1. Introduction
The semiarid rainfed Mediterranean areas in the European Union (EU) are facing a double challenge, namely to diminish the use of fertilizers by 20% and to halve nutrient loss [
1,
2]. These objectives are added to the already regulated N fertilization practices in nitrate-vulnerable areas [
3], which have been designated across EU land (e.g., [
4]).
This double challenge requires different strategies according to the framework of each region. In rainfed semiarid Mediterranean areas (<400 mm yr
−1), low water availability is a critical factor for winter crops, and it significantly influences their productivity [
5]. In fact, in the rainfed systems of Northeastern Spain (Catalonia), the traditional rotation is three years for barley and one year for wheat or fallow. Crop diversification has a constraint of low soil moisture conditions, although rapeseed and leguminous crops are being introduced [
6]. The average yields from barley (
Hordeum vulgare L.) and rapeseed (
Brassica napus L.) in the 2016–2020 period were 3.5 and 2.3 Mg ha
−1, respectively [
7]. This region also has an important rearing activity, mainly porcine (
Sus scrofa domesticus), with slurry production equivalent to 54.8 million kilograms of N [
8]. Therefore, slurry management is one of the most relevant environmental issues for the 1.1 million ha of arable land [
9]. In addition, sewage sludge is also available for use. In 2017, 93% of the 108.6 million kilos (over dry matter) of sewage sludge produced, which equated to 4.6 million kilograms of N, was destined for soil application. Legislative tools exist to prevent heavy metal accumulation when using sewage sludge as fertilizer [
10]. They include specific limits for concentrations of Cd, Cu, Ni, Pb, Zn, Hg, Cr (VI), and inorganic As in fertilizers. Also, they establish limits in the soil concentration for the first six mentioned above. In addition, the synthetic nitrogenous fertilizers consumed in the region rose to 26.9 million kilograms during the 2022–2023 season [
11].
Achieving the EU’s fertilizer-reduction targets requires a much better matching of crop demand and an increased nutrient-use efficiency [
12]. One of the aspects is the application period or splitting fertilization in different development stages of the plant, i.e., at sowing and later on as topdressing [
13]. Improving yields with organic fertilizers also depends on the type of organic fertilizer, crop rotation, soil type, method of application, and previous treatments [
14].
The use of organic fertilizers is also seen as a useful tool to increase organic matter content in soils [
15] in the context of climate change. In Spain, this is an important issue, as Spanish agricultural land holds the lowest European average of organic C soil content [
16]. Apart from its total concentration, it is a matter of interest to grasp the fate of organic C in soil, which can be elucidated through its physical fractionation and further chemical analysis [
17]. Furthermore, organic fertilizers can favor micronutrient availability [
18].
The hypothesis of this work is that, in rainfed semiarid zones where water availability is a constraint because of low rainfall with high annual variability, N applications should be reduced in relation to the general maximum N applications allowed in nitrate-vulnerable areas (170 and 120 kg N ha−1 yr−1 for fertilizers of organic and mineral origin, respectively) or in nitrate-non-vulnerable (semiarid) areas from Catalonia in NE Spain (190 kg N ha−1 yr−1 from fertilizers of organic origin). This reduction will not reduce general chemical nutrient fertility but rather avoid excessive nutrient build-up or potential nutrient losses out of the agricultural system while maintaining organic C content through the use of organic fertilizers. Furthermore, this work will also address new production scenarios that are linked to climate change, where drought periods are expected to increase.
The goal of this work is to evaluate different fertilization strategies in terms of crop yields and soil chemical characteristics to determine which fertilization strategy best matches the EU policy goals on reducing fertilizers (mainly N inputs) and potential nutrient losses in semiarid Mediterranean regions without affecting soil quality. Soil fertility aspects (mineral-N, organic matter, nutrients, and available and total heavy metals) will be assessed.
2. Materials and Methods
2.1. Location
The experimental field was located in NE Spain (41°46′ N, 01°05′ E, 346 m asl).
It was established in the 1997–1998 cropping season. The results of this paper include 13 cropping seasons, starting from the 2003–2004 one.
The soil is classified as a Typical Xerorthent [
19]. The soil is very deep (>1 m), without stones or coarse elements. The soil texture in the superficial layer (0.3 m) is loamy (USDA classification). The clay, silt, and sand contents are 157 g kg
−1, 463 g kg
−1, and 380 g kg
−1, respectively. The pH equals 8.1 (1:2.5, soil–water); it is not saline (electrical conductivity 1:5 soil:water equals 0.33 dS m
−1 at 25 °C), having a high content of equivalent calcium carbonate (280 g kg
−1). The cation exchange capacity is 6.7 cmol
+ kg
−1. In 1997, at sowing at 0–0.3 m depth from a composite soil sample, the organic matter, available phosphorus, and available potassium contents were 19 g kg
−1, 36 mg P kg
−1, and 303 mg K kg
−1.
The climate is dry Mediterranean according to the Papadakis classification [
20]. The maximum precipitation occurs in spring and autumn. The average annual rainfall of the studied period (thirteen years) was 387 mm. The recorded yearly variability oscillated between 201 and 621 mm, and the monthly values ranged from 8 to 100 mm. The annual average temperature was 13.8 °C, with an oscillation between 12.2 and 14.7 °C. The reference crop’s evapotranspiration averaged 1091 mm yr
−1 (Penman–Monteith equation; [
21]). In this research, the spring of the cropping seasons was considered dry if the accumulated rainfall between February and March was less than 50 mm. Those months coincide with the end of the tillering period and the beginning of stem elongation in winter cereals. In rapeseed, it coincides with leaf development, with nine or more leaves unfolded.
2.2. Experimental Design
The present study was conducted in the context of an initial demonstration field on the use of organic fertilizers, established in the 1997–1998 cropping season. In the 2003–2004 cropping season, the field was fully included in a research program. The different strategies, based on the combination of fertilization at sowing (first factor) and topdressing (second factor), were maintained throughout the whole experimental period. They were distributed according to a split-plot design with three replicates.
The first factor was the N fertilization applied before sowing (Fsow) with different fertilizers. It included five treatments: 0, 30, 141, 176, and 351 kg N ha−1. They were applied as mineral fertilizer (MIN), slurry from fattening pigs (PS), and the last ones as composted sewage sludge (L1 and L2), respectively. These five treatments were randomized against the block.
The second factor was N fertilization applied at the topdressing (Ftop) with mineral fertilizer. It included three N fertilization treatments: 0, 50, and 100 kg N ha−1. These treatments were randomized against the first factor within each block.
The above-mentioned averages of N applied with PS, L1, and L2 were calculated for the 2003−2016 period, and the standard deviation of applied rates were ±31, ±67, and ±136 kg N ha
−1, respectively. The rate variability was linked to the changing concentrations of organic materials (
Table 1), although they always satisfied the legislation requirements [
10].
Mineral N was applied as ammonium nitrate or calcium ammonium nitrate, while organic amendments were locally available. Composted sewage sludge was obtained by mixing one part of municipal wastewater sludge with three parts of agro-industrial and forest waste, always complying with agricultural use regulations [
10].
Each experimental plot measured 7 m wide and 23 m long.
In slurries and composted sewage sludges, dry matter was determined by a gravimetric method at 105 °C, organic carbon by ignition at 550 °C, organic nitrogen by the Kjeldahl method, and ammonium nitrogen by distillation and titration following APHA methods 4500-NH3B-C and 4500-NH3C [
22]. The total phosphorus and total potassium were analyzed by acid digestion (wet) and further determined using inductively coupled plasma atomic emission spectrometry [
23].
Organic fertilizers were applied with machinery, while mineral fertilizers were applied by hand. At sowing, the N mineral fertilizer was urea. After application before sowing, fertilizers in all plots were mechanically buried (0–0.15 m) within 24 h after application. In winter (from late January to early March), plots received an N topdressing enhancement in the form of a mineral fertilizer (calcium ammonium nitrate, ammonium nitrosulfate, or ammonium nitrate based on availability), applied at a V6–V8 Zadoks cereal physiological stage for barley and wheat [
24] and the leaf-development stage for rapeseed. The topdressing fertilizers were not buried. The mineral treatment was annually supplied with phosphorus (26.4 kg P ha
−1) and potassium (74.7 kg K ha
−1) as was CO in 6 out of the 13 cropping seasons. The annual average rates of P and K applied with PS were 38 kg P ha
−1 and 76 kg K ha
−1. For L1, the rates were 73 kg P ha
−1 and 62 kg K ha
−1, a rate which was doubled for L2.
The present work includes the yields from thirteen cropping seasons, starting at the 2003–2004 cropping season when the field was fully included in a research program. It also includes the comparison in NO3−-N between September 2003 and 2016. The rest of the analyzed parameters were evaluated at the end of the experimental period (July 2016).
Barley was sown in the autumns of 2003, 2006, 2007, 2008, 2010, 2011, 2012, 2014, and 2015. Bread wheat (Triticum aestivum L.) was sown in the autumns of 2004, 2005, and 2009. Finally, rapeseed was sown in September 2013. Management field practices followed the technical agricultural recommendations for the area. Harvesting was carried out between late June and mid-July. Straw was removed from the fields, except for the 2014 harvest, due to the low amount of straw production.
The total surface of each plot (161 m2) was harvested. A grain sample was taken to determine the humidity by drying at 60 °C. The yield data were adjusted to 1.2 g kg−1 of humidity for barley and wheat and to 0.8 g kg−1 for rapeseed.
At the start of this experimental period (September 2003) and at the end of it (September 2016), the plots were sampled from the surface to a 0.9 m depth using an Edelman auger (7 cm in diameter). The soil was fresh-sieved to pass through a 2 mm sieve, and 20 g were extracted with 60 mL of a solution of potassium chloride (1N) for the colorimetric determination of NO3−-N concentrations with a continuous flow analyzer (AA3, Bran + Luebbe, Norderstedt, Germany).
In the last cropping season (2015–2016), an additional sampling was previously performed in July, after harvest, at the depth (0–0.15 m) where Fsow fertilizers were annually buried. Plots from the Fsow treatments, without mineral-N topdressing fertilization (Fsow0), were sampled. The pH, electrical conductivity (EC), available P (Olsen-P), cation exchange capacity and exchangeable cations, heavy metals, and organic carbon fractionation were analyzed. In each plot, a composite sample was obtained from three sampling points. The samples were prepared for analysis according to UNE-EN 16179 [
25]. The soil pH was determined in an aqueous solution using a 1:2.5 (soil–water) ratio, salinity (EC) by conductimetry (1:5), and oxidizable organic carbon, as by [
26]. The organic carbon fractionation followed the procedure NF X 31-516 [
27]. Cation exchange capacity and exchangeable cations were evaluated by extraction with ammonium acetate 1N (pH = 7). The exchangeable cations K
+, Na
+, Mg
2+, and Ca
2+ were determined by following [
28], followed by further determination by atomic absorption spectrophotometry. The available P content was quantified by the Olsen method (sodium bicarbonate-extractable P at pH 8.5 [
29]. The available Mn, Fe, Zn, Cu, Ni, and Cd were extracted with a DTPA (diethylenetriaminepentaacetic acid) solution (1:2, w:v) following Baker and Amacher [
30]. From microwave soil digested samples (UNE-EN 54321; [
31]) with aqua regia (3:1, v:v, HCl:HNO
3) the elements P, K, Ca, Mg, Na, Fe, Cd, Co, Cr, Cu, Mn, Ni, Pb, and Zn were quantified using inductively coupled plasma mass spectrometry (UNE-EN 16171; [
32]).
2.3. Statistical Analysis
The statistical package SAS, version 9.4 [
33], was used for the statistical analysis.
The yield analysis considered the split-plot design in three randomized blocks. A mixed model procedure was performed for analyses of the wheat and rape yields. The block was considered a random factor. For wheat production, we accounted for repeated measurements over 3 years by using a compound symmetry covariance structure. In barley, a Glimmix model procedure was performed with a gamma distribution and inverse link function to satisfy the assumptions. The covariance structures for repeated measurements were modeled by using a compound symmetry covariance structure. The models’ performances were evaluated using the Akaike Information Criterion [
34]. Multiple comparisons of least square means of the main effects and interactions were conducted using the LSMEANS option. The value of 5% (i.e.,
p < 0.05) was selected as the minimum criterion for significance.
Furthermore, in barley, the correlation coefficient was used to assess the strength of the relation between barley yields and precipitation in March and April. The 2007 harvest was excluded because of previous problems in plant establishment. The 2016 harvest was also excluded because of the severe hailstorm close to the harvest period, which damaged all the barley plants.
The analysis of NO3−-N content (kg N ha−1) was performed twice, (i) once for the upper layer (0–0.3 m) and (ii) once for the soil profile (0–0.9 m). The analysis was performed using the mixed model procedure from SAS. The Satterthwaite approximation was chosen. The adopted fertilization treatment for analysis was the combination (Fcomb) of fertilization at sowing with fertilization at the topdressing. The Fcomb treatment and time (sampling year) interaction was included in the model. The Fcomb treatments referenced for each block were called subjects. In the statistical analysis, random intercepts for experimental blocks nested within subjects were included, thereby accounting for the hierarchical structure of the experimental design and the potential correlation between repeated measures within subjects. Multiple comparisons were made with the LSMEANS option. The value of 5% (i.e., p < 0.05) was selected as the minimum criterion for significance.
The rest of the parameters were analyzed for the fertilization treatments at sowing, without fertilization at the topdressing (Fsow0), and according to a randomized block design. The GLM model procedure for general linear models was used. The means were compared according to Duncan’s multiple range test (DMRT) (p = 0.05).
4. Discussion
The longest research period was devoted to barley (nine cropping seasons). Assuming this crop as a reference in this agricultural system, barley’s average yields were lower than the recorded average in rainfed Spanish areas [
7]. The exceptions were the 2009 and 2013 harvests (
Figure 2) with humid springs (>70 mm rainfall,
Figure 1) when the yields matched the results from other authors [
5,
13] for similar cropping seasonal rainfall.
Low water availability at critical stages for the development of winter cereals [
35,
36], like the stem-elongation period (February–March in this experiment), affected barley (2.49 Mg ha
−1, 5-year average) and wheat (1.97 Mg ha
−1, 2-year average) yields (
Figure 1,
Figure 2 and
Figure 4). When carbon assimilation during stem elongation is reduced by stress, the storage in stems is reduced. It implies a limitation on the quantity of reserves that might be remobilized during the grain-filling period [
35]. This lack of water also constrained the rapeseed yield in the 2014 harvest (maximum of 2.44 Mg ha
−1,
Figure 5), exacerbated by rapeseed’s lower water use efficiency in comparison to barley [
37]. However, rapeseed was able to take advantage of annual organic fertilization (PS, L1) and N topdressing, probably because rapeseed requires more N fertilizer input and has a lower utilization efficiency during its growth period when compared to other crops [
38].
In dry years, the lack of water availability constraints yields, but the absence of N leaching and further organic matter mineralization in these systems [
39] allows the control treatment to achieve the highest yields in the following rainier cropping season, as evidenced in the 2007 and 2013 harvests (
Figure 1 and
Figure 2). In fact, the absence of winter leaching in dryland systems [
40], the reduction of N losses through ammonia volatilization due to the rapid slurry soil incorporation after application [
41], and the prevention of large concentrations of N
2 and N
2O within the soil (in contrast to other agricultural systems [
42]) is the framework that explains full N availability in the following cropping season from organic matter mineralization or former N fertilizer applications. The amount of N atmospheric deposition is very low [
43,
44,
45]. Nonetheless, it must be considered that, given the projected increase of dry periods due to the climate-change scenario [
46,
47], residual N will be of greater importance in the N balance of these systems.
The no-N application strategy might be, a priori, an advantageous short-term strategy for barley and wheat, since it did not present significant yield differences versus other studied strategies in dry years while also being the most economical option and the one with the lowest potential environmental impact related to soil NO
3−-N content. However, it would be inadequate to consider the CO treatment as a long-term fertilization strategy, since “nutrient mining” is unsustainable [
48,
49] if the mineralized organic matter is not replaced. In fact, in 2016, the amount of organic C found in the CO treatment (14.4 g kg
−1) was lower than its initial value of 19 g kg
−1 in 2003, and it was significantly below (from 3 up to 12 g kg
−1) the organic C content in other treatments (
Table 2). This result challenges the international “4 per 1000” initiative that aims to increase carbon storage in agricultural soils in hopes of mitigating the effects of climate change [
50]. Therefore, other annual treatments appear as sounder options in order to achieve maximum yields while sustaining organic C, as is the case with the PS treatment or the L1 treatment, which even increases the content of organic C (
Table 2). However, it must be noted that, in pig slurry, the organic C is mainly sustained by the light fraction from 0.05 to 0.2 mm (
Table 3), which will then be mineralized and requires continuous maintenance. In contrast, L1 also affects organic C content in the heavy fraction (0.2–2 mm), potentially offering longer term organic C protection. In fact, the spectroscopic characteristics of soil humic-type substances after pig slurry applications enhance aliphatic structures and not aromatic ones, which implies a weak effect of PS in long-term C sequestration [
51].
Levels of NO
3−-N in the soil profile (
Figure 7) support annual application maintenance in PS and L1, as the soil’s NO
3−-N did not significantly increase over the 13-year period, provided that mineral topdressing fertilization will be avoided. However, the significant NO
3−-N increase in the top layer (
Figure 6) alerts about the potential N-leaching risk linked to exceptional rainfall or snow events that are, in fact, a definite risk in these Mediterranean environments [
39].
Considering both productivity and nitrate-leaching control, and taking into account the yields obtained in the CO treatment after a former drought season, it would be justified to avoid PS and L1 fertilization after such periods of water scarcity, on the understanding that C sequestration will be limited.
In wheat, no yield response to fertilization was recorded in the dry years (2005 and 2006 harvests;
Figure 1 and
Figure 4). In contrast, L1 achieved the highest yield at the end of a rainy cropping season (2010 harvest). In a humid year, a topdressing of 50 kg N ha
−1 of mineral-N fertilizer can also be feasible. Thus, wheat mirrored barley’s behavior in terms of the profitability of residual N. In fact, barley and bread wheat respond similarly with regard to grain nitrogen content and water availability changes in the typical Mediterranean environmental conditions [
52].
In rapeseed, the seed yield fluctuated from 1257 kg ha
−1 (in the CO, without N at topdressing) to 2649 kg ha
−1 (in PS plus 50 kg N ha
−1 at the topdressing), demonstrating the link between yield and N availability, in agreement with [
5]. However, in their study, they were able to reach 4000 kg ha
−1 when the water availability increased. In this crop, the application of PS and L1 organic amendments proved to be effective as fertilization strategies (
Figure 5). However, mineral N at the topdressing, from 50 kg N ha
−1 to 100 kg N ha
−1, also increased the yields (
Figure 5), which is in concordance with [
53]. In addition, the N supply from soil organic matter mineralization, which in these conditions can reach 100 kg N ha
−1 within the first 0.6 m depth [
39,
54], is also an important N source. However, after the highly productive barley season of 2012–2013, rapeseed probably took advantage of its deep root development [
55], with acceptable nutrient absorption during its initial phenological stages [
56].
If PS and L1 are adopted as fertilization strategies (necessarily adjusted to the previous season’s climate conditions), their impact on other fertility and soil quality parameters must be considered. Compared to mineral fertilization, pig slurry will increase K, Mg, Cu, and Zn availability (
Figure 8 and
Figure 9), and composted sewage sludge (L1) will increase the soil cation exchange capacity and Fe and Cd availability (in this case below the risk thresholds). However, an important constraint for PS and L1 annual use arises from the increase of available P (Olsen). In fact, 86 mg P kg
−1 (
Table 2) is a threshold P soil concentration because of the risk of its displacement through the profile of calcareous soils [
57], despite the tendency in total P increase (
Table 4). No other limitations of PS and L1 use are detected to be related to the total soil concentrations of nutrients or heavy metals (
Table 4), which is a fact that mainly disagrees with [
58], probably because of the different origins and treatments of the compost used in both cases.