were calculated for the plant height (7.5–10.2%).

According to the classification proposed by Dahiya et al. [51], the CV values for the grain and straw yields were moderate (15–75%) and low (0–15%) for the plant height. The asymmetry (skewness) of the grain and straw yields was positive (0.053–0.930), whereas that of the plant height ranged from positive 0.803 in 2018 to negative −0.052 in 2017. The kurtosis of the grain yield varied from 0.141 in 2018 to negative −0.445 in 2020. The corresponding ranges for the straw yield and the plant height were 0.480 in 2017 to −0.249 in 2019 and −0.349 in 2019 to −1.018 in 2020. In general, the skewness and kurtosis values indicate that the yield trait components were close to the normal distribution, which was slightly flattened in nine cases and slightly slender in three cases.

The response of the cereals to the SICS applied was related to the yield trait components and the study year. The differences in the mean grain yield between the particular treatments and the control in the first three study years (2017–2019) varied from 18.0% to −16.6% (Figure 2a). However, in the last study year (2020), the wheat grain yield increased (*p* < 0.05) in the M and M + L + LU treatments by up to 47.3% and 45.8, respectively, compared to that in the control (0.279 kg m−2). A substantially lower and statistically insignificant increase was observed in the lime (L) and catch crop (LU) treatments (by 10–18%). On average, the cereal grain yield during the experimental period (2017–2020) increased in the L, M, and M + L + LU variants by 2.5, 23.3, and 16.6%, respectively, and slightly decreased in LU (by 0.7%) compared to the control (0.224 kg m<sup>−</sup>2). Irrespective of the treatment, the grain yields were considerably lower in both dry years 2018 and 2019

(growing season rainfall: 306 and 308 mm) than in the wet years 2017 and 2020 (growing season rainfall: 424 and 440 mm). The grain yields averaged over the treatments were 0.123–0.143 kg m−<sup>2</sup> in the dry years (2018–2019) and 0.333–0.361 kg m−<sup>2</sup> in the wet years 2017 and 2020 (Table 1). The inter-annual variations in the grain yields were relatively greater than those between the SICS treatments in all study years.


**Table 1.** Basic statistics for cereal yield trait components during the period of 2017–2020.

The straw yield changes in response to the SICS applied varied in the first three years from 22.3% (in M + L + LU in 2019) to −11.4% (in L in 2018) (Figure 2b). The highest straw yield increment was observed in 2020 in the M and M + L + LU variants, where the straw yield increased by 58.2% and 65.0%, respectively, compared to that in the control (0.340 kg m−2). It is worth noting that this increase in the straw yield in both treatments was relatively greater than that of the grain yield and was reflected in lower grain/straw ratios (Figure 2c). Noteworthy, the similar mean grain yield of spring wheat in the dry years 2018–2019 (0.123–0.143 kg m−2) was accompanied by a considerably (almost twice) higher straw yield (0.395 kg m−2) in the dry year 2019 than in the other dry year 2018 (0.205 kg m<sup>−</sup>2).

The high straw yield in 2019 was clearly reflected in the considerably lower grain/straw ratios in all treatments (0.287 to 0.331), compared to those in the other years, i.e., 2017 (0.969 to 1.084), 2018 (0.633 to 0.787), and 2020 (0.733 to 0.874) (Figure 2c). The straw yield averaged over the whole experimental period (2017–2020) increased in L, LU, M, and M + L + LU by 1.5, 4.3, 23.8, and 29.0%, compared to the control 0.311 kg m<sup>−</sup>2, respectively.

The plant height at harvest in the first three years (2017–2019) in the particular SICS was slightly lower (to 5.8%) or higher (to 9.7%) and statistically insignificant compared to the control (Figure 2d). However, in 2020, the plant height was significantly (*p* < 0.05) higher in M (by 16.0%) and in M + L + LU (by 20.7%), compared to the control (86.7 cm). It is worth noting that the plant height response to the particular SICS was relatively lower than that for grain and straw, irrespective of the study year. The plant height averaged over the four study years increased in L, LU, M, and M + L + LU by 3.1, 0.8, 8.4, and 11.6%, respectively, compared to the control (69.4 cm).

**Figure 2.** Mean values of the grain yield (**a**), straw yield (**b**), grain to straw yield ratio (**c**), plant height (**d**), and in response to soil-improving cropping systems. C—control, L—liming, LU—leguminous catch crops, M—farmyard manure andM+L + LU—farmyard manure + liming + leguminous catch crops. The vertical bars indicate the standard deviations (n = 9).

#### *3.3. Correlation Coefficients between Yield Trait Components*

As can be seen from Table 2, the highest correlation coefficient (r) between straw and grain was determined in the wet and last study year 2020 (0.798), whereas the lowest value was reported in the dry 2018 (0.393); both values were statistically significant (*p* < 0.05). The correlation coefficients between the plant height and the grain yield were more closely correlated in the wet and last study year 2020 (0.776) (*p* < 0.05), compared to the first three study years (0.487–0.596). The lowest correlation coefficients between the plant height and the straw yield were calculated in the dry 2018 (0.189), and the highest values were recorded in the wet and last study year 2020 (0.833) (*p* < 0.05). Noteworthy, there were markedly different r values between the plant height and the straw yield in the two dry years, i.e., 2018 (0.189) and 2019 (0.785), with much higher straw yield (and higher plant height) in the latter at a similar grain yield in both years. In line with this finding, there are significant positive correlations for the grain yield and the plant height between 2018 (0.322) and 2019 (0.333) (*p* < 0.05) in contrast to the insignificant and negative correlation for the straw yield (−0.57). Overall, the highest coefficient correlations between all paired yield trait components were recorded in the last study year.

#### *3.4. Bland–Altman Analysis*

Bland–Altman plots including horizontal lines of the bias line (mean difference from the SICS and control plots), limits of agreement (LoA = bias ± 1.96 × SD) along with confidence intervals (CI), regression lines (y = ax + b), and Bland–Altman ratio (BAR, half the range of LoA to the mean differences between the SICS and control plots) describe quantitatively the impact of particular SICS vs. the control on the cereal grain and straw yields and plant height. They are shown in Figures 3–5.

241

**Figure 5.** Bland–Altman plots for plant height (H) (cm) (oats—2017), (spring wheat—2018, 2019), (oats—2020). C—control, L—liming, LU—leguminous catch crops, M—farmyard manure, and MLLU—farmyard manure + liming + leguminous catch crops, LoA—limits of agreement, CI—confidence intervals.


**Table 2.** Correlation coefficients (r) between grain, straw, and height in the study years. The correlation coefficients in bold are significant at *p* < 0.05.

The average differences (biases) indicate that the application of the particular SICS resulted in a lower grain yield (bias < 0) in seven cases and a higher grain yield (bias > 0) in nine cases (Figure 3). The negative biases varied in kg m−<sup>2</sup> from −0.002 for LU in 2019 to −0.017 for LU in 2018, and positive biased ranged from 0.006 for M + L + LU in 2017 and L in 2020 to 0128 for M and 0.132 for M + L + LU in 2020. It is worth noting that all negative biases occurred in the first three study years and the most positive ones were noted in the last study year. The ranges of LoA for the grain yields were in general similar in 2017 and 2018 in all comparable SICS treatments (except higher in LU in 2018). They decreased considerably in 2019 and then increased in 2020. The increase in 2020 was most pronounced in M and M + L + LU, where the ranges of LoA± in kg m−<sup>2</sup> (0.318, −0.062 and 0.241, 0.023) were several times greater than those in 2019 (0.077, −0.033 and 0.069, −0.016). The largest ranges of LoA in M and M + L + LU in 2020 correspond with the respective highest maximum values of root mean square residuals (RSMR) (0.16 and 0.143 kg m−2) and maximum relative residuals (MRR) (164.7 and 158.8%) (Figure 6). Irrespective of the treatment, the largest Bland–Altman ratio (BAR) values were noted in 2018 (0.809–0.929) and the lowest were recorded in 2017 (0.231–0.330), which indicates insufficient (BAR ≥ 0.4) and moderate (0.2 ≤ BAR < 0.4) agreement, respectively, between the grain yield in the SICS and control plots [48].

As to the straw yield, the negative biases occurred in seven of the 16 cases and changed in kg m−<sup>2</sup> from −0.002 for M + L + LU in 2018 to −0.036 for LU in 2017 (Figure 4). The positive biases varied from 0.011 for LU in 2018 to 0.198 for M and 0.221 for M + L + LU in 2020. The highest positive biases correspond with the highest RMSR (0.251–0.269 kg m<sup>−</sup>2) and MRR (196.2–204.3%). The ranges of the limits of agreements (LoA = bias ± 1.96 × SD) were in general relatively narrow and similar in all treatments in the first two study years 2017–2018, but increased largely in all SICS treatments in 2019–2020. This increase was most pronounced in 2020 in the M and M + L + LU treatments, where the LoAs ranges approximately doubled compared to those in 2017–2018. The lowest BAR values were recorded (0.311–0.425) in the first study year 2017, and the largest were noted in 2020 (0.490–0.753) in all comparable treatments (Figure 6), which indicates moderate (0.2 ≤ BAR < 0.4) and insufficient (BAR ≥ 0.4) agreement, respectively, between the straw yield in the SICS and control plots [48].

**Figure 6.** Root mean square residuals (RMSR) and maximum relative residuals (MRR), which are the difference in the yield trait components between the particular SICS and control plots, BAR—Bland–Altman ratio, Bias—mean of the differences, LoA—limit of agreements, SD—standard deviation of the differences.

The bias values for the plant height were mostly positive in 13 of the 16 cases (except two negative values for LU in 2017 and L in 2018 and 0 for LU in 2018). The lowest bias in cm (−4.33) was recorded in 2017 for LU and the highest in 2020 for M (13.33) and M + L + LU (17.22). In comparable treatments, the lowest biases occurred in general in the dry growing season 2018 with the lowest mean plant height among the study years and the highest values in 2020 with the highest plant height (Figure 5). The ranges of LoAs were wider in 2020 than in the other years. The highest BAR values for most of the comparable pairs were recorded in 2020 when the plants were the tallest. In all years except 2017, the highest BAR values were noted for M + L + LU. The BAR values in 8 cases (0.111 to 0.199) and in four cases (0.212–0.262) (Figure 6) indicated good or moderate and moderate agreement [48], respectively, between the grain yield in the SICS and control plots. As can be seen in Figure 6, the BAR values for the plant height were lower than those for the grain and straw yields, irrespective of the treatment and study year. The lower BAR values for the plant height correspond with the higher RMSR and lower MRR values.

The regression lines of the differences between the particular SICS and control plots against the average yield of both indicate that the trends for grain were descending, ascending, or almost unchanged (close to the bias line) depending on the SICS type and study year. Ascending trends were mostly observed for the paired treatment M and Control. As to straw yield and plant height, the regression lines indicate slightly descending or ascending trends.

Regardless of the SICS type, yield trait component, and study year, the Bland–Altman plots indicate that a bulk of the points are within the limits of agreement (LoA) and outliers—within the confidence intervals (CI) (Figures 3–5).

#### **4. Discussion**

#### *4.1. Impact of the Soil-Improving Cropping Systems (SICS) on Yield Trait Components*

Our study showed the most pronounced differences in all crop yield trait components between SICS in the fourth and wet study year. A statistically significant and similar increase in the crop yield was found in two SICS, i.e., M consisting of only farmyard manure and M + L + LU providing less farmyard manure and plus lime and cover crops. This significant impact of both treatments may result from the increased nutrient supply from farmyard manure and cover crops although the soil organic matter content increased only slightly (data not shown). Similarly, the high yields in the M and M + L + LU variants imply that organic matter from deficit farmyard manure in the former can be replaced in part by green manure/cover crops, with maintenance of the same productivity. The positive effect of the combined SICS (M + L + LU) on the crop yield in the acid soil can also be enhanced by yield-increasing liming improving the availability of essential nutrients to plants [52]. These results support the recent actions in several countries, including Poland, focused on promotion of incorporating legumes in the intercropping systems and extending agricultural lime application [53–55]. Application of the combined SICS should be considered not only in relation to crop productivity enhancement but also as a sustainable strategy to improve the supply capacity of essential nutrients including fixed atmospheric nitrogen [56], and alleviating the negative effect of soil acidity [52]. It should also be noted that increasing the organic carbon content or even keeping good levels in sandy soils requires a continuous supplying organic materials. This is due to the fact that sandy soils, especially tilled, are well aerated creating conditions conducive to rapid microbial decomposition of organic matter.

#### *4.2. Weather Influences*

Our results showed that the cereal yield trait components were largely influenced by both the total rainfall amount and their temporal distribution during the growing seasons. As could be expected, the wheat grain yield was appreciably lower (by more than 50%) in the two dry growing seasons compared to the two wet growing seasons. The analysis of the yield trait components and the weather course further revealed that, in both dry years (2018–2019) with almost the same total amount of rainfalls during the growing season (306– 308 mm), the straw yield of spring wheat was by 160–221% higher, depending on treatment, in 2019 than 2018. In turn, the grain yield of spring wheat in 2019, compared to 2018, was not different or slightly lower (by 6.4–24.3%) in the comparable treatments. This opposite response of the yield trait components can be explained by the different distribution of rainfalls during the analyzed growing seasons. The large amounts of rainfalls in May during intensive growth at shooting and the scarce precipitation during later growth in 2019 (Figure 1) may have stimulated top growth. Moreover, the shoot growth in May 2019 may have been favored by the lower temperature (12.5 ◦C) compared to that in 2018 (18.5 ◦C) (Figure 1) by changing evaporation rates. The more intensive shoot growth in 2019 vs. 2018 was reflected in the greater straw yield in the former season in all treatments (Figure 4). The results imply that a good water supply at shooting increases allocation of assimilates to shoots while reducing the grain yield. These diverse responses of the yield trait components emphasize the importance of the increasingly frequent episodic (extreme) drought and wet conditions during the growing season associated with climate change [57]. The sensitivity of the yield trait components to weather variation during the growing season in sandy soils can be enhanced by the high permeability and low water holding capacity of these soils, which do not allow storing water for a longer time and

efficient use of nutrients [5], and by the relatively shallow root system of spring cereals. Understanding the relations among the yield trait components depending on the weather course during the growing season is important in food and bioethanol production where grain and straw, respectively, are potential feedstocks [58,59].

#### *4.3. Usefulness of the Bland–Altman Method*

The use of the Bland–Altman method contributed to improvement of the quantification of the direct (separated) impact of a given soil-improving practice on the cereal yield trait components in reference to the control in different inter-annual weather conditions. For example, the small values of limits of agreement (LoA = bias ± 1.96 × SD) for the grain yield in the most yield-producing SICS (M and M + L + LU) in the dry growing season 2019 (with the lowest yield) increased by several times in the wet season 2020 (with the highest yield), indicating that the grain yields were less even in the latter. The reduced evenness of the yield in the wetter growing season 2020 may have resulted in part from the variability in soil water content and the related availability of nutrients from organic matter provided by these two SICS. The variable soil water content in these SICS treatments may have resulted from changes in the soil structure caused by the organic matter amendments and from the natural variability of the soil texture in the study area [19]. This explanation is supported by the fact that water deficit is a dominant crop yield–limiting factor in sandy soils [5,60].

The Bland–Altman plots indicate that the orientations of the regression equation lines for the grain and straw yield in M and M + L + LU, compared to the other SICS, were in most cases close together to the bias lines. This can be indicative of the stabilizing effect of the largest quantity of organic matter provided by both SICS on the yield and uniformity of the yield components. The regression equation lines below or above the bias line indicate a reduction in yield uniformity. It is important to add that if only one treatment, i.e., SICS or the control, in the pair has a wide range of limits of agreement (LoA), the Bland–Altman will always produce wide limits of agreement [61]. This means that poor agreement between the paired SICS and control do not necessarily indicate that the tested SICS has low evenness of crop yields in the replicate sub-plots.

The comparison of the Bland–Altman Ratio of the yield trait components revealed that grain and straw yields, compared to plant height, exhibit appreciably higher uncertainty (at the most comparable paired SICS and control). Even with high uncertainty, analysis of biases and LoA values facilitates assessment of the degree of the causal (positive or negative) effect of particular SICS on the crop yield. This observation along with interannual differences in crop yield trait components is important in modelling crop responses to SICS and weather conditions during growing seasons [62].

#### **5. Conclusions**

The results of this study indicate the following findings:


trait components, the highest limits of agreement (LoA) were recorded in 2020 in the M and M + L + LU variants, where all yield trait components reached the maximum values.


**Author Contributions:** B.U. and J.L. performed experiments and measurements. B.U. carried out data computation and statistical analysis. All authors have read and agreed to the published version of the manuscript.

**Funding:** The research leading to these results has received funding from the H2020 under grant agreement number 677407 (SOILCARE).

**Institutional Review Board Statement:** Not applicable.

**Data Availability Statement:** Data sharing is not applicable to this article.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **The Impact of Soil-Improving Cropping Practices on Erosion Rates: A Stakeholder-Oriented Field Experiment Assessment**

**Ioannis K. Tsanis 1,\*, Konstantinos D. Seiradakis 1, Sofia Sarchani 1, Ioanna S. Panagea 2, Dimitrios D. Alexakis <sup>3</sup> and Aristeidis G. Koutroulis <sup>1</sup>**


**Abstract:** The risk of erosion is particularly high in Mediterranean areas, especially in areas that are subject to a not so effective agricultural management–or with some omissions–, land abandonment or wildfires. Soils on Crete are under imminent threat of desertification, characterized by loss of vegetation, water erosion, and subsequently, loss of soil. Several large-scale studies have estimated average soil erosion on the island between 6 and 8 Mg/ha/year, but more localized investigations assess soil losses one order of magnitude higher. An experiment initiated in 2017, under the framework of the SoilCare H2020 EU project, aimed to evaluate the effect of different management practices on the soil erosion. The experiment was set up in control versus treatment experimental design including different sets of treatments, targeting the most important cultivations on Crete (olive orchards, vineyards, fruit orchards). The minimum-to-no tillage practice was adopted as an erosion mitigation practice for the olive orchard study site, while for the vineyard site, the cover crop practice was used. For the fruit orchard field, the crop-type change procedure (orange to avocado) was used. The experiment demonstrated that soil-improving cropping techniques have an important impact on soil erosion, and as a result, on soil water conservation that is of primary importance, especially for the Mediterranean dry regions. The demonstration of the findings is of practical use to most stakeholders, especially those that live and work with the local land.

**Keywords:** soil erosion; soil-improving crop systems; sustainable land management; sustainable agriculture

#### **1. Introduction**

Soil erosion is a primary biophysical process involving the detachment of soil particles from a given initial area and their transport and accumulation to a new depositional area [1]. It is considered one of the most severe natural threats worldwide, as it threatens soil fertility, water availability and crop productivity [2]. The risk of erosion is particularly high in Mediterranean areas, especially in areas that are subject to a not so effective agricultural management–or with some omissions–, land abandonment or wildfires [3].

Crete's Mediterranean soils are under imminent threat of desertification, characterized by loss of vegetation, water erosion, and subsequently, loss of soil. In particular, the serious impact of the expected climate change to the southern Mediterranean regions, together with the adoption of crop techniques by many olive groves' farmers that negatively affect the environment, such as intensive tillage, use of chemicals, burning of pruning branches in their fields, may lead to loss of ground's organic matter, putting the fields in possible drought hazard in the upcoming years [4,5]. Several studies have focused on the estimation of average soil erosion in the island. Most of the studies simulate

**Citation:** Tsanis, I.K.; Seiradakis, K.D.; Sarchani, S.; Panagea, I.S.; Alexakis, D.D.; Koutroulis, A.G. The Impact of Soil-Improving Cropping Practices on Erosion Rates: A Stakeholder-Oriented Field Experiment Assessment. *Land* **2021**, *10*, 964. https://doi.org/10.3390/ land10090964

Academic Editors: Guido Wyseure, Julián Cuevas González and Jean Poesen

Received: 3 August 2021 Accepted: 8 September 2021 Published: 12 September 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

erosion with the use of the revised universal soil loss equation (RUSLE) method. Kazamias et al. [6], for example, estimated average soil erosion rates for the Greek territory at 4.75 Mg ha−<sup>1</sup> yr−<sup>1</sup> and indicated that for over 12% of the Greek area higher ratings than 10 Mg ha−<sup>1</sup> yr−<sup>1</sup> are observed, mainly located at steep areas. Similar values are estimated for the Cyprus area (11.75 Mg ha−<sup>1</sup> yr<sup>−</sup>1) with October and November being the most erosive months [7]. Panagos et al. [8] estimated soil erosion rates between 6 and 8 Mg ha−<sup>1</sup> yr−<sup>1</sup> for the region of Crete and Kourgialas et al. [9] suggested ratings of an average 4.85 Mg ha−<sup>1</sup> yr−<sup>1</sup> for the western part of the island. Several other investigations assess soil losses as being orders of magnitude higher. Kouli et al. [10] provided an estimate of soil loss of up to 200 Mg ha−<sup>1</sup> yr<sup>−</sup>1, and Alexakis et al. [11] suggested that losses of more than 200 Mg ha−<sup>1</sup> yr−<sup>1</sup> were recorded for 2018. Polykretis et al. [12] assessed the intraannual and inter-annual fluctuations, providing similar estimates. Furthermore, changes in rainfall patterns are foreseen to affect soil erosion in the area. Grillakis et al. [13] examined projected changes in erosivity for the island of Crete under three concentration pathways (RCP2.6, RCP4.5 and RCP8.5). Simulations suggest positive changes exceeding 30% for the 2021–2050 period, while for the far future, erosivity decreased with the increase in concentration, ranging from −10% to +30% on average, depending on the scenario and as a result of changes in extremes [14].

Despite the extensive literature devoted in the investigation of soil erosion at the regional scale, few studies focus on the local field scale, and in particular, at the major land use types and associated land management practiced. Olives are the most important crop grown on the island of Crete [15], covering 64% of the arable land and representing 86% of the tree plantations on the island. Despite the problem of phyloxera in the 1980s [16] and the Common Agricultural Policy (CAP) to reduce the area of vineyards, viticulture remains one of the most important production activities of Crete. Olive orchards and vineyards in Crete often suffer from extreme soil erosion by water due to farm slope and recent intensification of tillage practices [17–19]. There is a need to find practices that prevent soil erosion without reducing the profitability of both crops. Less tillage at the olive sites can improve soil health by reducing organic matter decline, keeping soil microbiology intact, and limiting compaction through less machine passes across fields, as well as reducing fuel use and related emissions [20]. In addition, the simplest and most natural way to prevent erosion in vineyards is through planting vegetation. Cover crops keep ground covered over storm events with high rain rates and winds, which can cause erosion [21,22]. Plants establish root systems which stabilize the soil and prevent erosion. Moreover, cover crops can reduce the need for fertilizer and supply organic nitrogen if leguminous [23,24].

Average erosion rates for orange groves on the island are estimated at 1 Mg ha−<sup>1</sup> yr<sup>−</sup>1, whereasthe average rates are assessed at 8 Mg ha−<sup>1</sup> yr<sup>−</sup>1, which is still above other cultivations. Moreover, in the Chania Prefecture of Crete, orange cultivation is a major crop, but due to severe market competition, producer prices have significantly dropped, leaving little or no profit. Recently, avocado plantations have been proposed as a potentially sustainable alternative over orange groves, with high profitability and increasing demand, but soil erosion for avocados has not been measured [25]. The cultivation is demonstrating to fit in warm seasons of Mediterranean regions [26].

The objectives of the present experimental study are: (i) to compare different tillage practices in olive orchards, as tillage is known to affect soil erosion rates; (ii) to test the use of a vetch cover crop in a vineyard compared to no vetch, with vetch being a nitrogen fixing cover crop; (iii) to compare the erosion rates as well as other soil quality parameters between a field that has remained an orange grove for 45 years and one that was converted to an avocado farm 20 years ago.

#### **2. Materials and Methods**

#### *2.1. Study Site*

The experiment was conducted on three real farm fields managed by farmers in three different areas of Chania, Crete, Greece. The first field is an olive orchard located at Biolea

in Astrikas region, at an altitude of about 260 m and covers an area of about 3000 m<sup>2</sup> with a slope gradient of about 6%. The olive trees were planted in a dense of 90 trees per 1000 m2. The field is 25 years old and had not been tilled for 7 years before the beginning of the project. The second field is a vineyard located in Alikampos region, in an area of about 3000 m2 and an altitude of about 254 m. The slope gradient of the field is about 15%. The investment began in 2013. Since then, animal manure is the fertilizer applied every two years in the orchard, while moldboard ploughing at 20 cm depth is a standard farm operation. The field is also drip irrigated during the summer period. The third field is a fruit orchard (orange and avocado) located in Koufos region, in an altitude of about 86 m and covers an area of about 2000 m<sup>2</sup> with a slope gradient of around 10–15%. The orange trees were planted in 1988 and the conversion of part of the orange orchard to avocado orchard occurred in 1998. The Avocado trees' first plantation included 40 trees per 1000 m2, whereas the orange trees' first plantation included 120 trees per 1000 m2. The fruit orchards received ammonium sulphate and potassium fertilizers during the past 10 years, applied in the irrigation water during the summer period, whereas solid potassium nitrate is banded on the soil surface in the winter period. Every year, soil mulching with cut branches occurs in the form of wood chips. Manure is also applied every year on the avocado trees. Moldboard ploughing at 20 cm depth occurs every two years and glyphosate is banded on the soil surface every year for weed management. Finally, the field is drip irrigated according to the needs of each summer period. The topsoil of all sites has a clay loam texture according to the USDA classification system.

Each experimental site has a representative meteorological station. The closest meteorological station of the olive orchard field is Kolympari, whereas of the vineyards is Vrysses and of the fruit orchards is Alikianos. However, the time period of temperature, precipitation and evapotranspiration observations is not long. Vrysses started gauging in 2007, Alikianos in 2012, and Kolympari in 2016. Table 1 displays the yearly hydrometeorological records of the 2018–2020 period for the stations of interest.


**Table 1.** Overview of the yearly temperature, precipitation and ET0 for the experiments.

Crete has a typical Mediterranean island environment with about 53% of the annual precipitation occurring in the winter, 23% during autumn and 20% during spring, while there is negligible rainfall during summer [27,28]. The average precipitation for a normal year on the island of Crete is approximately 934 mm, with a markedly non-uniform distribution, a reduction of almost 300 mm from the west to the east part of the island and a strong orographic effect. Noticeable are the high rainfall winters and the dry summers in the Chania Prefecture [14].

Regarding the hydrometeorological conditions during the years of the experiment, on 26 October 2017, as well as on 15 and 24 February 2019, Western Crete suffered excessive rainfall and flooding. The October 2017 event was a high-intensity and short-duration rain event, resulting in flash floods in the low-elevation agricultural and urban areas on the northern part of the Chania Prefecture. Persistent storm events in February 2019 resulted in flooding, extensive riverbank erosion, landslides and rocks throughout the road network of Chania Prefecture, as well as in the collapse of the 111-year-old historical Keritis bridge over Alikianos River. For the entire Chania region, 2018 was a dry year followed by an exceptionally wet 2019, mainly due to the record high precipitation accumulations of February (1202 mm/month for Askifou station, Chania), and a normal 2020.

As for relative mean climate conditions between the study sites during the course of the experiment, safe results cannot be extracted due to the distance of the meteorological stations from the sites, differences in altitude and microclimate. In general, the vineyard site located in Alikampos receives the highest amount of mean annual rainfall (~1400 mm) and has the lowest mean temperature (due to lower minimum temperatures at place). The fruit orchard (orange and avocado) is located in probably the most fertile and intensively cultivated valley of Chania prefecture with an average precipitation of about 1200 mm/year, while the olive orchard site located in Astrikas receives less precipitation and has higher mean annual temperature, despite the higher altitude.

#### *2.2. Experimental Setup*

The beginning of the experiment was in 2017, involving a control versus treatment (soil-improving cropping system, SICS) experimental design. At the olive orchard field, two treatments about soil cultivation were tested. A normally tilled area, which served as the control plot and tilled twice within SoilCare project (November 2017 and May 2019), was compared with the no-tilled one, which was the SICS plot. The tillage method was moldboard ploughing at 20 cm depth. Two olive varieties were located in the experimental organic farm of 0.29 ha, *Olea europaea* and *Koroneiki*. At the vineyard site, the experiment compared a vetch (*Vicia faba*) cover crop plot with one without a vetch cover crop. The plot with no vetch served as the control area and the other plot, which was tilled and seeded with vetch, served as the SICS area. The grape variety was *Vitis vinifera* and the plots were located on a corporate organic farm of 0.46 ha. All the farms were fully operated and managed by the farmers, and because of practical management issues, no replicate plots could be designed. At the fruit orchard field, the experiment compared an orange orchard area, served as the control plot, with a rotation crop area of avocado trees, served as the treatment (SICS). The orange orchard variety was *Citrus* × *sinensis*, whereas the crop switch variety was *Persea Americana*, and the plots were located on a family conventional farm of 0.5 ha.

Soil loss rate assessments of both the olive orchards and vineyards' fields were undertaken through cross sections' measurements. The total soil loss is estimated by the erosion/deposition (ER) equation:

$$ER = \frac{VOL \times BD}{TA} \tag{1}$$

where *VOL* is the volume (m3), *BD* is the bulk density (kg/m3), and *TA* is the total effective area (m2).

In the olive orchard site, the soil loss rate monitoring occurred from November 2018 to June 2021 (32 months) with two cross sections, one per plot, having lengths of 5 and 5.8 m for the SICS and control plot, respectively. In the vineyard site, soil loss rate was monitored from January 2019 to June 2021 (2.5 years) through six cross sections, three per plot, of lengths ranging from 1.64 to 2.2 m. At the fruit orchards, soil loss rate assessments were undertaken through soil pins' measurements. The soil loss rate monitoring occurred from May 2018 to June 2021 (37 months), with three to four soil pins per plot, placed per 0.5 m. Figure 1 displays the three farm fields in which experiments were conducted.

**Figure 1.** (**a**) Tilled plot (up) and non-tilled plot (down) at the olive orchard site, (**b**) positions of the cross sections (CS) in which soil erosion measurements were performed at the vineyard site, and (**c**) oranges and avocado trees at the fruit orchard field.

Biophysical measurements were also performed both in the control and SICS plots regarding soil texture [29], saturated hydraulic conductivity [30,31], water stable aggregates [32], bulk density [33], mineral nitrogen [34,35], available phosphorous [36], exchangeable potassium, sodium and magnesium [37], soil organic carbon [38], soil pH [39], soil electrical conductivity [40] and earthworm count [41] in the three experimental fields at the end of October for the years 2019 and 2020.

Soil texture was measured with the Bouyoucos hydrometer method [42]. Saturated hydraulic conductivity was measured with the Beerkan method [43] but not in fully dry conditions as shown in Figure 2a. Soil aggregate stability was counted by sampling about 100 g of three to four soil aggregates from the topsoil per plot, which was air-dried for 20 days and thereafter immersed in water on a mesh of 0.4 cm diameter. The aggregates were observed for a few minutes for slaking [44]. The steps followed to perform the measurement are indicated in Figure 3. Bulk density was measured in a laboratory as an indicator of soil compaction. For its assessment, three soil samples from topsoil (10–20 cm) and three soil samples from subsoil (40–50 cm) per experimental plot (control and SICS) were taken with a metal ring with known volume of 246.42 cm3. Figure 2b displays the collection of a soil sample for the bulk density measurement. The following procedure concerned, first of all, the weighting of an ovenproof container in which each one soil sample per time was placed on; the soil was dried for one night in a conventional oven at 105 ◦C and then weighted. The difference between the two weight measurements divided by the soil volume gave the calculation of the bulk density [44]. A mixed soil sample was collected from each experimental plot, with soil from under ten trees using a Z-shape sampling methodology in order to estimate mineral nitrogen, available phosphorous, exchangeable potassium, sodium and magnesium, soil organic carbon (SOC), soil pH, and soil electrical conductivity. All samples collected were air-dried, grounded to pass a 2-mm sieve, and analyzed for selected chemical properties. Concerning the available forms of the nutrients, NO3 - -N was extracted with 1M KCl and determined with spectrophotometry at wavelengths 210 and 270 nm. Olsen P was extracted by 0.5M NaHCO3 with pH 8.5 and was quantitatively determined with molybdenum blue-ascorbic acid method [36] by using Vis-UV spectrophotometry. Exchangeable cations K+, Na+, and Mg2+ were extracted with 1M ammonium acetate, having pH 7 [45], and were analyzed by the inductively coupled plasma method ICP-OES. pH was measured in a soil/water suspension at a 1:2 ratio; SOC was determined with the wet oxidation method [38], whereas the electrical conductivity was measured in the saturation paste extract [46]. Earthworm density was evaluated as an indicator of the biological health and condition of the soil per experimental plot. The procedure followed was the mixing of 2 L of water with 20 g of mustard seed, the pouring of the half mixed on a 25 cm × 25 cm sample plot where vegetation and leaves were removed, and the observation of worms that came to surface over a period of 5 min, and then the pouring of the remaining mix and the waiting of another 5 min to gather

worms that came to the surface. Figure 2c,d indicate the procedure followed for earthworm counting in the olive and vineyard field respectively.

**Figure 2.** (**a**) Infiltration rate experiment, (**b**) soil sampling for bulk density measurement, (**c**) pouring mix of water and mustard for earthworm test in Astrikas, and (**d**) worms coming out to surface after the pouring of the mix in Alikambos.

**Figure 3.** (**a**) Soil aggregates from both plots (control and SICS) of the three study sites: air-drying before slaking test, (**b**) 100 g of three aggregates of the Astrikas field to be used for the slaking test, (**c**) soil aggregates into the mesh before being immersed to water, and (**d**) soil aggregates after the immerse to water.

#### *2.3. Stakeholder Engagement*

Throughout the SoilCare project, two stakeholders' workshops and two stakeholders' meetings were held, either with physical presence or virtually. The first workshop occurred on 21 March 2017 at the Technical Chamber of Crete (West Crete Chamber), where 12 persons (4 female and 8 male) participated. The main participants were farmers, agronomists and researchers. The stakeholders introduced themselves and justified their interest in the SoilCare project objectives. They were asked to place themselves on a stakeholder matrix that determined the scale of motivation and perceived influence, graded from low to high. The stakeholders also participated and contributed their experience and knowledge on drivers, barriers and solutions for the soil erosion threat in their area. During the workshop, commonly accepted and applied practices to combat soil erosion, along with their benefits and drawbacks, were discussed for further evaluation and potential outspread to all farmers. The stakeholders were also asked to rank, in a scale from 1 to 6, suggested ways to receive information about SoilCare during the lifetime of the project, as well as the display ways they would receive information regarding new SoilCare practices. The questionnaire they had to fill also concerned their preference on dissemination manners

of information and advice of farming practices being used in SoilCare, their three main questions that they would like to be answered when evaluating whether to apply a new practice, as well as the way they would normally find out about new farming practices, beyond the project.

A meeting was held in March 2018 on the premises of Technical University of Crete (TUC), which was attended by 6 people (2 women and 4 men). All the stakeholders were updated about the progress of the field experiments related to soil loss monitoring in the three agricultural fields and the installation of six sediment fences/traps for collecting deposited soil at all the study sites. The stakeholders were asked to evaluate the results of monitoring and soil loss collection thus far. Through a constructive discussion between stakeholders and researchers, it was agreed that the research should focus more on monitoring of soil erosion/deposition implementing additional approaches. In particular, among the following actions of the researchers, there would be the installation of triangular or square grid for monitoring sheet erosion, monitoring of soil roughness with means of images (stereopairs), multi-temporal monitoring and recording of rills in the study areas, as well as correlation of soil organic matter and spectral data (field spectroradiometric measurements) in terms of soil erodibility at a farm and watershed scales.

A research activity was afterward held within the period from April to June 2019 in Chania, which concerned the individual interviewing of 4 stakeholders/farmers (4 men) since it was not possible for them to be gathered in a group. Two of them were involved in the olive orchard experiment, one in the vineyard experiment, and one in the avocado/orange experiment. Toward the specific research activity, the stakeholders had to describe the expected benefits and impacts of the SICS being tested in their field. They also had to identify and describe the key barriers and enablers to SICS adoption. The factors evaluated were economic (farm and market) conditions, biophysical conditions (climate and geomorphology), technical barriers, knowledge and information barrier, and sociocultural factors, as well as institutional or policy regulations. Moreover, the stakeholders were asked to identify and assess feasibility of actions to promote SICS adoption at national and/or (sub)regional level by ranking the enabler and barrier factors from not so important to very important. In addition, they were requested to point out actions that would remove barriers and support enablers as well.

The final stakeholder workshop occurred via online meetings of small groups during various dates in February and March 2021 due to the COVID-19 pandemic restrictions in place. A total of 18 persons (6 female and 12 male) participated and discussed project findings. The main groups were farmers, researchers and agronomists. After the end of the online presentations, the participants, both men and women, and especially the farmers, raised useful questions. Indicatively, the olive groves' farmers wanted clarifications regarding tillage avoidance, especially in the dry season. The vineyard farmers showed particular interest in the application of the experiments. The orange cultivators were interested in understanding the way of further improving the biological health and condition of soil on avocado trees. An important question raised was whether an avocado market will exist for the trees currently planted which will be placed into production in five years. Afterward, the attendees validated whether the project findings were plausible and/or consistent with their understanding. They also identified the benefits that they gained from SoilCare already, as well as the ones that they found important for the future from the project findings. The stakeholders were requested as well to have an active role and state the way they can disseminate the project findings to more people who can benefit from them. The engagement of the stakeholders continued as they were asked to report the way that they would like to be supported in using or implementing project/research findings. Toward the end of the workshop, the attendees mentioned what impressed them most and how to implement what they learned from the workshop.

#### **3. Results**

*3.1. Impacts on Soil Erosion*

3.1.1. Olive Orchard Site

Concerning the sediment fences, although part of the study area was tilled, minimum difference was evident in the collected deposited soil between tilled and no-tilled area. This was considered as a common phenomenon as only few-deposited soil may be collected during winter when it rains more often, due to the presence of naturally grown winter cover crops retaining rainwater. Nevertheless, intensified tillage, which occurred twice in 18 months, contributed to increased soil erosion, as visually observed by the exposed rooting system. Apart from tillage, irrigation also increased soil erosion with the irrigated trees showing shallow roots accompanied with topsoil erosion. Regarding the cross-sectional soil loss measurements, results showed that the no-till treatment had a considerable impact on soil erosion rates. Soil loss rate monitoring revealed that the application of no-till treatment reduced mean soil erosion by over 14%, roughly from 3.3 to 2.9 Mg/ha during the 2.5 years experiment (November 2018 to June 2021).

#### 3.1.2. Vineyard Site

Extreme storm events occurred on 15 February 2019 and 24 February 2019. The nearby rain station recorded the exceptional accumulation of 726.2 mm during this period. These events created rills in the examined field. In the vetch plot, the rills were shorter compared to the no vetch plot when compared visually. The application of the vetch treatment had a direct impact on soil erosion over the 2.5-year monitoring period (January 2019 to June 2021). Soil loss rate monitoring revealed that the vetch coverage reduced mean soil erosion by over 12% (roughly from 3.4 Mg/ha in the no vetch plot to 3 Mg/ha in the vetch plot) during the 2.5 years experiment.

#### 3.1.3. Fruit Orchard Field

An extreme rainfall event occurred on 26 October 2017, leading to more than 2 kg of soil trapped in the sediment fences of 3 m<sup>2</sup> area, corresponding to about 7 Mg/ha. In the rest of the monitored period, the sediment traps did not collect considerable amounts of soil after events of light rain. Further extreme precipitation events, which caused severe flooding in the wider area, occurred in February 2019, triggering further erosion in the field. Field measurements showed that the crop switch to avocado trees significantly reduced the mean soil erosion compared to the orange orchards (control) over 3 years of monitoring (May 2018 to June 2021). Soil loss rate monitoring revealed that the avocado conversion caused over 34% reduce in mean soil erosion, roughly from 4.6 to 3 Mg/ha, during the 3-year experiment.

A one-way ANOVA was performed to compare the effect of the different treatments on the erosion rates, with 90% CI. The analysis revealed that there was a statistically significant difference in the erosion rates between the avocado and orange treatments (*p* = 0.096) (Figure 4iii) but not for the other treatments (*p* = 0.745 for the olive orchard, *p* = 0.561 for the vineyard). Figure 4i,ii indicate the mean soil erosion at the control and SICS plot of the olive orchard and vineyard site respectively after 2.5-year of monitoring through cross sections.

#### *3.2. Impacts on Soil Properties*

#### 3.2.1. Olive Orchard Site

Topsoil bulk density was slightly higher in the no-till plot. Bottom soil bulk density was found at the same levels in both plots. Exchangeable magnesium had an increasing trend in both plots from 2018 to 2020. Mineral nitrogen and available phosphorus concentrations were lower in the no-till plots, both in 2019 and 2020. The soil organic carbon rate had an increasing trend in both plots from 2018 to 2020, and was slightly higher at the last year, which was probably due to the animal manure application. The crop yield was the same at both plots (till and no-till) and was increased in 2020 compared to the years

2018 and 2019. Earthworms' density per m2, which can be used as a sensitive indicator to management changes were substantially higher in the non-tilled plot compared to the tilled one in the 2020 measurement, indicating better soil health and condition. Weed infestation was slightly higher (10%) in the non-tilled plot compared to the tilled one, which cannot be assumed as a considerable high hazard.

**Figure 4.** Mean soil erosion (Mg/ha) in (**i**) tilled and non-tilled plot at the olive orchard site during the 2.5-year of monitoring, (**ii**) no vetch and vetch plot at the vineyard site during the 2.5-year of monitoring, and (**iii**) orange and avocado plot at the fruit orchard field during the 3-year monitoring (\* denotes significant differences at a 90% CI level).

#### 3.2.2. Vineyard Site

By the end of 2020, top- and bottom soil bulk densities of the vetch plot were lower compared to the no vetch plot, indicating good soil functioning, improved water and solute movement as well as soil aeration. A soil aggregate stability test resulted in good soil stability and resistance to erosion for both plots; however, for the vetch applied plot, slaking effect was slightly less observed, indicating better structure maintenance. The soil organic carbon did not follow a specific trend; it was relatively satisfactory around 4% in both plots (control and SICS) from 2018 to 2020. The crop yield was the same at both plots (control and SICS), having a slightly decreasing trend during the 3-year monitoring. Earthworms per m2, which is a soil health indicator, were considerably higher in the vineyard with the cover crop applied. The percentage of weed infestation was 20% less in the vineyard with the cover crop.

#### 3.2.3. Fruit Orchard Field

The soil organic carbon rate was higher in the avocado trees compared to the orange orchards. The saturated hydraulic conductivity was considerably higher in the avocado trees plot compared to the orange trees plot, in the 2020 measurement. The exchangeable magnesium was also higher in the avocado trees compared to the orange orchards during the 3-year monitoring. The level of weed infestation was 10% less in the avocado field compared to the orange trees field. Electric conductivity values indicated high salinity levels in both plots, while higher values were observed for avocado trees.

Due to the lack of replicates, there was no efficient way to place error bars on the graphical values of soil properties. Nevertheless, the repetition of measurements every year demonstrated an important variation in properties values, as shown in Figure 5 regarding the 2.5-year of monitoring of the olive orchard site, Figure 6 concerning the 2.5-year of monitoring of the vineyard site, and Figure 7, which concerned the soil properties after 3-year monitoring of the fruit orchard field.

**Figure 5.** Soil properties in tilled and non-tilled plot at the olive orchard site during the 2.5-year monitoring.

**Figure 6.** Soil properties in no vetch and vetch plot at the vineyard site during the 2.5-year monitoring.

**Figure 7.** Soil properties in orange and avocado plot at the fruit orchard field during the 3-year monitoring.

*3.3. Assessment of Stakeholder Engagement*

3.3.1. First Stakeholder Workshop (21 March 2017)

Local stakeholders underlined that soil erosion mainly depends on geomorphology (slope), soil type, vegetation cover, climate, socio-economic and policy drivers, including human activities (land management, soil conservation techniques). They punctuated that soil erosion causes soil infertility, resulting in reduced production and thus income. Land users and agronomists showed the effects and the indicators of increased soil erosion on their cultivations, such as exposed tree roots, rills, reduced soil organic matter, soil pillars. They also seemed aware of the functions and services offered by the soil, as well as the impact of losing them. A few of the stakeholders perceived the soil functions and services as were presented in the meeting, but with high uncertainty on whether the knowledge will be consolidated. However, the meeting presentation was essential to discuss limits and effects of soil erosion before beginning a meaningful conversation at the local scale. The discussion with the stakeholders highlighted knowledge gaps of the interested parts regarding the extent that the erosion affects the performance and quality of production, the extent that different cultivation practices within the same crop or a total crop change affect the rate of soil erosion, and the most promising erosion mitigation approaches and technologies as well. Major barriers in adopting land management practices include the local administration lack of partaking in decision making, yet can implement and control what is already decided; agronomists perceive lack of interest of farmers (typically those with lower education level), including in regular testing of soil quality of their fields, being reactive rather than proactive; farmers' cooperatives often realize that competitive farmers are noncommitted to the cooperatives' objectives, whereas outsiders perceive cooperatives as less sustainable and profit oriented, or simply disagree with how resources are allocated; there is lack of financial motives (or motive awareness) for stakeholders; educational institutions lack the legislative freedom and means to interact with stakeholders for pilot/prototype practices for soil and land management. The stakeholders considered that

government, local municipalities and individual farmers were responsible for providing solutions.

#### 3.3.2. Research Activity (April to June 2019)

Based on the evaluation of the questionnaires, the olive farmers stated that the money saved from the no tillage practice is counterbalanced by the weed spraying costs or the time of cutting them off. The climate of Crete acts as an enabler for olives, whereas steep slopes, stones and rocks may significantly affect the crop. Drought may further raise the rooting system of the crop higher, affecting mostly the tilled plots. The vineyard farmer expected the yield to be increased through the cover crop. However, because of no directly visible benefits, the farmer, although willing to adopt the practice, stated the need for long-term experiments such that the benefits are proven and quantified. Yet, the presence of different crop within the vineyard competes with the soil structure in dry years. One farmer also noted that there is sufficient access to knowledge and information, but there is difficulty in convincing to test new alternative cultivation methods; it is preferable to recognize the benefits from other farmers and then adopt the new techniques. The avocado yield is expected to be profitable after the fifth year of the crop change, since the investment costs of the crop change process are high. However, there is a high interest for this investment since the climate conditions on Crete seem identical for the avocado crop. In addition, the avocado demand from European markets is expected to increase. The farmer asked for additional knowledge and advice from an experienced agronomist and identically visiting of demonstration sites.

The farmers claimed the need for guidance and advisory services of great expertise on soil data, soil analysis, fertilizers use, as well as extra skills to the specific SICS practices to remove barriers. They also indicated the necessity for financial support and incentives to adopt SICS practices. In addition, they suggested that the organization of workshops could support enablers with successful studies and practical applications that would emphasize the pros and cons of the proposed SICS. All farmers agreed that other farmers with experience in the specific SICS practices can provide important information to them. Social networks and videos can also help farmers adopt new techniques. New cultivation practices according to EU regulations can also be promoted from the State through seminars and programs.

#### 3.3.3. Final Stakeholder Workshop (February and March 2021)

After the end of the online presentations, the soil specialists/consultants and the researchers had well understood the project results in the three fields of its application. The farmers generally found the presentations helpful in understanding the conceptualization of the problems faced in the three field studies, as well as the results obtained. They found the vetch cover crop easy to be applied, and the no-tillage practice feasible. Several of the orange cultivators realized that switching crops to avocados would bring them a great financial profit in long-term, while at the same time soil erosion would be reduced in their fields.

Regarding the benefits gained from SoilCare thus far, all the farmers stated that they gained better knowledge of their fields and the soil properties and functions as well as of the soil erosion's negative impacts and the way these can be avoided. The consultants noted the necessity to inform farmers of soil improving techniques, as well as the requirement for proper training in applying these techniques correctly. Concerning the benefits of the project findings for the future, farmers were willing to apply the proposed SICS practices to new sites or to the rest of the parts of their fields already tested. Several were interested in examining the tests which concerned the soil properties. The consultants were motivated to use the results and present them in workshops and other organized events aimed at farmers. Certain researchers were interested in monitoring the study fields for another 2–3 years, with the agreement of the farm owners, to examine if soil erosion continued to decrease in the SICS plot and at what rate.

About the dissemination of the project findings to more people who can benefit, the farmers suggested that they may share the results with other farmers either through the cooperative or through discussion with nearby growers. Another suggestion was the co-organization of events with local organizations, municipalities, or farmer cooperatives at the local level. Others proposed the local media. The consultants offered to organize training events for farmers in order to strengthen their skills on innovative soil improving mechanisms. Among the suggestions of the researchers for dissemination were informative brochures and workshops about the findings, in situ exhibitions of SoilCare case studies in Crete, video demonstration of SICS solutions, as well as guidance documents about new soil practices addressed both to farmers and agronomists.

As regards to the way the stakeholders would like to be supported in using or implementing project findings, the farmers seek subsidies for new machinery, seeds, including to avoid loans in the case of avocado investment. Several look for policy opportunities, guidelines for crop change, further development of the agricultural associations, and consulting services as well. The consultants seek additional seminars organized by government agencies on the way that they should train the farmers about the benefits of SICS. The researchers look for project funding for new SICS mechanisms and involvement of both farmers and stakeholders.

#### **4. Discussion and Conclusions**

Soil-improving cropping systems (SICS) application seems to play an alleviating role in soil loss processes, therefore it is recommended that farmers be properly informed about the tested practices within their fields.

No tillage practice is substantially beneficial for controlling soil erosion (over 14%), improving soil health and keeping good soil structure. Olive farmers should consider reducing tillage practices in olive orchards, control the tillage depth, and at the same time, limit its application especially during severe drought periods. In addition, the biological health and condition of the no-tilled plots were clearly better compared to the tilled ones. Water and solute movement as well as soil aeration were appropriate, including in the case of no-tillage. Weed management is a deterrent factor for this practice.

Vetch application is an inexpensive solution and is recommended to control soil erosion. The correct application of cover crop is a determinant in improving soil quality. Specifically, the biological health and condition of the vetch cover plots were clearly better compared to the no vetch. Furthermore, water and solute movement as well as soil aeration were slightly improved in the case of cover crop application.

Avocado farms, besides having significantly higher financial benefits, can also maintain a comparably overall good soil quality. However, the earthworm density experiment displayed that the biological health, and condition of the avocado plot were inferior to the orange tree plots. Conversely, water and solute movement as well as soil aeration were in good status for both cultivations, as identified by the top and bottom soil bulk density experiments. Moreover, the application of regular manuring resulted in higher values of SOC in the avocado field. The improvement of soil quality and structure through the increase of SOC as well as the control of soil erosion was additionally achieved by the organic material that accumulates from the intense foliage of avocados trees, resulting in a thick layer of organic material on the ground.

Different trends were found for erosion/deposition for the various cultivations and different treatment methods; however, the only statistically significant results were obtained for the orange to avocado treatment. In the olive tree cultivation, average erosion/deposition for the SICS plot with no tillage was 2.86 Mg/ha, ranging from 0.07 to 15.66 Mg/ha, depending on the observation periods. The control applied in this site was tillage twice in the period of study (November 2017 and May 2019). Tillage resulted to increased spread and mean erosion/deposition values of 3.33 Mg/ha (ranging from 0.43 to 11.42 Mg/ha). The application of crop cover treatment had a direct impact on mean and spread of soil erosion/deposition in the vineyard cultivation. The application of

vetch treatment reduced mean erosion by 13% (reduced from 3.41 Mg/ha in the control to 2.98 Mg/ha in the SICS plot). A similar trend was found for the crop change experiment. Mean soil erosion/deposition reduced by over 34% (reduced from 4.66 Mg/ha in the orange to 3.04 Mg/ha in the avocado plot) by changing orange to avocado trees according to the measurements. The spread of the magnitude of these processes was also reduced. Comparing the three different experiments, lower values of soil erosion/deposition were obtained for olive orchards, and this may be due to the lower slope of the plot (~6%). Regarding vineyard plot, although the study site is located in an area with a slope of ~15%, the small sediment losses can be mainly attributed to the higher concentration of silt and clay and low sand content of the soil. The sandy soil (55.3% sand) of the orange and avocado plot may be the main reason of higher erosion/deposition values observed. An additional reason for not monitoring significant differences in the two experimental fields may also be the short-term application of the SICS treatments, whereas the conversion of orange to avocado happened already in 1998 and thus can be considered as a long-term experiment.

This experiment demonstrates that soil improving cropping techniques have a significant impact on soil erosion, and as a result, on soil water conservation, which is of primary importance, especially for the Mediterranean dry regions. As reported in other studies, tillage erosion is considered to be one of the most important processes of land degradation in cultivated areas. The effect of tillage in soil erosion was also recorded during the SoilCare experiment, including for the minimum tillage practice that was applied as a soil-improving cropping method. Results of the study also show that crop cover treatment (vetch) and crop type change have a substantial impact on soil erosion/deposition (12% to 34% lower, respectively). The proposed sustainable soil improving practices are already being applied in many parts of the region. In particular, the change in procedure from orange to avocado trees has been adopted by many farmers as a response to the reduced orange prices and the high income from avocado cultivation. These results highlight the crucial role of soil-improving cropping systems for sustainable land management.

**Author Contributions:** Conceptualization and methodology, I.K.T., K.D.S., S.S., I.S.P., D.D.A. and A.G.K.; validation, S.S. and A.G.K.; formal analysis, S.S. and A.G.K.; data curation, K.D.S.; writing original draft preparation, S.S. and A.G.K.; writing—review and editing, I.K.T., K.D.S., S.S., I.S.P., D.D.A. and A.G.K.; visualization, S.S. and A.G.K.; supervision, I.K.T.; funding acquisition, I.K.T. All authors have read and agreed to the published version of the manuscript.

**Funding:** The research leading to these results has received funding from the H2020 under grant agreement n◦ 677407 (SOILCARE).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data sharing is not applicable to this article.

**Acknowledgments:** The authors would like to thank Ioannis Daliakopoulos and Anthi-Eirini Vozinaki for their support.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **Soil Quality Assessment after 25 Years of Sewage Sludge vs. Mineral Fertilization in a Calcareous Soil**

**Ana Simoes-Mota 1, Rosa Maria Poch 2, Alberto Enrique 1, Luis Orcaray <sup>3</sup> and Iñigo Virto 1,\***


**Abstract:** The aim of this work was to identify the most sensitive soil quality indicators and assess soil quality after long-term application of sewage sludge (SS) and conventional mineral fertilization for rainfed cereal production in a sub-humid Mediterranean calcareous soil. The treatments included six combinations of SS at different doses (40 t ha−<sup>1</sup> and 80 ha−1) and frequencies (every 1, 2 and 4 years), plus a control with mineral fertilization, and a baseline control without fertilization. Twentyfive years after the onset of the experiment, 37 pre-selected physical, chemical and biological soil parameters were measured, and a minimum data set was determined. Among these indicators, those significantly affected by treatment and depth were selected as sensitive. A principal component analysis (PCA) was then performed for each studied depth. At 0–15 cm, PCA identified three factors (F1, F2 and F3), and at 15–30 cm, two factors (F4 and F5) that explained 71.5% and 67.4% of the variation, respectively, in the soil parameters. The most sensitive indicators (those with the highest correlation within each factor) were related to nutrients (P and N), organic matter, and trace metals (F1 and F4), microporosity (F2), earthworm activity (F3), and exchangeable cations (F5). Only F3 correlated significantly (and negatively) with yield. From these results, we concluded that soil quality can be affected in opposite directions by SS application, and that a holistic approach is needed to better assess soil functioning under SS fertilization in this type of agrosystem.

**Keywords:** soil quality assessment; sewage sludge; long-term effect; Mediterranean soils

#### **1. Introduction**

In the framework of circular economy and the European Green Deal goals, land application of sewage sludge (SS) is suggested as one of the most economical and ecological sludge disposal methods [1,2]. When properly managed, it is seen as way to prevent environmental pollution [3], recycle nutrients, and decrease the need for commercial fertilizers [4,5]. Sewage sludge, in general, has a high content of organic carbon, nutrients (particularly N and P) [6], and trace elements (S, Mg, Ca), and can promote the proliferation and activity of soil micro and mesofauna [2]. As a consequence, amending soils with SS can improve some soil properties, such as organic matter and nutrient content, soil porosity, bulk density, aggregate stability, or available water holding capacity [2,3,7–9].

However, SS can also contain trace metals and persistent organic pollutants, which present a harmful risk to the environment and can be transferred to crops [10]. Indeed, larger studies on the effect of different organic amendments on soil quality [11] observed that the overall effect can be positive, although some aspects, such as soil contamination or grain quality, may be compromised, depending on the type of amendment used. In particular, the consequences of SS application on soil chemical properties and the accumulation of contaminants have been extensively studied [5,12,13] and different strategies to minimize the risk associated with SS application have been developed [14].

**Citation:** Simoes-Mota, A.; Poch, R.M.; Enrique, A.; Orcaray, L.; Virto, I. Soil Quality Assessment after 25 Years of Sewage Sludge vs. Mineral Fertilization in a Calcareous Soil. *Land* **2021**, *10*, 727. https://doi.org/ 10.3390/land10070727

Academic Editors: Julián Cuevas González, Guido Wyseure and Jean Poesen

Received: 15 June 2021 Accepted: 8 July 2021 Published: 10 July 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

In this framework, the European Commission implemented the EU Directive 86/278/EEC with regulatory guidelines on SS application and the concentration of toxic elements [15]. Other regulations exist at the national and regional level worldwide. Nonetheless, the consequences of the continuous application of SS on other soil quality indicators, particularly the interaction between indicators and their relation to soil functions, have received less attention [11]. These consequences are dependent not only on the composition and frequency of SS application, but also on the pedoclimatic and agronomic characteristics of each site [2,16,17]. The importance of considering all dimensions and properties of soil (in terms of its physical, chemical, biological, and organic matter properties) is intrinsic to any such study [17–19].

In the last decades, the literature on the study of soil quality has grown substantially, with an increasing emphasis on the inclusion of this concept associated with agriculture and other land uses, which should consider the many relationships between soil functions and the ecosystem services they provide [20,21]. Within the diversity of the approaches developed, there is an agreement on the existence of a series of basic steps in the evaluation of soil quality, among which the selection of appropriate indicators stands out for its special relevance [22]. Global reviews of these indicators [22–24] point out the most frequently proposed ones as those related to the organic fraction and soil reaction, together with those referring to the status of some nutrients, porosity (density), and water retention.

In any case, the selection of soil quality indicators needs to be made by simultaneously considering the soil functions and/or the services associated with them that are to be evaluated, and the local conditions imposed by the soil-climatic characteristics at the site under consideration. Some examples of the use of this approach in Navarre, Spain [25,26] for the evaluation of soil quality in agrosystems managed according to conservation agriculture criteria showed that the most appropriate indicators can vary in a relatively short lapse of time with a change of context, such as the transformation from rainfed to irrigated. The identification of these indicators, therefore, must also consider the management context.

In the particular context of farmlands in semiarid and sub-humid Mediterranean regions, which are usually depleted in organic matter [27], and therefore especially sensitive to soil degradation, SS addition to croplands has been seen as a promising practice, due to its high content in organic matter and nutrients [28]. Still, physical degradation of the soil may occur depending on the quality of SS, the doses and frequency of application, and the pedo-environmental conditions [29]. An adequate assessment of soil quality in this context needs, therefore, to be holistic (comprise chemical, physical and biological soil indicators of relevance in this type of agroecosystem [30]). In addition, adequate soil quality indicators have ben conceptually defined not only as sensitive to changes in soil condition, but also as precocious in their reaction as possible, easy to measure, and, if possible, available in common soil datasets [17,31]. Ideally, soil quality indicators should also comprise information measured at the field level (in addition to laboratory analysis), and be easily understandable by famers and policy-makers [32].

Regional studies after long-term applications, and with extreme rates of application, seem useful to better understand the actual effect of SS application on soil quality in these conditions [23]. In this context, this study aimed to identify the most sensitive soil quality indicators and, by studying their correlations, to understand the effect of the long-term application of SS on the overall soil quality of a cultivated calcareous soil after 25 years of SS application at different rates and doses by comparing it with conventional mineral fertilization in a controlled experimental field in Mediterranean sub-humid conditions. We hypothesized that the amount and frequency of SS used might induce differences in the chemical, physical and biological condition of this soil that might be interrelated and explained by the selected soil quality indicators.

#### **2. Materials and Methods**

*2.1. Site and Experimental Design*

The long-term experimental field site in Arazuri, Navarra, NE Spain (42◦48 N, 1◦43 W, 396 m a.s.l.) was established in 1992 to assess the effect of the continuous application of SS on agricultural soil quality and productivity. The climate in the area is temperate Mediterranean, with a humid water regime, according to Papadakis [33]. Mean annual precipitation is 750 mm year−1, and mean annual Thornthwaite's evapotranspiration, 687 mm year−<sup>1</sup> [34]. The soil in this field is calcareous (approx. 20% of calcium carbonate in the tilled layer) with a clay-loam texture in the topsoil (31% clay, 30% silt, 39% sand) [35], is well-drained and has no salinity problems. It has been classified as a Calcaric Cambisol [36]. The soil's main physical–chemical characteristics in the tilled layer (0–30 cm) at the control plots are summarized in Table 1.

**Table 1.** Physical and chemical properties of the soil tilled layer (0–30 cm) for the control plots. Values are given as the mean ± standard deviation (*n* = 3).


The experimental design consists of a random factorial block design with eight treatments with three replicates (*<sup>n</sup>* = 3), each plot with an area of 35 m2 (10 m × 3.5 m). The treatments included six combinations of SS at different doses (40 t ha−<sup>1</sup> and 80 ha−1) and frequencies (every 1, 2 and 4 years), plus a control with the usual mineral fertilization in the area (46% urea and ammonium sulphate), and a baseline control without SS or mineral fertilization. Sewage sludge treatments were denoted after the dose and frequency (40-1, 40-2, 40-4, 80-1, 80-2, and 80-4). Mineral-fertilized and baseline controls were noted as MF and C, respectively. Both doses and frequency of SS were chosen according to the common practices in the area, and to get the highest possible rates in the plots with high doses and frequencies.

The crops used corresponded to the most frequent rainfed rotation of 3 years in the area (cereal–cereal–no cereal), managed with annual tillage with a 30 cm deep moldboard plow, and application of phytosanitary products according to the crops' needs each year. The most common cereal crops used were wheat (*Triticum aestivum* L.) and barley (*Hordeum vulgare* L.), with sunflower (*Helianthus annuus* L.) as the non-cereal crop in the rotation. Sewage sludge was produced in the urban wastewater treatment plant from the city of Pamplona (population 330,000), with primary and secondary treatments, stabilized through anaerobic digestion and mechanical dewatering. Sewage sludge characteristics, as described by [10], are summarized in Table 2. The SS was applied each campaign in September, 3 to 4 weeks before sowing, using a 3.5 m wide spreader trailer, followed by moldboard plowing down to 30 cm. Mineral fertilization with a commercial fertilizer purchased from a local provider was carried out before sowing. Wheat sampling was carried out in June in the year of study, at harvest. Each field replicate was harvested with a plot-scale combine, and grain yields were recorded. Grain weights were taken directly from the combine, and grain samples were collected to analyze their water content, to obtain yield data on a dry-mass basis.


**Table 2.** Physical and chemical properties of the sewage sludge. Values are given as the mean ± standard deviation (*n* = 3).

#### *2.2. Soil Sampling and Analysis*

Soil sampling was carried out 25 years after the onset of the experiment, in September, at each treatment and replicate at two depths (0–15 cm and 15–30 cm) after the crop cycle was completed, and at the furthest moment in time from previous soil alterations, except for samples for the physical properties, which were sampled in June before harvesting to avoid the possible effect of harvesting machinery. Disturbed and undisturbed samples were collected for the various analyses. Disturbed soil samples were collected for the 0–15 cm and 15–30 cm depths using an Edelman-type auger (Ø = 5 cm) or a shovel. Three subsamples were collected per plot for each depth increment and combined to obtain a composite sample. Immediately after sampling, a portion of the composite soil was stored at 4 ◦C for further biological analyses. Part of the sample was gently pushed through a 6 mm sieve. These aggregates were air dried and used for aggregate stability determinations. The remainder of the soil was air-dried and ground to pass through a 2 mm sieve. Undisturbed core samples were collected in triplicate using bevel-edged steel rings (Ø = 5 cm, total volume = 100 cm3) for the 0–15 cm and 15–30 cm depth increments to determine soil bulk density (ρb), permeability, and water retention characteristics. Undisturbed soil samples were also collected using Kubiena boxes for thin section analysis. For earthworm population assessment, two 20 × 20 × 30 cm soil blocks were extracted from each treatment in all replicates.

#### 2.2.1. Soil Physical Properties

The soil's physical condition was assessed using properties related to compaction and porosity, aggregation, and water flow and storage. Bulk density and penetration resistance (PR) were measured to assess compaction and porosity. As explained above, the core method was used to determine ρb [37]. Penetration resistance was measured at 9 points per field replicate to a depth of 60 cm using a field penetrometer (Rimik CP20, Agridy Rimik Pty Ltd., Toowoomba, Qld, Australia). Measurements were made after a rainy period to avoid differences in water content between treatments. Measurements were recorded every 15 mm, and PR for 0–15 and 15–30 cm were calculated as weighted depth averages.

Dry aggregate stability was determined by placing 100 g of dry aggregates (<6 mm) in the top of a column of sieves of 4, 2, 1, 0.5, and 0.25 mm openings, and shaking in a rotary movement at 60 strokes/min for 60 s in a Retsch VS 100 device (Retsch GmbH & Co., Haan, Germany). For wet aggregate stability, a constant shower-like flux (6 L/min) of distilled water was applied from the top of the same set of sieves while sieving (60 strokes/min, 60 s). We used a mechanical sample divisor (Retsch GmbH & Co., Haan, Germany) to ensure that the initial distribution of aggregates was similar among replicates. Aggregate size distribution and stability were expressed as the mean weight diameter (MWD) after dry and wet sieving [38]. The stability of the aggregates was also evaluated using the mass proportion of water-stable aggregates (WSA) > 0.25 mm [39]. Soil saturated permeability (Ks) was measured on undisturbed soil cores after saturation with deionized water under a vacuum using a laboratory permeameter (Eijkelkamp Soil & Water, Giesbeek, The Netherlands).

Soil water retention at −33 kPa, −50 kPa, and −90 kPa was determined on intact soil cores, and sieved (<2 mm) soil samples were used for water retention assessment at −1500 kPa. Samples were placed on pressure plate extractors (Soil Moisture Equipment Corp., Santa Barbara, CA, USA). Volumetric water was calculated using ρb. Available water-holding capacity (AWHC) was calculated as the difference between volumetric water content at field capacity (−33 kPa) and wilting point (−1500 kPa). From these data, as described in [40,41], the model proposed by [42] was used to estimate the equivalent pore diameter corresponding to each of the water potentials. According to this model, the equivalent pore diameter was 9 μm for −33 kPa, and 0.2 μm for −1500 kPa. This allowed us to obtain the equivalent size ranges of micropores in each sample, expressed as the proportion of each pore range (<0.2 μm, 0.2–9 μm, and >9 μm), as well as the proportion of pores able to retain water (0.2–9 μm) over those able to store water available for plants (>0.2 μm). These were denoted as PØ < 0.2, PØ 0.2–9, PØ > 9 and PØ 0.2–9 (>0.2), respectively.

Soil thin sections were prepared from undisturbed soil samples as described in [43]. Image analysis was used in these sections to determine parameters related to macroporosity. For this, a scanned image was obtained per thin section under two light conditions: parallel polarizers and crossed polarizers. They were processed using Image J [44] to obtain digital binary images. From each binarized thin section, five random images (10 × 10 mm) were selected using an adaptation of the method used by [45], where a grid of 27 squares (1 cm<sup>2</sup> each) was placed in each scanned section from which the eligible squares were chosen using a random number generator. From these, pore-size distribution analysis was carried out based on an open mathematical algorithm: the Quantim4 library [46]. The area occupied by pores was divided into five intervals according to the pore's apparent diameter: 100–400 μm; 400–1000 μm; 1000–2000 μm; >2000 μm. The proportion of the area (equivalent to volume proportion over total soil volume) occupied by pores with diameters between 400–1000 μm was selected for this study because of their special relevance when describing structure (size of planar voids or fissures), and also because these pores can result from the activity of mesofauna [47].

#### 2.2.2. Soil Chemical Properties

All chemical analyses were performed on air-dried sieved (<2 mm) samples. Total N was analyzed using the Kjeldahl digestion method. Available P was determined as described by [48]. Exchangeable K and Na were quantified using atomic absorbance after extraction with NH4OAc 1N [49]. The soil electrical conductivity (EC) and soil pH were measured in distilled water (1:2.5). Soil pH was determined with a Crison GLP22 pH meter (Crison Instruments, S.A., Barcelona, Spain). Conductivity was read with a Crison GLP32 conductivity meter (Crison Instruments, S.A., Barcelona, Spain).

Carbonate concentration was measured in a modified Bernard's calcimeter [50] by quantifying the CO2 produced after treating a soil sample with HCl. Available trace metals (Cu, Mn, Ni, Zn, Cd and Pb) at the 0–15 cm depth were analyzed as DTPA(C14H23N3O10) extractable concentrations from air-dried soil samples, using the extraction procedure described in the international standard ISO 14870:2001 [51], as described in [10]. In short, an extraction solution was prepared by mixing, first, 0.735 g of CaCl22H2O, 0.984 g of DTPA and 7.46 g of triethanolamine (C6H15NO3), diluted with 800 mL of deionized water, and the pH was adjusted to 7.3 with HCl. Subsequently, in a 100 mL wide-mouth polypropylene container, 20 g of soil and 40 mL of the solution were mixed and stirred for 2 h at 20 ◦C on a reciprocating shaker at 30 rpm. Then, a fraction of the extract was decanted and centrifuged for 10 min at 6000 rpm. The supernatant was filtered with a membrane filter with a pore size of 0.45 μm and collected for analysis. The extracts were analyzed earlier than 48 h from their preparation, by ICP-MS in a 7700x analyzer (Agilent Technologies, Santa Clara, CA, USA), following the UNE-EN 17053 standard [52].

#### 2.2.3. Soil Organic Matter and Biological Properties

Soil organic C (SOC) was determined by wet oxidation on air-dried sieved (<2 mm) samples [53]. The fraction of soil organic matter defined as particulate organic matter (POM) based on its size (>53 μm) [54] was isolated by dispersion and sieving of 10 g of air-dried soil [55]. Organic C in the form of POM (POM-C) was determined by wet oxidation.

Earthworms were collected crumbling the 20 × 20 × 30 cm soil blocks by hand, placing the worms in a glass jar, and weighing to obtain a fresh weight for each field replicate [56]. This allowed us to determine the total biomass (g per m−2), the abundance (number of individuals per m<sup>−</sup>2), and the average size (g per individual).

Microbial biomass carbon (MBC) was measured by comparing extractable C from non-fumigated and chloroform (CHCl3)-fumigated soil [57]. Carbon concentration in the extract (chromic acid dissolution) was analyzed by sulfuric digestion and subsequent spectrophotometry. The functional diversity of the soil microbial population was studied through the analysis of the community-level physiological profiles (CCPLs) in fresh samples by studying the C source utilization patterns observed using a Biolog EcoplatesTM microplating system (Biolog, Hayward, CA, USA), as described in [58]. EcoplatesTM were designed for determining CLPPs of terrestrial communities and comprise 31 C substrates that are major ecologically relevant compounds. One g equivalent dry weight of soil was mixed with 9 mL of autoclaved Mili-Q ultra-pure water and shaken in an orbital shaker at 125 rev min−<sup>1</sup> for 1 h. After shaking, samples were left to settle and then a 1:100 dilution was inoculated onto Biolog Ecoplates™. The plates were incubated at 30 ◦C and color development was read twice a day at 595 nm using a microplate reader (Thermo Scientific Multiskan® EX Waltham, Massachusetts, USA). Average well color development (AWCD) was determined by calculating the mean of every well's absorbance value at each reading time. The number of substrates used by the soil microbial community (NSU), equivalent to species richness [59], was quantified as the number of wells showing corrected absorbance values >0.25 at the onset of the exponential microbial growth in the Biolog EcoplateTM microplates (59 h).

#### *2.3. Statistical Analysis*

The selection of the most sensitive soil quality indicators, and the assessment of the effect of SS application on soil quality, was conducted in a two-step procedure, as in [26,60]. First, a multivariate analysis of variance (MANOVA) was performed to test whether there was a significant effect of the categorical independent variables (fertilization and depth) on at least one of the physical, chemical, or biological variables studied. For this study, the eight fertilization treatments (SS application, MF and C) were considered as a factor, because, despite some treatments receiving equivalent cumulative amounts of added SS, for some soil properties, the effect of SS or mineral fertilization might be different depending on the frequency, whereas some others might be affected by the accumulation of SS applications with time.

Then, a univariate analysis of variance (ANOVA) for the different soil variables was performed to examine for significant influences of fertilization treatment and depth. Only the variables for which the *F* statistic for SS application or fertilization treatments was significant (*p* < 0.05) were retained for further analysis. In a second step, factor analysis was used to group the retained variables into statistical factors based on their correlation structure. Principal component analysis (PCA) was used as the method for factor extraction. To eliminate the effect of different units of variables, factor analysis was performed using the correlation matrix on the standardized values of the measured soil properties; each variable had mean = 0 and variance = 1 (total variance = number of variables [61]). We used the determinant of the correlation matrix as an indicator to identify the existence of correlations among variables. As in Imaz et al. (2010) [25] and Apesteguía et al. (2017) [26], using the correlation matrix, principal components (factors) with eigenvalues > 1.5 were retained and subjected to varimax rotation with Kaiser to estimate the proportion of the variance of each attribute explained by each selected factor (loadings), and by all factors (communalities). A high communality for a soil attribute indicates that a high proportion of its variance is explained by the factors. In contrast, a low communality for a soil attribute indicates that much of that attribute's variance remains unexplained. Less importance should be ascribed to soil attributes with low communalities when interpreting the factors [62].

To evaluate the effects of the studied SS application or mineral fertilization treatments on the extracted factors, factor scores for each sample point were calculated and ANOVA was performed on the new score variables. Homogeneous groups among treatments were detected using Duncan's test (*p* < 0.05, unless otherwise indicated). Only factors that differed among treatments were retained for further consideration. Soil attributes were then assigned to the factor for which their loading was the highest [61]. For each retained factor, highly weighted attributes were selected as possible soil quality indicators. We considered as highly weighted those within 10% of the highest factor loading, as in [63].

Finally, a correlation analysis was conducted for wheat yields against the scores of the extracted factors with PCA. For all analysis, the significance level was set at *p* < 0.05 unless otherwise indicated. All statistical treatments were performed using IBM SPSS Statistics 27.0 [64].

#### **3. Results**

#### *3.1. Identification of Indicators*

The different treatments and sampling depth significantly affected the physical, chemical, and biological properties evaluated. The analysis of variance for each individual parameter studied showed some significant differences for the different treatments and depths, and for some parameters, a significant interaction of both (Table 3). As the goal of this study was to identify the most sensitive soil quality indicators and to assess overall soil quality from PCA, and not to assess each individual parameter, only the significance of the analysis is provided for each parameter. As depth had a significant effect in some of the parameters studied, the factor analysis performed to select soil quality indicators was performed separately for the two sampling depths. Soil parameters that were not significantly affected by treatment per studied depth were not considered for the following analysis.

At the 0–15 cm depth, these were Ks, water retention at −33 kPa and −90 kPa, AWHC, PØ < 0.2, PØ 400–1000 in equivalent diameter, carbonates, available Mn, available Pb, the POM-C/SOC ratio, MBC, diversity indexes (AWCD and NSU), and earthworm abundance (individuals m−2). At the 15–30 cm, among those studied at that depth, the parameters excluded for PCA were all physical parameters, except for PØ < 0.2, the POM-C/SOC ratio, and the microbial functional diversity indexes (AWCD and NSU).

#### 3.1.1. 0–15 cm Depth

The correlation matrix for the 21 selected indicators (determinant < 0.0001) showed several significant correlations on 101 pairs out of 210 (Table S1). The highest positive and significant correlations were found between available P vs. available Zn, available Cu, and available Ni (*p* < 0.001), and also between SOC vs. available Zn and available Ni (*p* < 0.001).

The PCA identified three factors (F1, F2 and F3) with eigenvalues > 1.5 for the 0–15 cm depth, which together explained 70.2% of the variance of the 21 selected indicators (Table 4).

The soil properties with high loadings for these factors were considered potential good soil quality indicators (Table 5). F1 showed high loadings (within 10% of the one with the highest loading, or close) for available P, EC, available Zn, available Cu, available Ni, available Cd, available Pb, and SOC. F1 can therefore be associated to organic matter and chemical parameters. F2 showed high loadings for water retention at −50, the total proportion of pores 0.2–0.9 μm (PØ 0.2–0.9), and the proportion of pores 0.2–0.9 μm over total pores > 0.2 μm (PØ 0.2–9(>0.2)). F2 can therefore be associated with water retention in the soil. Finally, F3 grouped earthworms' biomass (g m−2) and earthworm average size (g i<sup>−</sup>1) as the properties with the highest loading. F3 would represent the behavior of earthworm populations as affected by treatments in this field.


**Table 3.** Results of the analysis of variance (ANOVA) for all soil properties.

**Table 4.** Eigenvalue, percentage, and cumulative variance explained by factor analysis using the correlation matrix of the standardized data of soil parameters at 0–15 cm (F1, F2 and F3) and at 15–30 cm (F4 and F5) depths.


<sup>1</sup> Only factors with eigenvalues > 1.5 are shown.

#### 3.1.2. 15–30 cm Depth

Using the nine selected indicators for this depth, a correlation matrix was developed for the 15–30 cm depth (determinant < 0.0001), which showed significant correlations on 17 pairs out of 36 (Table S2). The most significant correlations (*p* < 0.001) were observed between available P vs. SOC and POM-C, EC vs. SOC and POM-C, and finally SOC vs. POM-C.


**Table 5.** Proportion of variance explained using varimax rotation for each of the retained factors and communalities for the selected soil properties for the 0–15 cm depth.

The PCA extracted two factors (F4 and F5) with eigenvalues > 1.5, explaining 68.5% of the variance between indicators at this depth (Table 4). For F4, the highest loadings corresponded to EC, available P, and SOC. Like F1 at 0–15 cm, F4 can be associated to organic matter and chemical parameters. Regarding F5, exchangeable K and Na were the properties with the highest loadings. F5 would therefore represent exchangeable cations (Table 6).

**Table 6.** Proportion of variance explained using varimax rotation for each of the retained factors and communalities for the selected soil properties for the 15–30 cm depth.


#### *3.2. Sensitivity of PCA Factors to Treatment*

All factor scores were sensitive to treatments (Table 7). The scores for F1 were significantly different in C and MF than in all treatments with different doses of SS. The scores for F2 differed significantly in C, MF, 40-4, 80-1, and 80-2 from 40-1, 40-2, and 80-4. The scores for F3 were significantly different in C than in MF, 80-2, and 40-2, with the rest of treatments showing intermediate values.


**Table 7.** Effect of treatment on factor scores from PCA (*p* < 0.05).

In each column, different letters denote different Duncan's homogeneous groups.

Factor scores for both F4 and F5 were sensitive to treatment (Table 7). For F4, scores were significantly different for C and MF than for all treatments receiving SS. Among them, 80-1 had the highest load, and all other treatments with SS displayed intermediate values. Finally, F5 scores were different for 40-1 than for 80-2, 80-4, and MF, with the other treatments showing intermediate values.

#### *3.3. Yield*

Average wheat yield was statistically different among treatments (*p* < 0.001, Table 8). MF treatment had the highest values, followed by 40-1, 40-2 and 40-4. The treatment with the highest rate of sludge, 80-1, showed the lowest yield apart from the baseline control treatment (C), for which yield was less than half compared with the MF and 40-1 treatments. No significant correlations were observed for the factor scores and yield, except for F3 (Pearson's correlation coefficient = −0.638, *p* < 0.05).

**Table 8.** Crop yield results (kg ha−1) treatments with the same letters are not statistically different (*p* < 0.001). Values are given as the mean ± standard deviation (*n* = 3).


Different letters denote different Duncan's homogeneous groups.

The interaction between yield and F1, F3, and F4 scores, as well as the interaction between F1 and F4 grouped by treatment are represented through scatter plot graphics in Figure 1.

**Figure 1.** Relationship between soil quality assessment factors selected through PCA and yield. Treatments (*n* = 3). (**a**) Yield × F1; (**b**) yield × F3; (**c**) yield × F4; (**d**) F1 × F4.

#### **4. Discussion**

*4.1. Selection of Soil Quality Indicators*

4.1.1. Sensitivity to Management

The results of the preliminary ANOVA (Table 3) indicated that some of the pre-selected soil properties were indeed sensitive to SS application and fertilization management in the experimental field used in this study. This response was also different for the two studied depths. At 015 cm, most of the physical and chemical indicators originally considered were shown to be sensitive to treatments, as well as organic matter indicators and earthworms, whereas indicators related to the soil microbial community seemed to be less sensitive. The sensitivity of those preselected was, however, lower in the 15–30 cm depth, despite all treatments receiving annual inversion tillage at 0–30 cm, which suggests a depth stratification in the response to treatments. The relevance of the stratification of the response of soil properties to management [39] and the relatively low sensitivity of SOC stratification have been largely discussed [65,66].

Regarding the selection of the most sensitive indicators, at 0–15 cm, sensitive indicators included water retention at −50, microporosity (PØ > 0.2–9, PØ > 9 and PØ > 0.2–9(>0.2)), PR, aggregate stability (WSA, MWDd, MWDw), available P, total N, EC, pH, extractable Na, available trace metals except for Mn and Pb, SOC, POM-C, earthworms' biomass, and average size. At 15–30 cm, they were PØ < 0.2, available P, total N, EC, pH, exchangeable

Na, and K, SOC, POM-C, and MBC. The most sensitive indicators in the physical indicator group were therefore those related to water retention, microporosity, or aggregate stability. These soil properties are well known to be sensitive to changes in soil when SS or other organic amendments are used [3,67], as they represent the changes induced in soil structure as a response to increased inputs of organic matter [68].

On the contrary, non-sensitive indicators included ρb, Ks, AWHC, and PØ 400–1000. This can be explained, at least partially, by the fact that all treatments receive the same mechanical management, comprising annual moldboard tillage, seedbed conditioning, and seeding, as well as mechanical harvesting. In terms of ρ<sup>b</sup> and Ks, the effect of these operations can counteract that of the addition of organic matter with SS, or from crop residues. Other studies have also shown that ρ<sup>b</sup> can be less sensitive to changes in the soil's physical condition associated to changes in organic matter than other physical indicators, such as those related to aggregate stability. For instance, [69] found no differences in ρ<sup>b</sup> between a conventional tillage treatment and long-term no-tillage inducing gains in SOC on a silt loam soil, while differences were found among tillage treatments concerning WSA and MWD.

These results can also be understood as some changes in the soil's physical condition being more noticeable than total porosity or AWHC. In this sense, there is evidence that some soil pore-size ranges can be more affected by changes in SOC gains than others. Indeed, Kirchmann et al. (2002) [70] reported microporosity (1–5 μm) as more sensitive to management than macropores in an Inceptisol amended with different exogenous organic matter, which they related to changes in SOC concentration. This coincides with our observation of a greater sensitivity of micro porosity indicators, especially in the 0–15 cm depth, than PØ 400–1000 in size, which may indicate that this porosity interval is more related to management (planar voids, fissures due to tillage), equal to all treatments, than to the activity of mesofauna.

Among chemical indicators, those related to nutrients (N and P), EC, and pH, as well as trace metals, proved to be the most sensitive ones at 0–15 cm. The response of these indicators to SS addition and/or mineral fertilization was expected, and has been reported in many previous studies on the use of SS in agriculture [12,71,72]. In particular, the accumulation of trace metals with SS addition has been already reported and studied in this soil [10,73]. Changes in pH and EC have also been systematically reported in soils amended with SS, in contrast with MF or non-amended soils [12,74,75] and related to the content in soluble salts in SS (Table 1).

Carbonate content was included in the original collection of indicators because it has been observed that the repeated addition of SS and other sources of organic matter can result in changes in the amount and typology of soil carbonates [76,77]. The content in carbonates of the studied soil (Table 1) seems, however, elevated enough not to be sensitive to these changes, although the observed sensitivity of pH (Table 3) suggests that some changes could be expected in the future if the repeated addition of SS results in some sort of acidification in this soil. In fact, changes in pH after SS application can occur due to proton release due to nitrification process, as observed by Tamir et al. (2013) [78] and Huang & Chen (2009) [79] after application of animal manure or sewage sludge compost, respectively, which causes CaCO3 dissolution. For instance, Eid et al. (2021) [80] have recently reported a significant decrease in soil pH (from 8.5 to 7.7) in a short-term pot experiment when the soil was incubated with SS at doses > 30 g kg−<sup>1</sup> soil.

At 15–30 cm (no data on trace metals were available), exchangeable K and Na were also observed to be sensitive. A positive correlation was found between Ex Na and pH (Table S2). Other studies confirm this correlation [9,81] between exchangeable cations and pH, related to the increase in the amount of exchangeable cations, Na and K in this case, resulting from the leaching process, contributing to the pH acidification. As in most of the studies conducted on SS application [8,12,82], SOC and POM-C were seen to be highly sensitive to treatments at both depths [83]. In our study, however, an important observation was that while both SOC and POM-C revealed to be sensitive to the treatment, the proportion of POM-C over SOC was not. This suggests that the differences in SOC and POM-C between treatments would correspond to the amount of both, and not to differences in the quality of organic matter, or at least, to the proportion of labile C. Other studies using SS have also observed that the long-term addition of SS results in changes mostly in the stock of SOC, rather than in differences in its composition [84], which they attributed to crop residues being the most relevant source of SOC compared to SS.

Finally, in relation to biological indicators, several studies have reported changes in the soil microbial biomass or diversity with SS application [75,85,86], since the microbial community is triggered by the increase of labile C [87], which was indeed sensitive to treatments in our experimental field (Table 3). However, other studies, such as Urra et al. (2019) [73], conducted in the same experimental field, or Picariello et al. (2020) [88] in an incubation experiment, did not find a significant response of soil microbial biomass to the treatments studied, as was the case in our study at 0–15 cm. This might be due to the changes observed within soil chemical parameters. For instance, Lloret et al. (2016) [89] showed that changes in EC can hinder microbial activity in a calcareous soil. Similar to our results, Roig et al. (2012) [7] found no correlation between basal respiration, a known indicator of soil biological activity, and the use of SS in the same experimental field 10 years before our collection of samples. In this sense, it has been reported that the soil microbial community can be more sensitive to tillage practices than to soil organic matter management [90,91]. Earthworm indicators were, however, clearly sensitive to treatments in our study. Their response to soil and organic management has been widely studied, and reported in the sub-humid and semi-arid areas of the region [55,92,93]. In contrast, at the 15–30 cm depth, MBC was found to be significantly sensitive to treatments (Table 3). This suggests some stratification of microbial biomass, as observed for other indicators such as ρ<sup>b</sup> and PR, also showing significant differences with depth. As stated above, microbial biomass can respond better to tillage and changes in the soil physical–chemical condition than to changes in organic matter, as can be seen by the lack of significant correlations between MBC and the organic matter parameters at this depth (Table S2). In fact, as in the study by [89], significant correlations were found between MBC, EC, and pH at this depth (Table S2).

In summary, those indicators showing the highest sensitivity to management included some of the originally selected ones, but not all. Among those with the highest sensitivity, physical, chemical, biological, and organic-matter related indicators were included, which supports the idea of a holistic approach being needed to understand changes in soil when SS is tested as an agricultural amendment.

#### 4.1.2. Grouping and Selection of Indicators

The most significant correlations were observed for P and SOC with trace metal availability at 0–15 cm, and for P and EC with SOC in the 15–30 cm depth. These correlations suggest that the addition of SS, which overall implied the addition of different accumulated doses of organic compounds, also implied an enrichment in P, trace metals, and, very likely, soluble compounds. Zoghlami et al. (2020) [12] reported a concomitant change in SOC, P, and exchangeable Na with higher doses of SS application. A correlation between organic matter accumulation and an increase in EC has been also observed in similar studies [12,75]. These observations again put in evidence the relevance of paying attention to all consequences of the addition of SS when assessing their effects in soil, as well as the interaction among soil properties. The correlations observed at both depths were reflected in the results of the PCA. At the depth of 0–15 cm, the selected indicators had different loadings in the three factors retained (F1, F2 and F3), so that F1 received high loadings from organic matter and chemical parameters, while F2 was mostly associated to water retention, and F3 to earthworm populations (Table 5). These results suggest that that the responses of soil water retention and earthworms were not directly correlated to that observed for organic matter, P, and trace metals. This can be explained by different means.

First, although the soil's physical condition and porosity are known to interact with soil organic matter in most soils, the observed discrimination of water retention indicators (F2) from those related to SOC and nutrients (F1) can be related to the particular mineral composition of this soil, which contained 16% carbonates in the studied depth (Table 1). Carbonates are known to interfere with the soil physical stability [94], and can be a factor of stabilization of soil structure in calcareous soils, making it less dependent on SOC than in other soil types [95,96]. In addition, the observed correlation between SOC and EC at both depths and exchangeable Na (at 15–30 cm) suggests that those treatments displaying SOC gains would also result in an increase in soluble salts, which are a known factor of soil structure destabilization [68]. Second, the lack of correlation between earthworm indicators (F3) and those related to SOC and nutrients (F1) and water retention (F2) indicated that their presence and abundance did not directly respond in this soil to the amount of organic C stored in each treatment, nor to physical indicators related to water retention. This suggests that their activity in the studied soil would be dependent on other factors such as toxicity or compaction.

In this framework, following the criterion for selecting the soil attributes with the highest sum of correlation coefficients (Table 5) as the most appropriate soil quality indicators [25,63], our results showed that SOC, available P and trace metals, microporosity and water retention at low water potential (−50 kPa), and earthworms would be those selected at the 0–15 cm depth. Available P would also be selected at 15–30 cm, together with SOC and EC, and exchangeable monovalent cations (Na and K). It must be noted that EC was also a secondary driving factor at 0–15 cm for F1. The relevance of SOC as an indicator of changes in soil resulting from exogenous organic inputs is logical and has been demonstrated in many cases [3,12,84,97]. Indeed, in a recent study conducted at a regional level, the addition of exogenous C has been proven to be the most efficient strategy to increase SOC stocks in the region of study [98]. SOC is also known to correlate well with other fertility indicators, such as the cation exchange capacity, in soils where clay mineralogy is rather stable, like the one used in this study. The linear relation between applying SS to the soil and the enhance of P, trace metals, and soluble salts, as stated above, is also well documented [12,81,99]. The calcareous nature of this soil can explain, at least partially, the accumulation of both P and trace metals [100], resulting in these indicators displaying a high correlation with the addition of SS. Their value as indicators for this type of soils seems relevant, as both are related to environmental risks. At the same time, the selection of earthworms as the most sensitive biological indicator supports their increasingly recognized role as universal soil biological indicators [22,101,102].

Finally, it can be noted in relation to the selection of indicators that, although the pre-selection of indicators was performed based on expertise, the approach used in this study was statistical. This approach has been seen to sometimes result in unexpected or contradictory selection of indicators [11]. Nevertheless, in our study, the most sensitive indicators were in harmony with most soil quality assessments [22], and seemed adequate if the aim of soil quality studies is to provide practical information, with low cost analysis and with influence on the ecosystem services provided by the soil in the particular conditions of calcareous soils under SS application.

#### *4.2. Soil Quality Assessment*

A soil quality assessment can be performed based on the scores of the factors selected, and on the link between these factors and the soil functions under study, and the ecosystem services provided by these functions [19,22]. In this case, the goal being to test the effect of SS and MF on an agricultural soil, the main function to be assessed would be biomass production (yield).

Our results showed that amending the studied soil with SS can result in similar yields as with mineral fertilizers, as reported by Jaber et al. (2005) [103] and Obriot et al. (2016) [11] on the use of municipal solid waste (MSW). When compared to the control, yields were high and like MF in the SS treatments with intermediate doses (40-1; 40-2; 40-4; 80-2; 80-4), but the treatment with the highest dose (80-1) implied a decrease in yield (Table 7). The same was previously documented by Mantovi et al. (2005) [72], who reported lower yields on the highest doses on a on a winter wheat–maize–sugar beet rotation fertilized, due to excess of N and wheat lodging. Differently, Cherif et al. (2009) [104] observed that a high dose of municipal solid waste compost (80 t ha<sup>−</sup>1) enhanced wheat yield by 239%.

In relation to the relationship between factors issued from PCA and yield, for F1, PCA discriminated the treatments with intermediate doses (40-1; 40-2; 40-4; 80-2; 80-4) from treatment 80-1 with the highest accumulated SS dose (Table 7). In addition, both MF and C were differentiated from the SS treatments for this factor. In the 15–30 cm depth analysis, F4, which, as F1, was associated with nutrient dynamics, trace metals, and organic matter, also displayed a clear discrimination of MF and C from the treatments receiving SS, and treatment 80-1 from the other treatments with SS (Table 8). A significant correlation was indeed found between F1 and F4 (Figure 1), indicating that the effects of the treatments tested in this group of soil properties were similar in the two studied depths. This suggests that the continuous application of SS as an organic amendment to the agricultural soil of this study had a different impact on the soil chemical indicators and organic matter than mineral fertilization, or even no fertilization, and that, within those treatments with SS, the dose would also have different effects in this sense.

However, although F1 and F4 correlated with nutrient dynamics, organic matter, and trace metals, no significant correlation was found between these factors and yield. This can be explained because the few indicators selected by our analysis of F1 and F4 (available P, total N, EC, trace metals, and SOC) are known to have implications on yield, which act in opposite directions [105]. Indeed, as can be seen in Figure 1a, the only treatment not following a correlation between F1 scores and yields was the one without any type of fertilization (treatment C). Among the other treatments, the correlation was negative, suggesting that the increasing use of SS (cumulative dose) would result in an overall negative effect on the scores in F1 and F4 [72]. From another point of view, and concerning trace metals, their high loadings in F1 indicate that the assessment of their bioavailability can be of use to assess changes in soil (which may affect yield) in the conditions of the study. The studies [10,73] have recently explained the link between SS application and trace metal accumulation in the soil and crops of this experimental field.

In relation to soil physical indicators, under the site environmental conditions, F2 appeared as a relevant factor to assess soil functions related to structure and water storage. However, the scores of F2 did not show a clear trend among treatments or different fertilization management practices and were not either correlated to yield. Still, the high loadings for microporosity and water storage parameters suggest that they might be of relevance when assessing soil quality in the field under study.

Finally, on the biological condition of this soil, factor F3 (obtaining the highest loadings from earthworm indicators) was the only factor that significantly correlated with yield. The analysis of this factor separated the C treatment from those fertilized (MF and all treatments with some amount of SS), supporting its potential as an indicator for changes induced in the soil by mineral or organic fertilization. The correlation with yield was negative, suggesting that the changes induced by mineral and organic fertilization in this soil, which would result in increased yield, might be detrimental for earthworm populations. Indeed, reports on the effect this type of fertilization has on the earthworm population are contradictory [106]. A better understanding on the relationship between earthworms and soil management in calcareous soils like the one studied here is needed. In the region, it has been observed that earthworms can be positively affected by the reduction of tillage, and the concomitant gains in SOC in long-term trials [55,92,107]. Our results suggest that, when conventional tillage is used, fertilization may become a major driver of earthworm abundance in this type of soil.

Overall, the soil quality assessment on a crop field with 25 years of SS application revealed several implications regarding this type of fertilization. Sewage sludge had a direct effect on nutrient and organic matter input, as well as on trace metals. Yield

results indicated that the soil amended with SS was capable of accomplish similar yields as with MF.

#### **5. Conclusions**

The goal of this study was to identify the most sensitive soil indicators to assess changes in soil quality after long-term application of SS and MF on a cultivated calcareous soil, and to understand these changes in a controlled experimental field in Mediterranean sub-humid conditions.

A selection of physical, chemical, and biological indicators, as conducted using PCA in this study, was possible, resulting in SOC, available P, total N, EC, trace metals, and earthworms as the most sensitive indicators of changes in the calcareous soil of study. These showed up, therefore, as the most reliable indicators in the long-term monitoring of the effect of SS application in the conditions of the study. These indicators have been frequently identified in other studies on the response of agricultural soils to management and are, in general, commonly reported and easy to monitor.

The study also showed that the overall response of soil quality to the managements tested (SS application and MF) was not linear or straightforward. As hypothesized, the amount and frequency of SS used induced differences in the soil's chemical, physical, and biological condition. However, the overall effect of SS application was more evident on organic matter, nutrients, and trace metals than on the soil's physical condition or earthworms. However, physical indicators and earthworms were highly sensitive to management, and therefore seem useful for assessing changes in soil, not necessarily related to yield. Indeed, the response of the factors issuing from PCA to the treatments tested, and their correlation with yield (which was not always positive or significant) showed that soil quality can be affected in opposite directions by the type of fertilization (mineral vs. organic), or even by the use or not of fertilizers. This supports the idea that a holistic approach, including soil chemical, physical, and biological indicators, is needed to assess soil functioning in this type of agrosystems, while using yield as the only indicator of soil performance may lead to incomplete diagnosis.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/land10070727/s1. Table S1: Correlation among measured soil attributes considered for FA in the 0–15 cm depth across all management treatments.; Table S2: Correlation among measured soil attributes considered for FA in the 15–30 cm depth across all management treatments. Earthworm total biomass (EW g m<sup>−</sup>2), abundance (EW I m−2), and average size (EW g i−1).

**Author Contributions:** Conceptualization, A.S.-M., R.M.P., A.E., L.O. and I.V.; methodology, A.S.-M., R.M.P., A.E., L.O. and I.V.; Validation, A.S.-M., R.M.P. and I.V.; formal analysis, A.S.-M., R.M.P. and I.V.; investigation, A.S.-M., R.M.P. and I.V.; resources, A.S.-M., R.M.P., A.E., L.O. and I.V. data curation, A.S.-M., R.M.P. and I.V.; writing—original draft preparation, A.S.-M., R.M.P. and I.V.; writing—review and editing, A.S.-M., R.M.P., A.E. and I.V.; project administration L.O.; funding acquisition, I.V. and L.O. All authors have read and agreed to the published version of the manuscript.

**Funding:** This project has received funding from the National Institute for Agricultural and Food Research and Technology (INIA) through the RTA2017-00088-C03-01 project and from the European Union's H2020 research and innovation programme under Marie Sklodowska-Curie grant agreement No. 801586.

**Institutional Review Board Statement:** Ethical review and approval were waived for this study, as no humans or animals were used in the study.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author. The data are not publicly available due to their belonging to a multi stakeholder project.

**Acknowledgments:** We are grateful for the services and disposition of the Commonwealth of Pamplona, which contributed the experimental plot and the sludge from the Arazuri treatment plant. We also thank C. Gonzaléz from the Soil Science Laboratory of the Public University of Navarra, and A. Zaragüeta and R. Antón for their assistance in sampling and some analysis.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Review* **Soil Compaction Prevention, Amelioration and Alleviation Measures Are Effective in Mechanized and Smallholder Agriculture: A Meta-Analysis**

**Peipei Yang 1, Wenxu Dong 2, Marius Heinen 3, Wei Qin <sup>4</sup> and Oene Oenema 1,4,\***


**Abstract:** Background: The compaction of subsoils in agriculture is a threat to soil functioning. Measures aimed at the prevention, amelioration, and/or impact alleviation of compacted subsoils have been studied for more than a century, but less in smallholder agriculture. Methods: A metaanalysis was conducted to quantitatively examine the effects of the prevention, amelioration, and impact alleviation measures in mechanized and small-holder agriculture countries, using studies published during 2000~2019/2020. Results: Mean effect sizes of crop yields were large for controlled traffic (+34%) and irrigation (+51%), modest for subsoiling, deep ploughing, and residue return (+10%), and negative for no-tillage (−6%). Mean effect sizes of soil bulk density were small (<10%), suggesting bulk density is not a sensitive 'state' indicator. Mean effect sizes of penetration resistance were relatively large, with large variations. Controlled traffic had a larger effect in small-holder farming than mechanized agriculture. Conclusion: We found no fundamental differences between mechanized and smallholder agriculture in the mean effect sizes of the prevention, amelioration, and impact alleviation measures. Measures that prevent soil compaction are commonly preferred, but amelioration and alleviation are often equally needed and effective, depending on site-specific conditions. A toolbox of soil compaction prevention, amelioration, and alleviation measures is needed, for both mechanized and smallholder agriculture.

**Keywords:** compacted subsoils; crop yield; mechanized agriculture; smallholder agriculture; soil bulk density; soil penetration resistance; tillage

## **1. Introduction**

Soil compaction is defined as the 'densification of soil and the distortion of soil structure', which cause the deterioration or loss of one or more soil functions [1,2]. Compacted soils have a relatively high soil bulk density and soil strength, a low number of macro pores, and a relatively high tortuosity, and thereby a low hydraulic conductivity and water infiltration rate [3,4]. These phenomena increase the risks of temporal water logging, runoff, and erosion [5]. Compacted soils impede root elongation and development, and thereby limit soil nutrient uptake and crop development, which in turn causes yield loss [6,7]. The altered soil aeration and wetness and the decreased root growth and crop production also affect soil biodiversity and biological activity, and thereby nutrient transformations and greenhouse gas emissions [4]. Decreased aeration and increased wetness may also predispose compacted soils to infection of root rot diseases [8]. Compacted soils are widespread and have

**Citation:** Yang, P.; Dong, W.; Heinen, M.; Qin, W.; Oenema, O. Soil Compaction Prevention, Amelioration and Alleviation Measures Are Effective in Mechanized and Smallholder Agriculture: A Meta-Analysis. *Land* **2022**, *11*, 645. https://doi.org/ 10.3390/land11050645

Academic Editors: Guido Wyseure, Julián Cuevas González and Jean Poesen

Received: 6 April 2022 Accepted: 24 April 2022 Published: 27 April 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

been recognized as a global threat for modern agriculture [9,10]. Greatest concerns relate to subsoil compaction, because of the difficulty to ameliorate subsoil compaction [11,12].

Compacted soils are not easily recognized. This relates especially to compacted subsoils. There are various measures to assess subsoil compaction, e.g., [3], but there is little routine monitoring of soil compaction in practice. Yet, the concerns for soil compaction in the scientific literature is steadily increasing (Figure S1). This increased attention is especially related to the impacts of the increasing mechanization and wheel loads of machines in agriculture [13]. It was noted that a significant fraction of arable farmers in Germany are aware of the risk of intensive field traffic and high axle loads for subsoil compaction, but that this awareness had not yet led to adequate changes in practice [14]. Indeed, the impacts of human-induced (sub)soil compaction seem to increase over time [10,15,16].

Next to human induced soil compaction, through trafficking and ploughing (forming traffic and plough pans in the subsoil), soils may become compacted through natural processes, e.g., during peri-glacial conditions, or as a result of the illuviation of soil colloids, cracking and swelling processes (combined with topsoil tumbling down to the subsoil when cracks are open), heavy rains, and soil trampling by animals. Soils may have a compacted subsoil also because of an abrupt textural or mineralogical change with depth, due to a different geo-genetic origin [3]. The susceptibility of soils to compaction differs greatly. Most susceptible are soils with low soil organic matter content and a high content of silt (particles with a size of 20 to 50 μm). These soils often have a low structural stability and may be characterized as 'sealing, crusting, and hardsetting' [8,17].

Measures to ameliorate compacted subsoils and/or to alleviate their impacts have been explored almost as long as the problem has been realized [18,19]. Hence, many studies have examined the effectiveness of amelioration and alleviation measures, including deep tillage, subsoiling, reduced tillage, crop rotation, reduced trafficking, and using soil amendments. Results of these studies have been discussed and summarized in some excellent reviews. For example, Ungar and Kaspar [6] reviewed studies examining root growth in compacted soils and suggested that tillage and growing deep-rooted crops in rotations will help avoid subsoil compaction and alleviate negative impacts. Soane and Van Ouwerkerk [20] summarized the early studies related to the nature and alleviation of soil compaction. While reviewing the literature since the early 1990s, Hamza and Anderson [21] identified eight practices to avoid, delay, or prevent soil compaction, and suggested that specific combinations of measures are most effective. The review of Batey [3] largely confirmed the suggestions of Hamza and Anderson [21] and emphasized the need for the monitoring of soil compaction in practice. Nawaz et al. [4] reviewed models simulating soil compaction and the effects of soil compaction, while Chamen et al. [22] reviewed studies examining the costs and benefits of measures aimed at ameliorating soil compaction. Schneider et al. [23] quantitatively examined the effects of deep tillage on crop yield, using a meta-analysis of data mainly from Europe and North America, and observed that deep tillage effects were highly site-specific. Shaheb et al. [7] reviewed how soil compaction affected different crop types and listed twelve management strategies to alleviate soil compaction. Most studies focused on mechanized agriculture and paid little attention to smallholder agriculture. Of a different nature, Kodikara et al. [24] reviewed how soil compaction can be improved in civil engineering and transport.

Evidently, soil compaction is a complex and persistent phenomenon affecting the sustainability of crop production in modern agriculture in large areas of the world. The threat of subsoil compaction for crop production is thought to be most severe in mechanized agriculture with high axle loads on wet soils [2,12,25,26]. However, there are also reports on subsoil compaction in smallholder agriculture in China, for example, as a result of longterm soil cultivation practices, irrigation, and natural conditions [27]. It is unclear whether the effects of amelioration and alleviation measures are different between mechanized and smallholder agriculture. Machine weight is much less and ploughing depth is also less in smallholder agriculture than in mechanized agriculture. We hypothesized that amelioration and alleviation measures are more effective in smallholder agriculture than in

highly mechanized agriculture, because compacted soil layers are likely more shallow in smallholder agriculture, and thus easier to remediate.

We conducted a systematic review of the quantitative effects of measures aimed at preventing and ameliorating compacted subsoils or at alleviating the impacts of soil compaction on crop yield and soil physical properties, using a meta-analysis of published studies conducted in areas with smallholder farms (mainly China), and in mechanized agriculture in Europe, America, and Australia. We categorized measures in three groups (Table S1), largely following Hamza and Anderson [21] and Chamen et al. [22]: (i) measures aimed at avoiding and preventing subsoil compaction, including minimized and controlled trafficking, zero and minimum tillage (rotary tillage and shallow harrowing); (ii) measures aimed at remediating compacted subsoils, including subsoiling, deep ploughing, and crop rotation; and (iii) measures aimed at alleviating the effects of compacted subsoils, including residue return, controlled irrigation, and manure application. This categorization of measures also fits in the DPSIR framework1 [2].

The objectives of our study were (1) to quantitatively examine the effects of measures aimed at avoiding and ameliorating soil compaction and at alleviating the impacts of compacted subsoils on crop yield, soil bulk density, and soil penetration resistance, using results of published studies; and (2) to examine the effectiveness of measures in smallholder and mechanized agriculture. We focused on the period 2000–2019/2020, because of the existence of some excellent reviews covering the earlier period, and because studies on smallholder agriculture conducted before 2000 are relatively scarce.

#### **2. Materials and Methods**

#### *2.1. Data Collection and Screening*

We searched for peer-reviewed publications investigating the effectiveness of measures to address compacted (sub)soils, using Web of Science and China Knowledge Resource Integrated Database (CNKI, for Chinese studies not published in English language). Search terms were ("soil compaction" OR "compacted soil" OR "compacted subsoil" OR "subsoil compaction") AND ("yield" OR "biomass") AND ("density" OR "penetration" OR "soil cone index") in titles, keywords, and abstracts. In Web of Science, conference proceedings and non-English publications were excluded. This search gave 719 publications published between 2000 and 2019 (until 1 August 2019). The search in the China Knowledge Resource Integrated Database yielded 74 additional publications (from 2000 to August 2019).

The search process was followed by a screening procedure that was based on the following criteria: (1) field studies must include side by side comparisons of soil compaction prevention, remediation and/or alleviation treatments, and control (or reference) treatments; (2) for each paired comparison, treatments and reference treatments have the same location, cropping system, cropping management, and year; (3) grain yields and/or biomass yields were reported; (4) soil bulk density and/or soil penetration index data were reported; (5) the test crops were cereals, including wheat, maize, barley, oat, and sorghum; (6) location(s), year(s), and basic soil information of the experiment(s) were stated. Only studies with cereal crops as test crops were included. One reason for this is the importance of cereal crops in global food supply [28], and the other reason is that the results are likely more robust when using crops with similar root morphology and physiology [7]. Grain yield and/or biomass yield were used as crop response indicators.

Following the aforementioned screening procedure, we obtained 400 comparisons (paired observations) of crop yields from 54 studies in 28 countries from Web of Science, and 157 comparisons of crop yields from 23 studies from CNKI. Treatment measures were recorded and grouped. THe results of crop yield and soil bulk density/penetration resistance were extracted from each study, as well as characteristics related to location, experimental year(s), and soil clay content (Table S1). In cases where crop yield and/or soil bulk density and/or penetration results were presented in figures only, values were extracted using the GetData Graph Digitizer (https://apps.automeris.io/wpd/ (accessed on 1 January 2020)).

#### *2.2. Categorization of the Measures*

The paired observations were allocated to a category of measures, i.e., prevention, remediation, or alleviation measures. There is some degree of arbitrariness in the allocation of measures. For example, the choice of crop type and crop rotation was categorized as remediation measure but could have been categorized as prevention or alleviation measures equally well. Further, alleviation measures were thought to alleviate the effects of soil compaction, but may contribute also to remediation or prevention, depending on the environmental and management conditions. Thus, irrigation, fertilization, manure application, and straw return were thought to alleviate the impacts of compacted subsoils on root growth (their limited ability to take up water and nutrients from compacted subsoils).

Conventional (random) traffic was chosen as reference treatment for controlled traffic. In this case, a comparison was made between random (deliberate) trafficking and minimal or controlled trafficking, to infer the effects of controlled trafficking indirectly. Thus, random trafficking was used as reference treatment (worst-case), while minimal trafficking or controlled trafficking as the remediation treatment. The reference treatment of manure application was no manure application, while residue return was compared to no residue return. Crop rotation effects were compared to effects of mono-cropping.

Soil bulk density and soil penetration resistance results were grouped into three depth intervals: 0–20 cm (topsoil), 20–40 cm (upper subsoil), and 40–60 cm (lower subsoil). This grouping was seen as a compromise for comparing smallholder and mechanized agriculture. The depth of soil cultivation in smallholder agriculture is commonly less than 20 cm but in mechanized agriculture often a bit deeper, depending also on tillage system. Moreover, about 80% of the roots of most cereal crops are in the upper 40 cm and more than 95% of the roots are in the upper 60 cm of the soil [29,30].

Smallholder farms are mostly found in east and south Asia, Africa, and some countries of Latin America [31], and mechanized agriculture with relatively high axle loads in North America, Oceania, Europe, and west Asia. Therefore, studies conducted in south and east Asia and Africa were considered to be small-holder farming, while studies conducted in America, Europe, Australia, and west Asia were considered to be in mechanized agriculture. For more detailed information of the database composition, see Tables S1 and S2.

#### *2.3. Data Analysis*

Our meta-analysis basically followed the same approach as the one described by Qin et al. [32]. We used the natural logarithm of the ratio of the response variable of two treatments as the effect size [33]: ln(R) = ln(xt/xc), where R is the ratio, x is the response variable, and subscripts t and c refer to the specific treatment and control treatment. The response variable was either crop yield (x = Y), dry bulk density (x = BD), or penetration resistance (x = PR).

For the calculation of a grouped effect size, a linear mixed-effect model was used for which we used the R-package 'nlme' [34]. Mixed-effect models are preferred to fixedeffect models for statistical testing in ecological data synthesis because their assumption of variance heterogeneity is more likely to be satisfied [33]. In our study, results of treatments addressing soil compaction were set as fixed effects and study numbers were set as random effects, to allow accounting for variances among studies. We used the equal weighting method (e.g., [35]) when comparing studies with different number of replicates. The ln(R) of the individual pairwise comparison was used as the dependent variable. The mean effect size and the 95% confidence intervals (CIs) of each categorical group were estimated. The significance of the effects was statistically assessed at the 0.05 confidence level. In the graphs (forest plots), the effect-size of each treatment was transformed back and converted to a percentage change in crop yield, dry bulk density, or penetration resistance relative to the control or reference treatment, i.e., data were presented as (R − 1)∗100%. In case the value zero in such a forest plot falls outside the 95% CI, the given average value (effect size) is assumed to be significantly different from zero.

#### **3. Results**

#### *3.1. Overview of the Dataset*

Our dataset consisted of 557 yield comparisons, 620 soil bulk density comparisons, and 592 soil penetration resistance comparisons. About half of the number of bulk density comparisons dealt with the topsoil (346), and half with the subsoil (274). More yield comparisons were from countries with predominantly small-holder farming (S-farming) (323) than from countries with predominantly mechanized agriculture (M-agriculture) (234). More yield observations were related to prevention (221) and remediation measures (205) than alleviation measures (131, Figure 1a). Yield observations of prevention measures were found more in M-agriculture countries than in S-farming countries. The number of yield observations related to remediation and alleviation measures was two times larger with S-farming than M-agriculture (Figure 1b,c).

**Figure 1.** Relative changes in crop yield (%) in response to soil compaction prevention, remediation and alleviation measures; means of all results (**a**); means of results from countries with mechanized agriculture (**b**); means of results from countries with small-holder farming (**c**). Dots show means of treatments, error bars indicate 95% confidence intervals. Numbers in the parentheses indicate number of comparisons.

#### *3.2. Effects of Measures on Crop Yields*

Five out of ten measures examined had positive effects on crop yields, including prevention, remediation, and alleviation measures (*p* < 0.05, Figure 1a). Relatively large mean effect sizes were noted for controlled traffic (+26%) and irrigation (+51%). Mean effect sizes were also significantly positive for subsoiling, deep ploughing, residue return, and crop rotation (+8% to +11%). Minimum tillage and manure application did not display significant effects, while no tillage had a negative mean effect on crop yield (−6%).

Differences between S-farming and M-agriculture in the mean effect sizes of prevention, remediation, and alleviation measures on crop yields were relatively small (Figure 1b,c). The mean effect size of controlled traffic on crop yield was two time higher in M-agriculture (+38%) than in S-farming (+16%). However, the number of comparisons was much larger in M-agriculture (88) than in S-farming (21). Subsoiling was more studied in S-farming than in M-agriculture during the last 20 years and the mean effect on crop yield in S-farming was positive (+8%). Controlled irrigation and manure application were examined in S-farming but not in M-agriculture as possible measures to alleviate the effects of compacted subsoils. Evidently, controlled irrigation had a large effect size, but it is not realistic to ascribe this effect merely to the alleviation of soil compaction. Likely, crop yields in the reference treatments were limited by drought and not only by compacted subsoils.

#### *3.3. Effects of Measures on Soil Bulk Density*

The measures had a relatively small effect on the soil bulk density of the top soil and subsoil (Figure 2a,d), compared to their effects on crop yields (Figure 1). Relative mean changes in bulk density were in the range of 0–9%. For the subsoil, which is most critical, controlled traffic, deep ploughing, subsoiling, residue return, and crop rotation decreased soil bulk density by on average 2–9% (*p* < 0.05; Figure 2d). Controlled irrigation increased bulk density in the topsoil and subsoil, while minimum tillage increased subsoil bulk density by 3% (*p* < 0.05; Figure 2d).

Essentially all comparisons related to the effects of subsoiling and deep ploughing on subsoil bulk density originated from S-farming. As a consequence, no proper comparison can be made between S-farming and M-agriculture on the effects of subsoiling and deep ploughing. This holds for alleviation measures as well. Controlled trafficking decreased soil bulk density in both topsoil and subsoil, and S-farming and M-agriculture.

#### *3.4. Effects of Measures on Soil Penetration Resistance*

Soil penetration resistance responded to the measures in a similar way as bulk density, but the relative changes were larger (Figure 3a,d). Controlled traffic treatments had on average 33% lower penetration resistance in topsoils and 26% lower resistance in subsoils than the reference treatments. Subsoiling and deep ploughing decreased penetration resistance by 13% to 20% (*p* < 0.05, Figure 3d). No tillage increased penetration resistance in the topsoil but not in the subsoil.

Observations on subsoiling and deep ploughing originated mainly from S-farming countries, where these measures decreased penetration resistance. Residue return decreased penetration resistance in both topsoil and subsoil in S-farming. The number of comparisons for residue return was too low in M-agriculture to make firm statements. Irrigation slightly decreased penetration resistance in the topsoil but not in the subsoil in S-farming.

#### *3.5. Effects of Experimental Duration*

More than 80% of the comparisons dealt with short-term experiments (1~3 years; Table S1). Tillage treatments (deep ploughing, subsoiling, no tillage, minimum tillage) accounted for almost half (47%) of the long-term experiments (≥4 years), followed by controlled traffic (23%). For controlled traffic, the relative effect size for crop yield and for subsoil bulk density tended to increase over time (Figure 4a). For crop yield, the effect size was 33% in short-term and 37% in long-term experiments, while subsoil bulk density was 4% lower in short-term and 6% lower in long-term experiments compared to the reference treatments (*p*< 0.05; Figure 4b,c). For deep ploughing, the relative effect size for crop yield and bulk density decreased over time. In short-term (1~3 yrs) experiments, mean effect sizes were statistically significant on crop yields and bulk density (*p* < 0.05), but not in long-term (≥4 yrs) experiments. Similar results were found for no tillage (Figure 4).

#### *3.6. Effects of Soil Texture*

Soil texture (silt and clay contents) and soil organic matter content affect the susceptibility of soils to compaction and also likely influence the effect sizes of measures. A clay content of 17.5% is commonly used as a threshold value in soil compaction evaluation. Soils with <17.5% clay are considered to be more susceptible to compaction than soils with ≥17.5% clay [36]. Thus, we compared the effect sizes of measures for soils with <17.5% clay with soils having ≥17.5% clay. Yield effects were on average similar for the two textural classes (Figure 5). However, light-textured soils (<17.5% clay) showed greater responses to prevention and amelioration measures than heavy-textured soils (≥17.5% clay). This was most notable for controlled traffic. Effect sizes for yield differed by more than a factor two

(+49% vs. +19%; *p* < 0.05), for subsoiling (+12% vs. 3%), and deep ploughing (13% vs. 8%; *p* < 0.05).

**Figure 2.** Relative changes in soil bulk density (BD) in response to soil compaction prevention, remediation and alleviation measures for the topsoil (**a**–**c**) and for the subsoil (**d**–**f**); means of all results (**a**,**d**); means of results from M-agriculture (**b**,**e**); means of results from S-farming (**c**,**f**). Dots show means of treatments, error bars indicate 95% confidence intervals. Numbers in the parentheses indicate number of comparisons.

**Figure 3.** Relative changes in soil penetration resistance (PR) in response to soil compaction prevention, remediation and alleviation measures for the topsoil (**a**–**c**) and for the subsoil (**d**–**f**); means of all results (**a**,**d**); means of results from M-agriculture (**b**,**e**); means of results from S-farming (**c**,**f**). Dots show means of treatments, error bars indicate 95% confidence intervals. Numbers in the parentheses indicate number of comparisons.

**Figure 4.** Relative changes in crop yield (**a**) and soil bulk density (BD; for top soil, (**b**); and subsoil, (**c**)) in response to various soil compaction prevention, remediation and alleviation measures; means and standard deviations of results from short-term (<4 years), and long-term (≥4 years) field experiments.

**Figure 5.** Relative changes in crop yield (%) in response to soil compaction prevention, remediation and alleviation measures; means of results from clay soil (clay content ≥17.5%) (**a**); means of results from sandy soil (clay content < 17.5%) (**b**). Dots show means of treatments, error bars indicate 95% confidence intervals. Numbers in the parentheses indicate number of comparisons.

#### **4. Discussion**

#### *4.1. Understanding the Cause-Effect Relationships*

The cause–effect relationships of soil compaction and its mitigation measures can be analyzed and understood through the 'driving forces, pressures, state, impact, responses' (DPSIR) framework [2]. In agriculture, the driving forces often stem from the economic incentives to produce more and to lower costs, especially in affluent countries [11,13]. This leads to more intensive soil cultivation and the use of larger and heavier machines, which exerts literally pressure on the soil. This pressure may lead to a densification of the (sub)soil, i.e., compacted (sub)soils, with impacts on water infiltration, root and crop growth, microbiological processes, and gaseous emissions, e.g., [3]. The response of farmers and land managers may be directed towards avoiding or preventing soil compaction, i.e., addressing the driving forces and pressures, or they may focus on the amelioration of compacted soils, i.e., addressing the state, or at alleviating the impacts of compacted soils, or both (Figure 6). Thus, the three categories of measures distinguished in our meta-analysis (Table S2; Figure 1) address different aspects of the cause–effect chain of soil compaction.

**Figure 6.** The Driver-Pressure-State-Impact-Response (DPSIR) concept with focus on soil compaction. The response measures indicate which part of the DPSIR chain is being addressed by the measures.

Avoiding, preventing, and precautionary strategies are preferred above amelioration and alleviation strategies, also because of the complexities and imperfections of the latter [2,37]. However, large areas in the world have naturally compacted subsoils (e.g., [8,17]), or have been compacted by human activities in the past [15], and thus will need amelioration and alleviation strategies. Moreover, the susceptibility of soils to densification and the farming and environmental conditions greatly differ across the world, suggesting that region- and farm-specific strategies will be needed, and thus a toolbox of options and strategies. Our meta-analysis contributes to this toolbox by examining quantitatively the effects of both prevention, amelioration, and alleviation measures.

Depending on the strategy, different indicators may be used for evaluating the effectiveness of the strategy. Lebert et al. [37] discussed indicators for precautions against soil compaction (pressure indicators) and for the impairment of subsoil structure through compaction (state indicators). For the first, they proposed the 'pre-compression stress' and 'loading ratio', which can be calculated for different soils, but need soil type specific calibration [37]. For assessing the impairment of subsoil structure, they proposed three indicators, i.e., air capacity (>5% air filled porosity at a water suction of pF 1.8), saturated water conductivity (<10 cm day−1), and a visual classification of the soil morphology (combination of a 'spade diagnosis' and measurements of the effective bulk density and packing density). The second suggested indicator (saturated water conductivity) is basically an

impact indicator (and not a state indicator). Soil bulk density was not recommended as an indicator for identification of 'harmful' soil compaction, because 'there is no critical threshold and classification scheme' according to the authors [37]. However, for the related 'packing density' (bulk density corrected for clay content) indicator, there are criteria [38]. Håkansson and Lipiec [39,40] reviewed the usefulness of the relative soil bulk density, or the degree of compactness, which was defined as the dry bulk density in percent of a reference dry bulk density of the same soil obtained by a standardized, long-term, uni-axial compression test at a stress of 200 kPa. Evidently, the measurements of the state of soil compaction are labor-intensive, and thus costly, especially when considering spatial withinfield variations [41,42]. As a result, routine monitoring of the state of soil compaction in farmers' fields is not common practice. Indeed, it appears costly and there is debate about appropriate indicators and their interpretation. We observed that soil bulk density and penetration resistance are most commonly used as indicators for assessing the state of soil compaction in field experiments to test measures aimed at preventing, ameliorating, and/or alleviating soil compaction. However, bulk density is not a sensitive indicator (e.g., relative changes in soil bulk density following the implementation of measures are relatively small; Figure 2), while penetration resistance is very sensitive to variations and changes in soil moisture content. Based on uni-axial tests, Panayiotopoulos et al. [43] showed that for a compression stress up to 300 kPa the dry bulk density changed up to 5~15%. This suggests that extreme changes in dry bulk density are not likely to occur. Further, measurements of penetration resistance should be performed at pressure heads of about −100 cm. It is, however, unlikely that this was the case in all studies. This may explain why a large variability in penetration resistance was found in the reviewed studies.

Impact indicators relate to the changes in soil ecosystem functioning following a change in the densification of the soil and associated changes in pore size distributions, tortuosity, and soil structure. Possible impact indicators are crop yield, hydraulic conductivity, run-off and ponding, and emissions of CO2, CH4, and N2O [3,44]. There are no critical thresholds and classification schemes for assessing changes in soil functions, perhaps apart from hydraulic conductivity [37]. Yet, comparisons can be made between situations without and with compacted (sub)soils as in our meta-analysis. Crop yield is probably the most powerful indicator in farmers' practice, because of its influence on farm income, although part of a yield penalty may be nullified through alleviation measures, including irrigation and fertilization.

In conclusion, the DPSIR framework is useful for analyzing and understanding the cause–effect relationship of soil compaction, but further work is needed to derive a proper set of indicators and threshold values.

#### *4.2. Impacts of Measures in Small-Holder Farming and Mechanized Agriculture*

The mean effect of controlled traffic on crop yield was 38% (range 32–45%) in mechanized agriculture (M-agriculture) and 16% (range 6–27%) in small-holder farming (S-farming). The wide range of yield effects is roughly in the same range as reported by Antille et al. [16] in a review of 20 studies for various crops. The yield of crops was 0–98% higher when grown in the absence of field traffic compared to the yield of crops grown under typical traffic intensities. Controlled traffic was introduced in commercial-scale farming in the 1990s, initially in Australia and subsequently in Europe and northern America [45,46]. The net economic benefit of controlled traffic increases with farm area. Conversely, the yield effect of controlled traffic needs to be relatively large to make controlled traffic economically attractive in small farms [16,22]. It is therefore no surprise that the number of experimental studies was much larger in M-agriculture than S-farming (Figure 1b,c). Interestingly, the mean yield effect of controlled traffic was on average a factor of two smaller in S-farming than in M-agriculture, which may indeed reflect differences in axel loads between S-farming and M-agriculture.

Zero-tillage minimizes the traffic of soil-cultivating tractors and was therefore considered to be a preventive measure for soil compaction, but it does not necessarily control

the traffic of other (e.g., harvesting) machines in the field. There is a lot of interest in zero-tillage and minimum tillage (e.g., [47], as it saves labor and fuel cost, minimizes erosion (especially when combined with surface mulching), and contributes to enhanced soil carbon sequestration. However, it increases N2O emissions and decreases crop yield. The latter is in agreement with our findings (Figure 1). Further, it tends to increase the soil bulk density and penetration resistance of the topsoil (Figures 2 and 3). The no-till (or reduced-till) compacted topsoils limit root penetration and plant growth [48], while crop residues remaining on the soil surface may increase the incidence of viruses and plant pathogens [49], and lower the soil temperature [50,51]. Our study indicates that current zero-tillage and minimum tillage practices are much less effective as a preventive measure for soil compaction than controlled traffic. However, there is a need for more soil physical and soil structural measurements (including bulk density) of the subsoil in no-till systems to confirm our findings.

Deep ploughing and subsoiling increased crop yields by on average 10% and 9%, respectively, though with relatively large uncertainty bars (Figure 1a). These mean effects were derived mainly from studies conducted in S-farming and reported between 2000 and 2019/2020. Schneider et al. [23] reported rather similar mean positive effects of deep tillage on crop yield (6%), based on a meta-analysis of 45 studies (67 field experiments) that were mainly conducted in Europe and North America between 1918 and 2014 (only three studies were reported after 2000, namely one from North America, one from Argentina, and one from China). They noted that the popularity of deep tillage decreased from the 1970s. Peralta et al. [52] also found positive mean effects of subsoiling on the yield of maize (+6%) and soybean (+26%) in no-till systems in Argentina, using a meta-analysis of 32 field studies. Our study indicates that positive effects of deep tillage on crop yields also hold for smallholder farming, notably China, for both deep tillage and subsoiling. Schneider et al. [23] found that the mean effect size of deep tillage on crop yield depended on the silt content of the topsoil, the density of the subsoil, and drought, but not on the deep tillage method (subsoiling vs. deep ploughing and deep mixing) and tillage depths. The strong interference by drought agrees with our observation that irrigation alleviates the effects of compacted subsoils and greatly increases crop yield (Figure 1). The effect of deep ploughing on crop yield decreased over time (Figure 4). A similar trend was observed in the meta-analysis studies of Schneider et al. [23] and Peralta et al. [52]. The decreasing effect of deep tillage over time is likely the result of re-compaction [22,53]. Our analyses indicate that deep tillage decreased soil bulk density (Figure 2) and penetration resistance (Figure 3) of the topsoil and subsoil. Similar decreases were noted for the topsoil by Peralta et al. [52], but neither Peralta et al. [52] nor Schneider et al. [23] reported changes in soil bulk density and/or penetration resistance for the subsoil in response to deep tillage.

Alleviation measures mainly aim to lessen the negative impacts of compacted subsoils on root and crop growth. Roots elongate less in compacted and dry soils due to a combination of mechanical impedance and water stress [54], and thereby have less access to soil moisture and nutrients. Irrigation thus greatly alleviates the negative impacts of compacted subsoils on crop yield. The mean effect size of irrigation on crop yield was 50% (Figure 1). However, irrigation increased soil bulk density in the topsoil and subsoil (Figure 2). These results are based on observations in S-farming countries only, i.e., mainly China. Crop residue return or surface mulching also had a positive on crop yield, likely because of its effect on soil water preservation [32]. Crop residue return decreased soil bulk density (Figure 2), possibly as a result of enhanced soil carbon sequestration [47]. Only a few studies explicitly examined the effects of manure application on alleviating impacts of compacted subsoils on crop yield. No significant effects on crop yields were found, but manure application in S-farming tended to decrease soil bulk density, possiblY through enhancing soil organic carbon contents [55,56]. In summary, alleviation measures 'treat the symptoms but not the root cause', yet some of these measures can be highly effective, also in cases where amelioration measures were not much effective.

#### *4.3. Managing Soil Compaction*

A common opinion is that 'the best way to manage soil compaction is to prevent it from happening'. The popularity of controlled traffic and reduced or no till practices reflects this opinion. The increasing wheel loads and weight of agricultural machinery in practice in especially Europe and North America during the last 60 years do not reflect this opinion. The increase in machinery weight has resulted in an increase in subsoil compaction, which may have contributed to crop yield stagnation and to an increase in the incidence of flooding in Europe [13]. The cascade of possible impacts from soil compaction beyond field and farm scales (e.g., increased risk of flooding, runoff, and erosion) could be seen as driver for actions by policy [57,58]. However, soil compaction is not subject to a coherent set of rules in, for example, the European Union (EU), and is also not mentioned in the recent EU soil strategy for 2030 [59]. Thus, farmers depend on the insights and guidelines of their own and their advisors when it comes to handling soil compaction, while there are essentially no monitoring data concerning farmers' fields.

There is less risk of soil compaction by machines in small-holder farming in China, for example, than in the mechanized agriculture of Europe, North America, and Oceania. There is also no governmental policy aimed at preventing soil compaction in China. However, the intensive cultivation practices and irrigation, and the silty texture of the dominant loss soils in north China are conducive to soil compaction, and there is therefore a continuous search for soil conservation practices that decrease the risk of soil compaction and improve soil structure [60,61]. A combination of tillage practices in sequence appears to be the best strategy [62–64]. This holds for no-till as well. However, it has to be combined with subsoiling once in a few years, as also discussed for the no-till agriculture in Argentina by Peralta et al. [44]. The need for combining tillage practices in China also follows indirectly from the increasing interest in subsoiling during the last two decades (e.g., Figure 1 [24]).

The FAO voluntary guidelines for sustainable soil management do provide technical and policy recommendations to prevent and mitigate soil compaction [65]. Though qualitative and without threshold values, these guidelines are interesting because they address not only the machines and vehicles in the field, but also the importance of crop type and crop rotation, soil organic matter content, soil macrofauna, and microbial and fungal activities. Amelioration measures are not explicitly mentioned, apart from the recommendation to also grow crops with strong tap roots able to penetrate and break up compacted soils. Next to soil compaction, the FAO guidelines also present recommendations to prevent and mitigate nine other soil threats [65]. The need for a more coherent and integrated soil management concept was also recently emphasized by Rietra et al. [47]. They presented a roadmap for developing high-yielding, soil-improving, and environmentally sound cropping systems. This roadmap involves an iterative selection and optimization of site and farm specific crop husbandry and soil management practices, including the selection of machines that minimize soil compaction.

Evidently, preventing soil compaction from happening is too simple a strategy to address soil compaction. Rather, a toolbox of strategies and management practices is needed, which can be used to develop and implement site-specific management measures. Our study provides evidence that both prevention, amelioration, and alleviation measures have value, depending on the site-specific conditions. These measures provide net economic benefits for farms in most cases, through increases in crop yields and resource use efficiency [22,66]. The selection of the most appropriate measures will likely improve, and the effectiveness of these measures will likely increase, when more data become available at the farm level, related to the state and impact of soil compaction, through routine monitoring.

#### *4.4. Limitations of Our Study*

We focused on the recent literature (2000–2019/2020), because there are some excellent papers that reviewed and analyzed the older literature, e.g., [23,67], and not many studies have been conducted in small-holder agriculture before 2000. We examined literature from both mechanized agriculture and small-holder farming to make comparisons between these two types of agricultural systems, based on the literature from 2000–2019/20. We note that the literature from S-farming countries from before 2000 has not been analyzed in a systematic manner yet, apart from the studies by Hoogmoed et al. [68], and the reviews by Laker and Nortjé [8], and Peralta et al. [52].

Further, we note that the machine weight is rapidly increasing over time [69], not only in M-agriculture countries, but also in some S-farming countries. Hence, the rough categorization in S-farming and M-agriculture countries may not be the best way to examine differences between mechanized and smallholder agriculture, although this comparison provided new insights, e.g., related to the type of measures applied in the two types of agriculture.

Crop types may respond differently to compacted soils and thereby also to prevention, amelioration, and alleviation measures, because of differences in root morphology and physiology [54,70]. We selected cereals as test crops because these were mostly used and have a more or less uniform response. Thereby, we excluded 183 studies with non-cereal test crops out of the 719 available studies (25%).

Further, we excluded studies that combined various measures, e.g., controlled traffic combined with no tillage, controlled traffic combined with deep tillage, tillage combined with residue management levels, and irrigation combined with subsoiling. The exclusion of these studies does not mean that these studies are less relevant. Instead, it requires another study to infer useful conclusions from these combined-measures studies.

#### **5. Conclusions**

Our meta-analysis included 77 studies from 28 countries (32 studies from 16 countries for mechanized agriculture (M-agriculture), and 45 studies from 12 countries for smallholder farming (S-farming)) all related to the effectiveness of soil compaction prevention, amelioration, and alleviation measures. These studies were published between 2000 and 2019/2020 and thus are relatively recent. Prevention measures were mostly studied in Magriculture, while remediation and alleviation measures were mostly studied in S-farming.

Soil compaction prevention, through controlled traffic, had a positive effect on crop yield in both M-agriculture (+38%) and S-farming (+16%) countries, and led to a lower soil bulk density in topsoil and subsoil (−4% to −6%), and to a lower soil penetration resistance (−26% to −33%). These results confirm earlier estimates for M-agriculture countries but now show that controlled traffic also holds promise for S-farming. However, it is not clear whether controlled traffic is economically profitable in S-farming. Soil compaction prevention through no-till had negative effect on crop yield, while bulk density was increased, in both M-agriculture and S-farming.

Soil compaction amelioration through deep tillage (including subsoiling) had positive effects on crop yields (+9% to +10%), while soil bulk density was decreased by about 3%. These results confirm earlier observations for M-agriculture, but we show that these observations are also valid for S-farming. The relatively large number of studies related to deep tillage in S-farming suggest that subsoil compaction is increasingly seen as a constraint to crop production in the countries with S-farming.

Irrigation was an effective alleviation measure for subsoil compaction, though only reported for S-farming. The large mean effect size for crop yield (+51%) reflects that compacted soils impede root elongation and thereby enhance the impacts of drought, though the effect of irrigation likely relates not only to alleviation of drought related to compacted subsoils. Crop residue mulching and manure application had a small effect on alleviating compacted subsoils.

Soil penetration resistance and bulk density were mostly used as state indicators. Effect sizes of measures on soil bulk density were small (<10%), indicating that bulk density is not a sensitive indicator for assessing the effects of measures. Effect sizes of crop yield as an impact indicator were relatively large, but variable because of interfering factors (climate, soil texture).

A toolbox of soil compaction prevention, amelioration, and alleviation measures is also needed because the cause of soil compaction and the responses of measures are site-specific. Our meta-analysis indicates that such a toolbox is needed for M-agriculture and S-farming.

**Supplementary Materials:** The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/land11050645/s1, Figure S1: Changes in the number of papers published per year, studying the relationships between soil compaction and crop yield.; Table S1: Data and information used for this meta-analysis; Table S2: Summary overview of effects sizes of soil compaction prevention, remediation and alleviation measures on crop yields, soil bulk density (BD), and soil penetration resistance (PR), shown for three categories of studies.

**Author Contributions:** Conceptualization and writing, P.Y. and O.O.; data collection and data analysis, P.Y. and W.Q. Review and editing, P.Y., W.D., M.H. and O.O. All authors have read and agreed to the published version of the manuscript.

**Funding:** The present study was carried out in the EU project SoilCare ("Soil care for profitable and sustainable crop production in Europe"), EU grant agreement 677407; https://www.soilcare-project.eu (accessed on 1 February 2022), and National Key Research and Development Program of China (2021YFD190100202). First author (P.Y.) received funding from China Scholarship Council (File No. 201408130093).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** This study builds on SoilCare report 07 (https://www.soilcare-project.eu/ resources/deliverables; accessed on 1 February 2022); we thank the co-authors of that report for their contributions.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Notes**

<sup>1</sup> The DPSIR framework stands for Driving forces, Pressure, State, Impact and Responses. It allows for analyzing and understanding the cause-effect chain of soil comapction in a systematic manner, as further discussed in the Discussion section.

#### **References**


## *Review* **A Review of Crop Husbandry and Soil Management Practices Using Meta-Analysis Studies: Towards Soil-Improving Cropping Systems**

**René Rietra \*, Marius Heinen and Oene Oenema**

Wageningen Environmental Research, Wageningen University & Research, Droevendaalsesteeg 3, 6708 PB Wageningen, The Netherlands; marius.heinen@wur.nl (M.H.); oene.oenema@wur.nl (O.O.) **\*** Correspondence: rene.rietra@wur.nl

**Abstract:** Coherent improvements in crop varieties and crop husbandry and soil management practices are needed to increase global crop production in a sustainable manner. However, these practices are often discussed separately, and as a result there is little overview. Here, we present a database and synthesis of 154 meta-analysis studies related to ten main crop husbandry and soil management practices, including crop type and rotations, tillage, drainage, nutrient management, irrigation and fertigation, weed management, pest management, crop residue management, mechanization and technology, and landscape management. Most meta-analysis studies were related to tillage (55), followed by crop type and rotations (32), nutrient management (25), crop residue management (19), and irrigation and fertigation (18). Few studies were related to landscape management (6) and mechanization and technology (2). In terms of outcome, studies focused on crop yield and quality (81), soil quality (73), and environmental impacts (56), and little on economic effects (7) or resource use efficiency (24). Reported effects of alternative practices, relative to conventional practice, were positive in general. Effect sizes were relatively large for environmental effects (nutrient leaching, greenhouse gas emissions), and small for soil quality (except for soil life) and crop yield. Together, meta-analysis studies indicate that there is large scope for increasing cropland productivity and minimizing environmental impacts. A roadmap is provided for integration and optimization of all ten practices, and recommendations are formulated to address the gaps in meta-analysis studies.

**Keywords:** crop residue; crop rotation; crop yield; environmental effects; irrigation; nutrient management; resource use; soil-improving cropping systems; soil quality; tillage

## **1. Introduction**

Global yields of main crops (wheat, rice, maize, and soybean) have increased by an average 1 to 2% per year during the last decades [1,2], in response to the increasing global food and feed demands, and facilitated through technological improvements. Forecasts suggest that mean crop yields per ha of cropland have to increase by as much as 2.4% per year to be able to meet the food and feed demands by the human population in 2050, also because further expansion of global cropland area and/or increased frequency of harvesting are not feasible [3,4]. The slow-moving mean increase in global crop yields during recent decades are in part related to areas where crop yields have been stagnating and to areas where crop yields have not increased at all or have fallen. Recent analyses suggest that crop yields are not increasing on 25 to 40% of the harvested global cropland area [1]. Yield increases of wheat, maize, and rice tend to be lowest in low-income countries because of lack of resources and poor crop husbandry practices. In high-income countries, yield increases may be less than average when actual yields approach attainable crop yields, suggesting that yields reach biophysical limits [5,6]. Crop yields may also stagnate in some countries because of climate change and environmental regulations [7–9] and soil degradation [10–12].

**Citation:** Rietra, R.; Heinen, M.; Oenema, O. A Review of Crop Husbandry and Soil Management Practices Using Meta-Analysis Studies: Towards Soil-Improving Cropping Systems. *Land* **2022**, *11*, 255. https://doi.org/10.3390/land 11020255

Academic Editors: Guido Wyseure, Julián Cuevas González and Jean Poesen

Received: 13 January 2022 Accepted: 30 January 2022 Published: 8 February 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

The yield increases per unit of surface area during recent decades have mainly been the result of improved germplasm and improved crop husbandry and soil management practices, including inputs of fertilizers, irrigation, and pesticides [13]. Availability of highyielding cultivars, fertilizers, irrigation water, and pesticides are commonly considered to be the dominant yield-controlling factors, next to climate and soil quality. However, the importance of precise timing and careful execution of the various crop husbandry practices in the proper order should not be neglected [13]. The crop husbandry and soil management practices together determine how far actual crop yields deviate from attainable crop yields and from potential crop yields [14]. Attainable crop yields, defined as the best yield achieved by the best farms through skillful use of the best available technology [14], are on average 70 to 80% of the potential yield. Potential crop yields are commonly defined as the yields obtained when cultivars adapted to the local environmental conditions are grown with minimal stress, achieved with best management practices [11,15,16]. Actual yields on farmers' fields range from 30 to 100% of attainable yields, depending on region [1].

Crop husbandry and soil management practices also influence the environmental sustainability of crop production systems, especially in cases where the pressures to increase crop yields are high. Concerns have arisen about intensive crop production systems with poor crop husbandry and soil management practices, as these pollute groundwater and surface waters with nitrogen (N), phosphorus (P) and pesticides, and emit greenhouse gases and ammonia (NH3) into the atmosphere [15–17]. There are also concerns about soil degradation through processes such as erosion, salinization, compaction, and declines of soil organic matter content and soil biodiversity [10]. The United Nations (UN) Sustainable Development Goals (SDGs) address essentially all of these concerns and indirectly guide the actions of nations in the pursuit of a more sustainable world. Of the 17 SDGs, at least five have a direct relation with cropping systems and soils, while others have a more indirect relation [18]. SDG-2 aims to 'end hunger, achieve food and nutrition security, and promote sustainable agriculture' and is key to the success of the SDG agenda [19].

While there are several spatially explicit assessments of changes in crop yields over time (e.g., [2,20], there are no spatially explicit, integrated assessments of the sustainability of crop husbandry and soil management practices. The main reason for this lack of assessments is the diversity of crop husbandry and soil management practices, and the lack of methods and procedures for making such integrated assessments. Wezel et al. [21] analyzed 15 agroecological cropping practices qualitatively in terms of possible advantages and drawbacks, for temperate areas. Others have reviewed the impacts of one or a few specific crop husbandry practices (e.g., [22–24], often on the basis of a meta-analysis of published studies. There is as yet no coherent overview and comparison of the effects of all main crop husbandry and soil management practices.

The aim of this study was to provide a review of crop husbandry and soil management practices on the basis of meta-analysis studies. Meta-analysis papers commonly analyze and synthesize many experimental studies related to topical research questions and/or ambiguous research findings. The term 'meta-analysis' was first used in 1976 and referred to 'the statistical analysis of a large collection of analysis results from individual studies for the purpose of integrating the findings' [25,26]. Most meta-analysis studies related to crop husbandry and soil management practices date from the last 10 to 20 years, and the cumulative number has increased exponentially (Figure 1).

Thus, our main hypothesis is that meta-analysis studies summarize and synthesize vast amounts of research results, and unravel underlying mechanisms of variations, and thereby provide overview. By reviewing and synthesizing meta-analysis studies related to several crop husbandry and soil management practices, we aimed to (i) summarize the main impacts of these crop husbandry and soil management practices, (ii) identify the most topical research areas, and (iii) suggest guidelines for 'sustainable cropping systems'. The crop husbandry and soil management practices examined were assessed in terms of (a) crop yield and quality, (b) soil quality, (c) economic effects, (d) resource use efficiency,

(e) environmental effects, and (f) human health impacts. However, none of the reviewed studies addressed human health impacts; as a consequence, this aspect is not reported here.

**Figure 1.** Exponential increase in time of published studies used in our overview (the bar left of 2000 refers to all studies before the year 2000). In total 163 unique studies (peer-reviewed publications) were used: 154 meta-analysis studies and 9 reviews.

We reviewed meta-analyses studies related to crop husbandry and soil management practices (henceforth 'practices'). A total of ten main categories of practices were examined: (1) crop type and crop rotations including intercropping, cover crops, (2) nutrient management, (3) irrigation and fertigation, (4) controlled drainage, (5) tillage practices, (6) pest management, (7) weed management, (8) crop residue management including mulching, (9) mechanization including precision technology, and (10) landscape management, including hedgerows, tree lines, buffer strips. These ten categories of practices relate to the main crop yield defining, limiting, and reducing factors [11,15,16], and with climate-related factors (not included here) have very dominant effects on crop yield and quality, soil quality, and the environmental impacts of crop production.

For each of these categories, quantitative effects of specific practices were distilled from the meta-analysis studies. In most studies, an improved or modified practice was compared with the conventional practice. We focused on the following five outcomes (impacts): (a) crop yield and quality, (b) soil quality, (c) farm income, (d) resource use efficiency, and (e) environmental effects. We attributed the indicators that were used in the meta-analysis to these five outcomes. For crop yield and quality, farm income, and resource use efficiency, a limited number of straightforward indicators were used commonly, but for soil quality and environmental effects a wide range of indicators have been reported. We made no selection in these indicators. We focused on effect sizes defined as the standardized mean difference between the effect of a specific treatment practice relative to that of the control treatment. It is often given as the response ratio (RR) which is the ratio of the effect of a specific treatment (*X*t) and the control treatment (*X*c), with or without natural log of the ratio.

We collected data from peer-reviewed meta-analysis publications only. The publications were identified using the online database Scopus (https://www.scopus.com/sources, accessed on 5 May 2021) within the period 1997–2020. Publications were searched using the keywords "meta-analysis" and the results were refined using additional words—"crop type", "crop rotation", "nutrient management", "fertilization", "irrigation", "fertigation", "drainage", "tillage", "pest management", "disease management", "weed management", "crop residue management", "mulching", "mechanization", "landscape management", "hedgerows", and "buffer strips". These keywords were searched for in the title, abstract,

**2. Materials and Methods** *2.1. Data Collection*

and keywords. Additionally, we used the forward and backward snowballing technique when applicable and a few review studies that presented quantitative data, such as a meta-analysis, were included as well. Further information about the selection and analysis of data is provided in the Supplementary Materials.

A list of abbreviations used in this publication is given in Table A1 (Appendix A).

#### *2.2. Data Compilation and Analysis*

Data from the meta-analysis studies were compiled in Windows Excel. The region of the study, the specific practice, the conventional (control) practice, the results, the units and the number of observations were recorded. No further data processing and analyses of the data were undertaken. The Windows Excel database, with all results extracted from the meta-analyses studies, is in the Supplementary Materials.

#### **3. Results**

#### *3.1. Overview*

Table 1 presents an overview of the meta-analysis studies across categories of practices. Most of the meta-analysis studies dealt with tillage practices (55). Crop type and crop rotations (32), nutrient management (25), irrigation and fertigation (18), and crop residue management (19) have also been analyzed frequently. In contrast, only two studies related to mechanization and (precision) technology.

**Table 1.** Summary of the number of reviewed meta-analysis studies across crop husbandry and soil management practices, and across aspects (outcome). Note that the sum of the studies for the different aspects can be larger than the number of meta-analysis studies since some studies reported on several aspects.


#: one reference included the human health related aspect survival time of zoonotic pathogens; &: The total number of studies reported here consisted of 163 unique publications, some of which considered more than one crop husbandry or soil management practice.

Most meta-analysis studies examined the effects of specific practices on crop yield and quality (81). For soil quality (73 studies), soil organic matter content was the main focus. For environmental effects (56 studies), the focus was mainly on greenhouse gas emissions and nitrate leaching. Resource use efficiency was examined mainly for irrigation and fertigation, nutrient management and tillage. Only seven meta-analysis studies included economic aspects (Table 1).

The meta-analysis studies reviewed covered a large number of experimental studies and practices in different parts of the world. Each meta-analysis study was based on a large number of underlying studies (on average more than 100; range 8 to 678). In order to attain an impression of how often literature sources have been used in multiple metaanalysis studies, we collected and examined the literature sources of the 55 meta analyses on tillage. In most cases, references to the original studies were provided in the supporting

information, but for seven out of the 55 meta-analyses studies no references were made available. For the remaining 48 studies we collected in total 5465 references to original studies. These were then manually checked on replicate use. Over two-third of these references were used only once, 26% of these were used in two meta-analysis studies, 4% were used three times, and 2% were used four times. Three references were used in seven meta-analyses studies. We conclude that essentially all meta-analyses related to tillage were based on unique studies, which replicated use of original studies is relatively small (given the large number of meta-analyses related to tillage), and that the results of these meta-analyses are largely independent on each other therefore. We did not check repeated use of original studies for other categories of practices.

#### *3.2. Crop Type and Crop Rotation*

Selecting the proper crop varieties and crop rotations is often farm and region-specific and key to successful crop farming. Crop rotation is the practice of planting different crops sequentially on the same field, mainly to combat pest and weed pressures and improve soil quality, and thereby to enhance crop yield sustainably. Crop rotations have been the subject of many meta-analysis (Table 1), whereby almost equal attention has been given to crop yield, soil quality, and environmental effects, but little attention to the economic aspects and to resource use efficiency. Specific crop varieties and cultivars have not been the subject of meta-analysis.

Effects of crop rotation, intercropping, and cover crops on crop yield, soil quality, and the environment were positive in almost all studies (Figure 2). Pre-crops before wheat [27] and especially legumes as pre-crops [28] had positive effects on wheat yield, soil quality, and pesticide use [29]. However, effects of pre-crops depend on the nitrogen fertilization rate: yield benefits are highest under low nitrogen fertilization [28]. Indeed, interactions with other crop husbandry and soil management hold for many crop rotation effects; nutrient management, irrigation, pest, disease, and weed management all have a large impact on the effect size of crop rotations [30–32].

The simultaneous cultivation of two or more crop species within one field for at least a part of the growing period (intercropping) also has positive effects on crop yield, but the effect size strongly depends on the crop types and intercropping patterns [30,33]. Growing cover or catch crops after the main crop reduces soil erosion and nitrate leaching and contributes to soil carbon sequestration [34,35], but requires labor and the suppressive effects on pest, diseases and weeds are not always positive. Growing mixtures of varieties of cereals [36] or mixtures of grasses [37] has positive effects on yield (stability) and nitrogen use efficiency.

Effects of crop rotations on GHG emission are variable [38]; this holds also for the effects on cover crops on GHG emissions [39].

#### *3.3. Nutrient Management*

The 25 meta-analysis studies related to nutrient management have paid more or less equal attention to crop yield and quality, soil quality, and environmental effects, but little or no attention to economic effects and resource use efficiency (Table 1). Almost all studies reported significant positive effect sizes of the studied nutrient management practices relative to conventional practices (Figure 3).

A main focus has been the characterization of differences between fertilizer types, especially between organic and mineral fertilizers [65,66,78,80] and between 'conventional' fertilizers and fertilizers with inhibitors [71,74], in relation to fertilizer effectiveness, soil quality, and environmental impacts. Deriving the optimal nutrient application rates have been the topic of many experimental studies in the past, and this has also been the subject of several meta-analysis studies [64,68,70]. Better timing of fertilization and placement of fertilizers gave positive effects on yields in most cases [75,84]. Soil liming increased crop yields, especially when pH was low [67]. Positive effects of organic soil amendments and mineral fertilizers on soil biological activity and microbial biomass were found, while

the response of soil enzyme activity depended on enzyme type [68,70,80,81]. Nitrous oxide emissions from cropland increase with nitrogen fertilization, but the increase can be mitigated through better compliance with fertilizer recommendations, and the use of nitrification inhibitors and biochar [72]. Slurry acidification, deep placement, and urease inhibitors decreases ammonia emissions from slurries and urea fertilizers applied to soil [76,86]. No meta-analysis studies related to the effectiveness of manure products from different manure processing techniques [87]. Increasing grazing intensity of pastures increased C, N, and P losses from these pastures ([83] as well as the transfer of zoonotic pathogens to water courses [77]). Only few studies pointed at the effects of interactions between categories of practices, including interactions between intercropping, tillage, and fertilizers types in fruit yield [63], interactions between fertilization, and irrigation in fruit yields [70,88] and in maize yields [70].

**Figure 2.** Overall effect sizes (response ratios) reported in meta-analysis studies on cropping (split in crop types and crop rotation, cover crops, intercropping, and perennial crops) grouped per area of interest (light green: agronomic; light blue: soil quality; light orange: resource use efficiency; light brown: economic (no data); light yellow: environmental impacts). The management or treatment comparison is indicated outside the y-axis and the variable to which the data refer are listed inside the y-axis (abbreviations can be found in Table A1). Green bars indicate improvement, red bars indicate worsening. See also Table S1. [a]: [24]; [b]: [28]; [c]: [27]; [d]: [40]; [e]: [41]; [f]: [42]; [g]: [38]; [h]: [43]; [i]: [44]; [j]: [30]; [k]: [45]; [l]: [46]; [m]: [47]; [n]: [48]; [o]: [49]; [p]: [32]; [q]: [31]; [r]: [50]; [s]: [51]; [t]: [52]; [u]: [33]; [v]: [53]; [w]: [54]; [x]: [39]; [y]: [55]; [z]: [56]; [aa]: [57]; [ab]: [58]; [ac]: [59]; [ad]: [60]; [ae]: [61]; [af]: [62].

**Figure 3.** *Cont*.

**Figure 3.** Overall response ratios reported in meta-analysis studies on nutrient management grouped per area of interest (light green: agronomic; light blue: soil quality; light orange: resource use efficiency (no data); light brown: economic (no data); light yellow: environmental impacts). The management or treatment comparison is indicated outside the y-axis and the variable to which the data refer are listed inside the y-axis (abbreviations can be found in Table A1). Green bars indicate improvement, red bars indicate worsening. See also Table S2. [a]: [63]; [b]: [64]; [c]: [65]; [d]: [66]; [e]: [67]; [f]: [68]; [g]: [69]; [i]: [70]; [j]: [71]; [k]: [72]; [l]: [39]; [m]: [73]; [n]: [74]; [o]: [75]; [p]: [76]; [q]: [77]; [r]: [78]; [s]: [79]; [t]: [80]; [u]: [81]; [v]: [82]; [w]: [83]; [x]: [84]; [y]: [85]; [z]: [86].

#### *3.4. Irrigation and Fertigation*

A total of 18 meta-analysis studies related to irrigation and/or fertigation, mainly examining the effects of irrigation methods and amounts on crop yield and water use efficiency for different cropping systems and regions (Table 1). The relative strong focus on water use efficiency reflects that irrigation water is a scarce resource. Most studies reported positive effect sizes of irrigation practices on crop yield and water use efficiency relative to conventional irrigation practices (Figure 4). Effect sizes of water productivity of optimal irrigation and deficit irrigation ranged from 20 to 80%. However, some studies also reported negative effects of deficit irrigation practices relative to conventional irrigation practices, possibly because irrigation was reduced too much in deficit irrigation treatments.

**Figure 4.** Overall response ratios reported in meta-analysis studies on irrigation and fertigation grouped per area of interest (light green: agronomic; light blue: soil quality; light orange: resource use efficiency; light brown: economic (no data); light yellow: environmental impacts). The management or treatment comparison is indicated outside the y-axis and the variable to which the data refer are listed inside the y-axis (abbreviations can be found in Table A1). Green bars indicate improvement, red bars indicate worsening. See also Table S3. [a]: [94]; [b]: [90]; [c]: [95]; [d]: [96]; [e]: [91]; [f]: [97]; [g]: [98]; [h]: [99]; [i]: [89]; [j]: [100]; [k]: [101]; [l]: [88]; [m]: [102]; [n]: [103]; [o]: [104]; [p]: [92]; [q]: [93]; [r]: [70].

Two meta-analysis studies examined the effects of irrigation method on emissions of N2O [89,90], while one study examined flood irrigation practices on emissions of methane from paddy rice [91]. Two studies examined effect of irrigation on soil respiration and soil carbon contents [92,93]. Only two studies pointed at large effects of interactions between irrigation and fertilization in water use efficiency and nutrient use efficiency [70,88].

#### *3.5. Controlled Drainage*

Effects of controlled drainage on the loss of water, nutrients, and greenhouse gases were assessed through six meta-analysis studies (Table 1). Controlled drainage is defined as the use of adjustable head structures to prevent discharge when the water table is lower than the outlet elevation. In this way the loss of water and nutrients may be altered, depending on the target. The quantitative effects of controlled drainage on reducing drainage volumes, N losses, and methane emissions were relatively large (range 17 to 85%) (Table 2). Generally, controlled drainage resulted in reduced drainage volumes, depending on soil type [105]. Controlled drainage also reduced N-losses via drainage water to surface water [106–108] and methane (CH4) emissions from peat lands [109]. No impact on yield was found by [108]. Alternating wetting and drying cycles in paddy rice greatly decreased CH4 emissions, but increased N2O emission; yet total greenhouse gas emissions decreased through improved water management [110].

**Table 2.** Controlled drainage: effect sizes as reported in meta-analysis studies. See also Table S4.


[a]: [109]; [b]: [105]; [c]: [106]; [d]: [107]; [e]: [108]; [f]: [110].

#### *3.6. Tillage*

Tillage refers to the preparation of the soil for growing crops, with or without incorporation of crop residues in the soil and/or weed control. In conventional or traditional tillage (TT), the topsoil (usually the upper 15 to 25 cm) is turned and/or milled. Conservation tillage (including no-tillage (NT) or reduced tillage (RT)) is the practice of minimizing soil disturbance, whereby crop residues commonly remain on the soil surface to protect the soil, while herbicides or precision mechanical weeding tools are used to control weeds. Tillage practices are debated because of high fossil energy and labor costs, and their effects on soil erosion, crop yield, soil organic carbon, and soil biodiversity. This debate is reflected in the high number (55) of meta-analysis studies (Table 1). The focus of most meta-analysis studies has been on soil quality (36), followed by crop yield effects (19) and environmental effects (14). Two studies synthesized economic implications of different tillage practices (Table 1). Of the total number of studies, 20 were global studies, 14 studies related to (parts of) China, 5 to the Mediterranean, 3 to US, 1 to South Asia, 1 to Brazil, 1 to Europe and none to Africa.

Overall, conservation tillage decreased crop yields, increased soil organic carbon contents in the topsoil, increased soil biodiversity and the abundance of soil organisms, and increased N2O emissions relative to conventional tillage, but the magnitude of the differences depended on climate and the particular study (Figure 5). Yield penalties of no-till depended on crop residue return and crop rotation and were larger in tropical than temperate regions, and tended to decrease with an increase in the duration of no-till [22,111]. The South-Asian study was probably the most integrated one, as it examined effect sizes of crop yield, water use, soil organic C sequestration, emissions of CO2, CH4, and N2O, and economic costs [112]. The cost of production was significantly lower under no-till than under conventional tillage in all the selected crops, and the net economic returns increased by 5 to 32%. Manley et al. [113] examined the economic cost of soil carbon sequestration in the US through no-till. They found that the additional carbon sequestration of no-till compared to conventional till was small and variable, and as a result, the net economic benefit also varied widely.

**Figure 5.** *Cont*.

**Figure 5.** *Cont*.

**Figure 5.** Overall response ratios reported in meta-analysis studies on tillage grouped per area of interest (light green: agronomic; light blue: soil quality; light orange: resource use efficiency; light brown: economic (no data); light yellow: environmental impacts). The management or treatment comparison is indicated outside the y-axis and the variable to which the data refer are listed inside the y-axis (abbreviations can be found in Table A1). Green bars indicate improvement, red bars indicate worsening. See also Table S5. [a]: [112]; [b]: [61]; [c]: [123]; [d]: [123]; [e]: [124]; [f]: [125]; [g]: [124]; [h]: [120]; [i]: [115]; [j]: [118]; [k]: [119]; [l]: [126]; [m]: [127]; [n]: [63]; [o]: [128]; [p]: [129]; [q]: [130]; [r]: [131]; [s]: [132]; [t]: [133]; [u]: [49]; [v]: [114]; [w]: [134]; [x]: [53]; [y]: [135]; [z]: [122]; [aa]: [136]; [ab]: [137]; [ac]: [138]; [ad]: [139]; [ae]: [140]; [af]: [117]; [ag]: [74]; [ah]: [141]; [ai]: [142]; [aj]: [113]; [ak]: [143]; [al]: [116]; [am]: [111]; [an]: [22]; [ao]: [121]; [ap]: [144]; [aq]: [24]; [ar]: [145]; [as]: [146]; [at]: [147]; [au]: [148]; [av]: [149]; [aw]: [150]; [ax]: [151]; [ay]: [152]; [az]: [153].

Several studies found positive effect sizes of no-till versus conventional till for number of earthworms and for the diversity of the (micro) biological community (e.g., [114–116]). No-till tended to increase the bulk density in the lower part of the topsoil (10–20 cm) and the water infiltration rate significantly [117,118]. However, no-till combined with occasional conventional tillage decreased soil bulk density compared to conventional tillage [119]. Effects of no-till on erosion are strongly affected by crop type and soil surface mulching; on average no-till and conservation agriculture reduced erosion [120,121], but pesticides in runoff tended to increase [122].

#### *3.7. Pest Management*

Pest management refers to the control of the number of undesirable organisms (pathogens, pest organisms) below an acceptable threshold, which is often based on economic principles. Methods of control can be crop rotation, chemical, biological, physical/mechanical, and/or genetic. There are often interactions with crop residue management, tillage, nutrient management, irrigation, and landscape management [154]. We found seven meta-analysis studies related to pest management (Table 1), of which three were in the context of comparing organic versus conventional agriculture (Table 3). Muneret et al. [155] found that organic farming experiences higher levels of pest infestation, but is able to match or outperform conventional pest control practices against plant pathogens and animal pests. Lesur-Dumoulin et al. [156] found that yields in organic horticulture were on average 10 to 32% lower than yields in conventional horticulture (Table 3). Garratt et al. [157] observed that organic farming practices can increase natural enemy numbers and also pest responses. Fertilization tends to increase insects and fungal plant pathogens [158,159]. Biofumigation through incorporating Brassicaceae plants and crop residues, which release glucosinolates and isothiocyanates, in soil reduced pest abundance and subsequently increased crop yield by 30% [160]. Anaerobic soil disinfestation, through temporal soil sealing following incorporation of labile organic carbon in the soil, is also effective against soil borne pathogens [161]. Furthermore, it has been indicated that addition of organic amendments and improving soil quality and biodiversity may result in fewer pests [162].

**Table 3.** Pest management: main effects as reported in meta-analysis studies. See also Table S6.


[a]: [160]; [b]: [161]; [c] [156]; [d]: [159]; [e]: [158]; [f]: [157].

#### *3.8. Weed Management*

Weed management refers to the control of the number of weed plants (especially noxious weeds) to below an acceptable threshold, as weeds compete with the crop for light, water and nutrients. Weed management often includes a number of methods, including crop rotation/intercropping/cover crops, soil cultivation (weeding, hoeing), mulching (crop residues or plastic covers), herbicides spraying, and burning. We found four metaanalysis studies related to weed control (Table 2).

Verret et al. [163] found that intercropping with legume companion plants enhanced weed control, generally without reducing the yield of the main crop (Table 4). Cover crops can also decrease the incidence of weeds and may have other ecosystem services [164]. Crop rotation with different planting dates and crop diversification, combined with limited soil disturbance, can disrupt weed-crop associations in addition to reducing yield loss and rebuilding soil fertility [165–167]. Glyphosate is the most used chemical weed control

**Parameter Comparison of Treatments Result** Weed biomass Legume intercropping vs. conventional, both non-weeded and weeded −56%, −42% [a] Weed density, biomass Parasitic nematodes Cover crops vs. traditional tillage <sup>−</sup>10%, <sup>−</sup>5%, +29% [b] Number of studies with increase soil organic matter Reduced tillage vs. traditional tillage +40 and <sup>−</sup>7 out of 78 studies [c] Soil microbial respiration Soil microbial biomass Glyphosate vs. no use, <10 mg kg Glyphosate vs. no use, >10 mg kg logarithm of ratio: 0.064 ± 0.126, 0.04 ± 0.09 [d]

agent [168], but is debated because of its effects on soil biodiversity and soil microbial respiration [169] and human health [168].

**Table 4.** Weed management: main effects as reported in meta-analysis studies. See also Table S7.

#### [a]: [163]; [b]: [53]; [c]: [170]; [d]: [169].

#### *3.9. Crop Residue Management*

Crop residues may be left on the soil surface, incorporated in the soil, burned or removed from the field for use as livestock feed or biofuel. Evidently, there are trade-offs in managing crop residues [171]. Conservation agriculture promotes the return of the crop residues to the soil to increase soil quality and reduce soil erosion, often in combination with zero-tillage or reduced tillage (Section 3.6). In this review, we distinguished crop residue management as a separate management practice, because of the relatively large number (19) of meta-analysis studies related to just crop residues (Table 1). Crop yield, water and nitrogen use efficiency, emissions of N2O, and soil carbon sequestration were the main topics of these studies.

In most cases, crop residue management and mulching increased crop yields, and water and nitrogen use efficiencies by 0 to 50% (Figure 6). Mulching greatly reduced soil evaporation and thereby provided a greater fraction of soil water to the crop, which boosted crop yields. Crop residue return has a positive effect on soil carbon sequestration and soil microbial activity, but N2O emissions increased as well. Nine out of the 19 meta-analysis studies dealt with soil mulching effects in China, as it is a common practice in dry-land farming in China (and India). One study examined the performance of biodegradable plastics to determine the optimal type of mulching for maize, wheat, potato, and cotton [172], and another [173] compared the performance of biodegradable films relative to polyethylene films.

#### *3.10. Mechanization*

Mechanization has greatly increased labor productivity in modern crop production systems, especially during the last century, and thereby has greatly contributed to farmscale enlargement and withdrawal of labor from agriculture [189]. However, mechanization has also contributed to increased fossil fuel use and increased soil compaction [190]. During the last decades, research emphasis has shifted to precision technology, controlled traffic, and robotization. However, only two meta-analysis studies have touched mechanization, precision technology and robotization (Table 1). Ampoorter et al. [191] concluded on the basis of an analysis of 11 studies with 35 forest stands that mechanical harvesting of trees has led to the compaction of the top 30 cm of forest soils, with the largest effects on the top 10 cm. One study was of a different nature: It examined the change in the ratio of maize grain yield to labor input following the introduction of specific sustainable intensification practices technologies in sub-Saharan countries [192]. No firm conclusions could be derived because of lack of sufficient empirical studies.

**Figure 6.** Overall response ratios reported in meta-analysis studies on crop residues and mulching grouped per area of interest (light green: agronomic; light blue: soil quality; light orange: resource use efficiency; light brown: economic; light yellow: environmental impacts). The management or treatment comparison is indicated outside the y-axis and the variable to which the data refer are listed inside the y-axis (abbreviations can be found in Table A1). Green bars indicate improvement, red bars indicate worsening. See also Table S8. [a]: [174]; [b] [175]; [c]: [176]; [d]: [177]; [e]: [178]; [f]: [22]; [g]: [179]; [h]: [172]; [i]: [180]; [j]: [181]; [k]: [182]; [l]: [183]; [m]: [97]; [n]: [184]; [o]: [185]; [p]: [173]; [q]: [186]; [r]: [187]; [s]: [188].

#### *3.11. Landscape Management*

Landscape management is a relatively new concept and has increased in importance following the approval of the UN Sustainable Development Goals and the recognition that the landscape is often the best scale for managing interactions, synergies, and tradeoffs for natural resource management [193–195]. Landscape management in the context of sustainable food production may include hydrological measures, terracing, hedgerows, tree lines, wind breaks, flower strips, corridors, and agroforestry, depending on the landscape, environmental conditions, and stakeholders.

We identified six meta-analysis studies related to landscape management practices (Tables 1 and 5). Three of these quantified the benefits of windbreaks on crop yields [196–198]. Three studies analyzed the effects of hedgerows and flower strips on pollination, pest control, and crop yield [197–199], and two studies analyzed the effects of hedgerows on runoff and erosion [198,200]. Both large positive effects and negative effects have been reported (Table 5).

**Table 5.** Landscape management: summary of the results as reported by meta-analysis studies. See also Table S9.


[a]: [196]; [b]: [198]; [c]: [197]; [d]: [199]; [e]: [200]; [f]: [201].

#### **4. Discussion**

#### *4.1. Main Findings*

Most meta-analysis studies reported positive effects of alternative/improved practices relative to conventional practices. The 32 studies related to crop type and crop rotation clearly indicated the positive effects of crop rotations versus continuous cropping, legumes in crop rotations versus no legumes in crop rotations, intercropping versus monocultures, and cover cropping versus no cover cropping on crop yield and soil quality. Positive effects of especially cover crops and perennial crops on erosion control and minimizing nitrate leaching were also found, depending on, e.g., N fertilization.

The 25 studies related to nutrient management examined a diversity of nutrient sources, and application methods, timing, and strategies. Most studies reported positive effects of alternative/modified practices on crop yield and on minimizing environmental pollution, relative to conventional practices. Impacts of nutrient management strongly depended on environmental conditions.

The 18 studies related to irrigation and fertigation focused on the method, timing and volume of irrigation. Drip irrigation, deficit irrigation, and subsoil irrigation were all effective in increasing water use efficiency compared to sprinkling irrigation and especially flood irrigation. No economic assessments were made, and long-term impacts on soil quality and environmental pollution were also not reported.

Six studies related to drainage, with a focus on controlled drainage in response to variable rainfall patterns. Results indicate that controlled drainage increased farm income when compared to no human-induced drainage.

A total of 55 meta-analysis studies were devoted to tillage practices. Reduced tillage tended to reduce crop yields, but increased farm income (one study only), water use efficiency, soil carbon contents, and emissions of nitrous oxide (N2O), which is a potent greenhouse gas. Reduced tillage in combination with crop residue return (mulching) and crop rotation had a slight positive effect on crop yield compared to conventional tillage.

Most of the seven studies related to pest management compared organic farming and conventional farming management practices. In general, organic farming management practices greatly decreased the use of pesticides, but lowered crop yields as well, depending on crop type and rotation, N application rate, soil quality, and (soil) biodiversity.

The four studies related to weed management did not provide a coherent view. Legume intercropping, cover cropping, and reduced tillage had positive effects on soil carbon contents but the effects on weed and crop yield were not clear.

The 19 studies related to crop residue management and mulching in dryland and/or irrigated conditions reported in general positive effects of mulching on crop yield and water use efficiency, but also increases in N2O emissions, which are unwanted.

The six studies related to landscape management reported positive effects of windbreaks and hedgerows on crop yields and erosion control, but depending on site specific conditions, and provided that the surface area of windbreaks and hedgerows is in balance with the cropping area.

Evidently, most of the studies reported positive effects of the examined alternative/improved practices, relative to the common practice, on either crop yield, soil quality, resource use efficiency, and the environment (decreased emissions). While global assessment studies often paint rather pessimistic views on the state of food production, agriculture, and the environment [10,16,202–204], it is clear that the 174 studies reviewed here present a picture of optimism and hope. Indeed, there is large body of scientific/empirical evidence that some specific practices are more effective than others, i.e., have positive effect sizes relative to conventional practices (Figures 1–6; Tables 2–5), and that these positive effects may contribute to the sustainability of crop and food production. However, large steps still have to be made to integrate, optimize, and transfer the scientific findings of meta-analysis in current practice. We note that only few meta-analysis studies examined interactions between categories of practices, while essentially no meta-analysis study made in-depth comparisons at cropping system level in which all ten categories of crop husbandry and soil management practices had been optimized. Hence, there is need for further integration and optimization of all ten crop husbandry and soil management practices, and show the effectiveness of optimized practices through experimental studies and ultimately meta-analysis studies. There is also a need to transfer the positive messages of meta-analysis studies to practice through demonstration, extension services and possibly economic incentives. Cropping systems with all crop husbandry and soil management practices optimized may be termed 'soil-improving cropping systems', to emphasize the two-way interaction between soil and crop (see Section 4.3).

#### *4.2. Uneven Coverage of Meta-Analysis Studies*

Some crop husbandry and soil management practices have been studied extensively and repeatedly, while some other practices have received little research attention (Table 1). Further, most studies have examined the effects of practices on crop yield, soil quality and environmental effects, while farm income (cost-benefit ratios) and resource use efficiency have received less attention (and human health aspects not at all). Evidently, the coverage of meta-analysis studies across practices and outcomes has been uneven; 75% of all studies addressed four practices, in the order: soil tillage > crop type and crop rotations > nutrient management > irrigation/fertigation (Table 1).

The large interest in soil tillage (55 meta-analyses studies) is certainly related to the importance of soil conservation, and the envisaged reduction in soil erosion, net greenhouse gas emissions, energy use, and labor through minimum or zero tillage. The effect-size of tillage practices were relatively small (0–10%) for crop yield, modest (0–50%) for greenhouse gas emissions and nutrient leaching, and relatively large and positive for soil quality, especially for soil life (0–150%).

The relatively large attention for nutrient management and irrigation/fertigation is related to the role of nutrients and water in boosting crop yields across the world (e.g., [205], to the depletion of fresh water resources [206] and rock phosphorus resources [207,208], and to the ecological impacts of excess nitrogen and phosphorus in the environment [16,209]). Nutrient and irrigation water inputs often form a relatively large economic cost to farmers, especially in developing countries, but this aspect has not been addressed.

We found only two meta-analyses related to mechanization and technology in agriculture (including forestry). However, several recent textbooks on precision technology for cropping systems do address the possible economic and environmental impacts of technological applications for sensing, field operations, and data handling, analysis, and control (e.g., [210–212]. Indeed, mechanization has revolutionized crop production systems during the past century but differently in different regions of the world. It has made large-scale crop production systems possible, has led to an exodus of laborers, has contributed to international trade of food and feed, and has indirectly affected essentially all crop husbandry and soil management practices. Robotization goes a step further and may revolutionize crop production systems again in the near future; it also offers the opportunity the reduce the impact of heavy machines on soil compaction. Keller et al. [190] estimated that the increase in weight of agricultural vehicles has caused an increase in soil bulk density, and thereby decreased root growth, crop yields, and soil hydraulic properties. They speculate that heavy machinery has contributed to yield stagnation and increased flooding in Europe [190].

We recommend that future meta-analysis studies related to crop husbandry and soil management practices should pay more attention to the socio-economic impacts of practices including possible barriers and constraints for their implementation in practice. Next, we recommend that more emphasis has to be given to interactions between multiple crop husbandry and soil management practices, and to comparisons of region-specific optimized packages of these practices. Further, Africa should not be neglected, as much of the increased food demand (and food production) during the next few decades will occur in Africa.

#### *4.3. Towards High-Yielding, Soil-Improving, and Environmentally Sound Cropping Systems*

The effect of specific crop yield defining, yield limiting, or yield reducing factors is largest when all other crop yield defining, limiting, or reducing factors are optimal, i.e., at a level where these do not affect crop yield [213]. This 'law of the optimum' may have also influenced the outcomes of meta-analyses studies; optimality of all factors will have enhanced the effect size of an alternative practice relative to the control practice, and vice versa. We have no insight in the degree of optimality of yield factors in the studies underlying the reviewed meta-analyses, but simply note here that there is often a gap between actual and attainable yields, and between actual and attainable environmental performances in practice. These gaps have to be narrowed to be able to produce adequate amounts of food in a sustainable and region-specific way [214].

The reviewed meta-analysis studies provide many suggestions for improved practices, but the optimization of all practices has to be done for specific regions, at farm level and/or regional levels. The possible steps in the optimization process have been summarized in Figure 7; it provides a roadmap for developing high-yielding, soil-improving and environmental-sound cropping systems. Steps 1 and 2 deal with the analyses and description of the current cropping systems, including its socio-economic and environmental environments. Steps 3 to 12 then deal with the selection and optimization of the 10 main specific crop husbandry and soil management practices, while taking the results of steps 1 and 2 into account. The actual process of optimization will be iterative, until the most optimal combination of practices has been identified. Variants of this road-map have been tested within the EU-funded project SoilCare, and results are presented in this special issue.

**Figure 7.** Towards sustainable cropping systems; a step-wise roadmap for developing high-yielding, soil-improving and environmentally sound cropping systems. The steps (1 to 12) have to be taken in a consecutive-iterative manner so as to find the optimal combination of practices.

#### *4.4. Concluding Remarks*

Crop husbandry and soil management practices are of critical importance for closing yield gaps, raising farm income and soil quality, and minimizing the environmental impacts of cropping systems in the world. We identified ten categories of crop husbandry and soil management practices, based on the concept of crop yield defining, limiting and reducing factors, and tried to quantify the effects of improved or modified practices relative to conventional practices, by using results of meta-analysis studies.

Our review was based on the premise that meta-analysis papers and reviews synthesize large numbers of experimental studies related to topical research questions and important research findings. For example, closing yield gaps and decreasing environmental impacts are topical, and thus we expected that in the course of the last 20 years when meta-analysis studies blossomed, a wealth of synthesized information would become accessible to help improve crop husbandry and soil management practices and thereby increase crop yield and soil quality, and decrease the environmental impact of crop production. The meta-analysis studies reviewed covered a huge number of experimental studies and practices in different parts of the world, albeit uneven. The number of studies per category

of practices seem to reflect topics of hot societal debates and/or studies with controversial research findings. The number of meta-analysis studies per category of practices seem not to reflect those topics and practices that have largest impacts on crop yields, soil quality, and the environment.

Most meta-analysis studies reported positive effects of specific practices relative to conventional practices, on crop yield, soil quality and the environment. However, most meta-analysis studies examined single practices, with limited emphasis on interactions between categories of practices, and on the optimization across practices. Further, the coverage of studies was uneven, both in terms of practices, sustainability aspects and world regions. Notably, economic aspects were rarely addressed.

Based on this review, we derived a roadmap with twelve steps for integrating and optimization of all main crop husbandry and soil management practices, so as to develop high-yielding, soil-improving, and environmentally-sound cropping systems. We call these 'soil-improving cropping systems' to emphasize that cropping systems must maintain and improve soil quality to remain sustainable. This roadmap has been tested in practice and some results are presented in other papers of this special issue. We also made a number of recommendations.

**Supplementary Materials:** The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/land11020255/s1, Table S1: Cropping: effects on (a) crop yield and quality, (b) soil quality, (c) economic effects, (d) resource use efficiency, (e) environmental effects, and (f) human health impacts as reported in meta analysis studies; aoi = area of interest. Table S2: Nutrient management: effects on (a) crop yield and quality, (b) soil quality, (c) economic effects, (d) resource use efficiency, (e) environmental effects, and (f) human health impacts as reported in meta analysis studies; aoi = area of interest. Table S3: Irrigation and fertigation: effects on (a) crop yield and quality, (b) soil quality, (c) economic effects, (d) resource use efficiency, (e) environmental effects, and (f) human health impacts as reported in meta analysis studies; aoi = area of interest. DI = deficit irrigation, PRD = partial rootzone drying, FI = full irrigation, AI = aerated irrigation, NAI non aerated irrigation, RDI = regulated deficit irrigation, CDI = conventional deficit irrigation, CI = conventional irrigation, OI = over irrigation, UI = under irrigation, OPTI = optimal irrigation. Table S4: Controlled drainage: effects on (a) crop yield and quality, (b) soil quality, (c) economic effects, (d) resource use efficiency, (e) environmental effects, and (f) human health impacts as reported in meta analysis studies; aoi = area of interest (see Table 1). Table S5: Soil tillage: effects on (a) crop yield and quality, (b) soil quality, (c) economic effects, (d) resource use efficiency, (e) environmental effects, and (f) human health impacts as reported in meta analysis studies; aoi = area of interest. NT = no tillage, TT = traditional tillage, CA = conservation agriculture, RT = reduced tillage, MT = minimum tillage. Table S6: Pest management: effects on (a) crop yield and quality, (b) soil quality, (c) economic effects, (d) resource use efficiency, (e) environmental effects, and (f) human health impacts as reported in meta analysis studies; aoi = area of interest. Table S7: Weed management: effects on (a) crop yield and quality, (b) soil quality, (c) economic effects, (d) resource use efficiency, (e) environmental effects, and (f) human health impacts as reported in meta analysis studies; aoi = area of interest. Table S8: Crop residue management & mulching: effects on (a) crop yield and quality, (b) soil quality, (c) economic effects, (d) resource use efficiency, (e) environmental effects, and (f) human health impacts as reported in meta analysis studies; aoi = area of interest. Table S9: Landscape management: effects on (a) crop yield and quality, (b) soil quality, (c) economic effects, (d) resource use efficiency, (e) environmental effects, and (f) human health impacts as reported in meta analysis studies; aoi = area of interest. Table S10: Explanation of the main columns in the accompanying Excel sheet.

**Author Contributions:** Conceptualization and writing by all authors; data collection and data analysis by R.R. + M.H. (contributed equally to this study). Review and editing were done by all authors. All authors have read and agreed to the published version of the manuscript.

**Funding:** The present study was carried out in the EU project SoilCare ("*Soil care for profitable and sustainable crop production in Europe*"), EU grant agreement 677407; https://www.soilcare-project.eu (accessed on 12 January 2022).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** All tabulated data are available in the Supplemental Information (Word document with tables; Excel file with extended data).

**Acknowledgments:** This study builds on SoilCare report 07 (https://www.soilcare-project.eu/ resources/deliverables; accessed on 12 January 2022); we thank the co-authors of that report for their contributions.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Appendix A**

**Table A1.** List of abbreviations.


#: which is the conservation tillage-induced change in N2O emission compared to conventional tillage when N fertilizer is applied; \$: CH4 and N2O emissions per unit yield.

#### **References**


MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Land* Editorial Office E-mail: land@mdpi.com www.mdpi.com/journal/land

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel: +41 61 683 77 34

www.mdpi.com ISBN 978-3-0365-6254-4