**Effects of Different Land Uses (Abandoned Farmland, Intensive Agriculture and Forest) on Soil Hydrological Properties in Southern Spain**

**Manuel Esteban Lucas-Borja 1, Demetrio Antonio Zema 2,\*, Pedro Antonio Plaza-Álvarez 1, Vesna Zupanc 3, Jantiene Baartman 4, Javier Sagra 1, Javier González-Romero 1, Daniel Moya <sup>1</sup> and Jorge de las Heras <sup>1</sup>**


Received: 2 February 2019; Accepted: 6 March 2019; Published: 11 March 2019

**Abstract:** A detailed knowledge of soil water repellency (SWR) and water infiltration capacity of soils under different land uses is of fundamental importance in Mediterranean areas, since these areas are prone to soil degradation risks (e.g., erosion, runoff of polluting compounds) as a response to different hydrological processes. The present study evaluates the effects of land uses on SWR and soil hydraulic conductivity (SHC) by direct measurements at the plot scale in three areas representing (1) intensive agricultural use, (2) abandoned farmland, and (3) a forest ecosystem in Southern Spain under Mediterranean climatic conditions. The physico-chemical properties and water content of the experimental soils were also measured. Significant SWR and SHC differences were found among the analyzed land uses. Forest soils showed high SWR and low SHC, while the reverse effects (that is, low SWR and high SHC) were detected in soils subjected to intensive agriculture. Organic matter and bulk density were important soil properties influencing SWR and SHC. The study, demonstrating how land uses can have important effects on the hydrological characteristics of soils, give land managers insights into the choice of the most suitable land use planning strategies in view of facing the high runoff and erosion rates typical of the Mediterranean areas.

**Keywords:** soil water repellency; soil hydrological conductivity; soil physico-chemical properties; vegetal cover; vegetation cover

#### **1. Introduction**

Mediterranean areas are very prone to soil degradation risks (e.g., surface runoff, erosion, transport of nutrients and other polluting compounds): the soils are generally shallow with low levels of organic matter, low aggregate stability, and nutrient content [1], and the climate is characterized by frequent and intense rainstorms producing a high magnitude of flash floods with high erosive power [2]. This combination of soil type and climate leads to a peculiar hydrological response with high runoff rates that have high erosive power. This response is also affected by both land use and soil cover and their temporal and spatial variability, which are considered the most important factors

affecting the intensity and frequency of surface runoff and soil erosion [3,4]. Inappropriate land use or a poor soil cover may accelerate water runoff and soil erosion dynamics [4,5], leading to unsustainable land degradation processes. The main causes of such negative environmental impacts are agricultural practices, deforestation, overgrazing, land abandonment, wildfires, and civil works [6,7]. For instance, agriculture is thought to generate high erosion rates [4,8,9], and the use of land for intensive agriculture, in general, may cause soil damage [10,11]. As well known, agriculture, particularly when intensive and subject to frequent tillage operations, may strongly affect the physico-chemical properties of soil, making it more prone to erosion and quality decay.

Erosion problems have also been associated with land abandonment of both agricultural and marginal farmlands [1,12] with particular reference to the Mediterranean of Western Europe [9,13,14]. For instance, in SE Spain, where drastic human impacts were recorded in the second half of the past century [15], anthropogenic land use changes have triggered soil erosion and led to severe land degradation [1]. Here, since the mid-twentieth century, the use of land for intensive agriculture and the gradual process of vegetation recovery of abandoned farmlands and marginal areas have been the main causes of land degradation [10,11]. These land use changes, whose environmental impacts are worsened by the specific climate and the soil fragility, make this region very prone to runoff generation and soil erosion [3] with consequent pollution of water bodies, soil organic matter decline, devastating floods, reservoir siltation, and mass failures [1,15].

In Mediterranean areas, the hydrological processes generating water runoff, soil erosion, and transport of polluting compounds are dominated by the infiltration-excess mechanism [16]. In soils of the semi-arid Mediterranean climate, exhibiting low hydraulic conductivity, surface runoff, and soil erosion, can be high. Moreover, such soils could be expected to also be affected by water repellency [17,18], further decreasing infiltration rates [19], which in turn leads to increased runoff and erosion [20,21], to accelerated leaching of agrochemicals [22], and to a reduction in the vegetal cover of soils, leaving the latter bare and thus prone to erosion [23,24]. Thus, depending on the water repellency level, Mediterranean soils may have an infiltration rate of up to several orders of magnitude lower than would be expected (e.g., [20,21,25,26]). It is thus evident how a detailed knowledge of the soil water repellency (hereinafter SWR) and water infiltration capacity (in terms of soil hydraulic conductivity, hereinafter SHC) under different land uses is of fundamental importance in Mediterranean areas to control hydrological risks and other environmental impacts linked to these risks.

SWR has been studied worldwide [20,21], in both forest [27] and agricultural soils [24,28]. The agricultural soils are usually considered wettable [29] and thus a little subject to SWR, while other studies have demonstrated that some management practices may induce SWR in cultivated soils [30–32]. The attention paid to surface runoff and soil erosion rates in marginal and abandoned lands (e.g., [33]) has highlighted sometimes contrasting results in the Mediterranean [34] and, consequently, the difficulty to fully understand the effects of land abandonment on hydrological response and soil erosion [9], as modified by both SWR and SHC. Therefore, it is important to know about repellency and infiltration under typical land uses in the drier parts of the Mediterranean basin [18]. A better comprehension of these fundamental soil parameters is important for both agricultural production and protection of abandoned farmland from hydro-geological risks [24].

This study evaluates the effects of land use on SWR and SHC by direct measurements at the plot scale in three areas representing (1) intensive agricultural use, (2) abandoned farmland, and (3) a forest ecosystem in Southern Spain. In addition, the physico-chemical properties and water content, which, as it is well known, can influence SWR and SHC, were also measured on these representative Mediterranean soils. The objective of the study is to evaluate which of the analyzed land uses (agriculture, abandoned farmland and forest) shows the highest SWR and lowest SHC in the experimental conditions, also linking these properties to important soil characteristics. By demonstrating how land uses can have important effects on the hydrological characteristics of soils, we want to give land managers insights on the choice of the most suitable land use planning in view of facing the high runoff and erosion rates typical of the Mediterranean areas.

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

#### *2.1. Study Area*

The study area is located in the municipality of Villamalea (39.36422◦ N, −1.59689◦ E, Albacete, SE Spain) (Figure 1). The climate is typically Mediterranean, "*Csa*", according to the Köppen-Geiger classification [35]. The average annual rainfall and temperature are 407 mm and 14 ◦C, respectively (Figure 2). The rainfall is mainly distributed in spring and autumn, with a long drought in summer that usually lasts from June–September.

Elevation of the studied plots ranges between 760 and 770 m a.s.l., a flat terrain, typical for the Iberian Plateau (Meseta). The plateau is characterized by high elevated (>600 m a.s.l.) region with an undulating (hilly) landscape that is prone to high erosion rates due to the dry climate conditions that reduce the vegetation cover.

The study area is a traditional Mediterranean forest and agricultural land, where pine plantations, natural forest, almonds, olive, cereals, and vineyards are widespread (Figure 1). The natural tree vegetation of the study area mainly consists of *Pinus pinea* L. and *Pinus halepensis* M. The main shrubs and herbaceous species found at the study site are *Rosmarinus officinalis* L., *Brachypodium retusum* (Pers.) Beauv., *Lavandula latifolia* Medik., *Thymus vulgaris* L., *Stipa tenacissima* (L.), *Quercus coccifera* L. and *Plantago albicans* L. Agriculture consists mainly of vineyards, olive orchards and maize crops, with vine production providing the main agricultural income in the study region.

**Figure 1.** Location of the study area (**A**, Villamalea, Castilla La Mancha, SE Spain) with the typical land uses (**B**: Forest; **C**: Intensive agriculture; **D**: Abandoned farmland).

**Figure 2.** Annual precipitation (total P, mm) as well as mean maximum, average and minimum air temperatures (Tmax, Tmed and Tmin) in the study area (Villamalea, Castilla La Mancha, SE Spain).

According to FAO-UNESCO [36] and the IUSS Working Group WRB [37], the soils of the study area are classified as *Calcic cambisols*. Organic matter content is on average 1.1 g/m3, varying from 1.05–1.12 g/m3. The texture is sandy clay loam (60% sand, 10% silt, and 27% clay) over a limestone parent material.

#### *2.2. Experimental Design*

#### 2.2.1. Plot Description

In autumn 2018, nine 10 × 10 m<sup>2</sup> plots were set up in each of the three land uses (27 plots in total) of the same study area with the typical climate and soil characteristics described in Section 2.1. The analyzed land uses were (1) intensive agriculture; (2) abandoned farmland; (3) natural forest. The plots with similar slopes (1–5%) were distributed selecting certain site characteristics, slopes, and aspects to ensure comparability among the 27 plots. Distances between plots were always over 500 m.

The land use related to intensive agriculture (hereinafter "IA") consisted of an olive grove about 20 years old. Cropping operations follow the usual standards of the local farmers. The orchard is tilled three times per year, scheduled depending on the soil moisture and weed removal needs. Tillage is carried out just before or on the occasion of weed germination in order to keep the soil surface clean and tidy and is refined by eliminating herbs near the trunk of each tree (in June or July) by hoes. No fertilizers are used in the olive grove. Residues of pruning, operated in February or March, are concentrated and burned at the margins of the grove, but not chipped, as usually made for other crops (e.g., citrus and apricots).

The abandoned farmland (hereinafter "AF") was an olive orchard, abandoned about 15 years ago and now mainly covered by herbs and shrubs (mainly *Rosmarinus officinalis* L., *Brachypodium retusum* (Pers.) Beauv., *Lavandula latifolia* Medik., *Thymus vulgaris* L., *Macrochloa tenacissima* L., *Quercus coccifera* L., *Plantago albicans* L. *Eryngium campestre* L., and *Pistacia lentiscus* L.).

Forest plots (hereinafter "FO") are covered by *Pinus halepenis* M., about 50 years old, with a density of 560 trees per ha. Shrubs found underneath the trees are *Quercus coccifera* L. and *Macrochloa tenacissima* L.

#### 2.2.2. Soil Property Measurements

Herbal cover, rock fragments, dead woody matter, and bare soil covers were measured at three 5-m transects in each plot, measuring the percent cover of each property over a grid of 1 m × 1 m. Bulk density was calculated on triple samples per plot as the weight of soil in a given volume of the core extracted by a small cylinder at a depth of 20–30 cm. Soil water content (SWC) was estimated using a HOBOnet Soil Moisture Sensor, which integrates the field-proven ECH2O™ EC5 Sensor and provides readings directly in volumetric water content. SWC was measured hourly for 24 h on the occasion of the SWR and SHC surveys and the measures were then averaged.

The main topsoil properties (0–5 cm) were determined on six samples per plot. A total of 162 soil samples (6 samples × 9 plots × 3 land uses) were collected. Soil characterization was carried out by measuring the following parameters: (i) texture (sand, silt and clay percentage) following the methodology of Guitian and Carballas (1976) [38]; (ii) soil organic matter (OM), estimated from organic carbon following the methodology proposed by Nelson and Sommers (1996) [39]; (iii) electrical conductivity (EC) and pH, measured in deionized water (1:2.5 and 1:5 w:w, correspondingly) at 20 ◦C [40]; (iv) content of (total nitrogen, using the Kjeldahl method [41]) and available phosphorus [42]); (v) content of potassium; magnesium; sodium; and calcium, exchangeable cations measured using the barium-chloridetriethanolamine method [43]).

Based on the parameters above, the carbon-to-nitrogen ratio (C/N) and cation exchange capacity (CEC) were calculated.

The water infiltration capacity of soils was estimated measuring SHC by a MiniDisk infiltrometer (MDI). In more detail, first the measured cumulative infiltration values (I, [m]) were fitted against the measurement intervals (t, [s]), both given by MDI, using Equation (1):

$$\mathbf{I} = \mathbf{C}\_1 \mathbf{t} + \mathbf{C}\_2 \sqrt{\mathbf{t}} \tag{1}$$

and the coefficients C1 [m/s] and C2 [m/s1/2] were estimated by interpolation. Coefficient C1 is related to SHC, and C2 is the sorptivity [44]. Then, SHC (k, [mm/h]) was calculated using the following equation:

$$\mathbf{k} = \frac{\mathbf{C}\_1}{\mathbf{A}} \tag{2}$$

where coefficient A is a value relating the Van Genuchten parameters (n and α) for a certain soil type to the suction rate (h0) and the infiltrometer disk radius (2.25 cm). Entering the values of n, α, and h0 (assumed in this study to be equal to −2 cm) of the experimental soils in the table reported in the MDI manual [44], a value of 2.8 was derived for A. Equations (1) and (2) were proposed by Reference [45].

SWR was measured as follows: 15 drops of distilled water were released, using a pipette, on the soil surface of a 1-m transect, to homogenize the changing soil conditions; the time necessary for drops to infiltrate completely into the soil was measured by a stopwatch. Before measurement, litter cover was removed, and the soil surface was cleaned using a brush. The high number of replicates (ten points per plot) assured the best reliability for this measurement. The method used is recognized as one of the most appropriate for evaluating the SWR degree in field measurements [46].

#### *2.3. Statistical Analysis*

A one-way ANOVA (using land use as the independent factor) was applied to evaluate the statistical significance of the variations in soil cover, physico-chemical properties, SWR and SHC. All the plots were considered spatially independent. An independent Fisher's minimum significant difference test (LSD) was used for the post hoc analysis comparisons. A *p* < 0.05 level of significance was adopted. It was not necessary to perform data transformations for the analysis. ANOVA assumes normality and this assumption was checked using QQ-plots. All measured variables were used to perform the principal component analysis (PCA), the latter being based on a Spearman rank correlation matrix, to reduce the dimensionality of the data set. The statistical analysis was performed by version 3.24 of the R Project for Statistical Computing.

#### **3. Results**

The soil physical characteristics as measured for the three land uses are shown in Table 1. The texture of the experimental soils was statistically similar between AF and IA; the texture of the FO soil was significantly different from the other land uses. In general, the soil was finer in the AF and IA land uses (clay loam and loam, respectively) and coarser in FO. Soils subject to IA were more compacted (mean bulk density of 1.25 g cm<sup>−</sup>3), while the FO soils had a lower bulk density (0.83 g cm−3) (Table 1).

**Table 1.** Soil physical characteristics (mean ± standard deviation) in the study area (Villamalea, Castilla La Mancha, SE Spain). Different lower-case letters indicate statistically significant differences following LSD test (*P*-value < 0.05).


\* AF: Abandoned Farmland; IA: Intensive Agriculture; FO: Forest.

The IA showed the lowest vegetation cover (on average only 1% of the total sampled area). This latter parameter was similar between the other land uses (about 30% and not significantly different between abandoned farmland and FO). On average, only 1% of the soil in the FO plots was bare against more than 70% in IA (Table 1). The percentage of dead woody matter (plant material less than 2 mm in diameter) was statistically significantly different for all land uses, being highest in FO plots, followed by IA and AF soils.

The soil chemical characteristics as measured in the three land uses are shown in Table 2. The values of pH (on average 8.5–8.7) and EC (0.21–0.24 mS cm<sup>−</sup>1) were not significantly different among the investigated land uses. The OM content, in tune with the soil total carbon, was similar between AF and FO (mean values between 1.3 and 1.5%), and significantly lower in IA (on average 0.6%). OM was not correlated to dead woody matter, presumably due to the different mineralization levels of the investigated soils, on which the previous and current tillage practices may have played a role. Also, the contents of P, K, and cations (Na+, Ca++ and Mg++) were significantly different between land uses. FO soils had the lowest concentrations (except for Ca++) and the AF the highest (except for P). Conversely, there were no statistically significant differences in N content (on average 0.05–0.1%). Due to this, the C/N ratio was quite similar among all the studied land uses (mean values between 11.9 and 16.3), in spite of the differences recorded in C mean contents (from 0.6% of IA to 1.5% of AF). Finally, CEC was significantly higher in AF (on average 16.3 meq/100 g of soil), while FO soils showed the lowest value (6.8 meq/100 g, this latter not significantly different from IA, 9.9 meq/100 g) (Table 2).

**Table 2.** Soil chemical characteristics (mean ± standard deviation) in the study area (Villamalea, Castilla La Mancha, SE Spain). Different lower-case letters indicate statistically significant differences following the LSD test (*P*-value < 0.05).


\* AF: Abandoned Farmland; IA: Intensive Agriculture; FO: Forest; OM: Organic Matter; CEC: Cation Exchange Capacity; EC: Electrical Conductivity; C: Total carbon; N: Total nitrogen.

The analysis of the hydrological properties of the investigated soils showed much higher mean SWR values on soil surface compared to measurements made at 2 cm below ground (4-fold, AF, and 20-fold, FO). Only in soils under IA, the surface, and below ground SWR were similar. Moreover, FO soils had the highest SWR at both investigated depth (on average 74 s, on the soil surface and 3.5 s, 2 cm below the ground), while the lowest values were measured in IA (2.5 s) on soil surface, and AF (2.0 s) at 2 cm below the ground. Soil water content was 10.0% for AF and FO and 15.2% for IA) at the date of measurements (Figure 3a).

(**b**)

**Figure 3.** Soil water repellency (SWR, measured on the soil surface and at 2 cm below ground, (**a**), water content (SWC, **b**) and hydraulic conductivity (SHC, **b**) in the study area (Villamalea, Castilla La Mancha, SE Spain). Different lower-case letters indicate statistically significant differences comparing land uses for SWC and SHC following LSD test (*P*-value < 0.05).

In soils under IA, the highest SHC was detected (on average 17 mm/h), while the FO soils showed the lowest values (5 mm/h), similar (but significantly different) from SHC measured in the AF (mean value of 7 mm/h) (Figure 3b and Table 3).

**Table 3.** Results of ANOVA applied to land uses (Abandoned Farmland; Intensive Agriculture; Forest) for soil water content (SWC), repellency (SWR) and hydraulic conductivity (SHC) in the study area (Villamalea, Castilla La Mancha, SE Spain).


\* SWC = Soil Water Content; SWR = Soil Water Repellency; SHC = Soil Hydraulic Conductivity.

The PCA evidenced a clear clustering of the three investigated land uses with regard to the properties of soils, assumed as original variables (Figure 4). The PC1 and PC2 explained 51% and 37% (in total 88%), respectively, of the total variance of the original variables (Table 4). EC, pH, CEC, and nutrient contents were the variables with the higher loading factors on PC1, whereas plot % of vegetation cover, bare soil, dead woody matter, total carbon content, and bulk density showed the higher loadings on PC2.

**Figure 4.** Principal component analysis applied to properties of soils subject to different land uses (AF: Abandoned Farmland; IA: Intensive Agriculture; FO: Forest) in the study area (Villamalea, Castilla La Mancha, SE Spain) (the ellipses represent the evident clusters achieved by coupling ANOVA and PCA and using land uses as factors).


**Table 4.** Factor loadings on the first three principal components (PC) applied to properties of soils subjected to different land uses in the study area (Villamalea, Castilla La Mancha, SE Spain).

Note: in parentheses the percentage of the total variance explained by each PC is reported.

#### **4. Discussion**

Many studies have demonstrated how and by what extent land uses influence the physico-chemical and hydrological characteristics of soil (e.g., [4,11,16,34,47]). However, these characteristics also depend on the specific soil properties, such as texture and vegetal cover. For the investigated soils (sandy to clayey texture with variable vegetation cover), it has been shown that some of the chemical characteristics (namely pH, EC, N content and the C/N ratio) were very similar among the investigated land uses. Conversely, significant gradients in the OM content and CEC, two very important soil properties regarding soil fertility and productivity, were noticed when comparing AF (showing the highest values) to soils subject to IA. This soil depletion in OM and CEC may be attributable to plant uptake, due to the crop growth [48]. Moreover, soil under IA has lower OM and higher C/N ratio on average, although not optimal [49]; the other two land uses (AF and FO) have much less favorable C/N ratio for plants.

The most important variations among the investigated land uses were noticed in the soil hydrological properties: SWR and SHC varied significantly among the experimental land uses. More specifically, the topsoil of forested land had a much higher SWR compared to the surface soils subject to IA and AF. Even a slight SWR of FO soil may have noticeable effects on water infiltration rates, generating more runoff—particularly in summer, when the soil is drier—and thus enhancing soil erosion [27]. However, although it has been reported that SWR is particularly common under rangeland or forestland such as permanent grassland and deciduous shrub and forest terrain [50–52], it can also occur on agricultural land [27]. SWR affects both coarse and fine-textured soils [21,29,53] and occurs at low to moderate water contents [26,54]. In particular, coarse-textured soils, as those of the forestland analyzed in this study, are more prone to water repellency, even when they are permanently vegetated [55]. The level of SWR depends on the soil particle fraction with a hydrophobic surface coating [52] and is influenced by the surface area of the particles, which varies considerably with soil texture [56]. Therefore, in the sandy soils of our forestland, which have the lowest surface area, a hydrophobic surface impacts a larger proportion of particles than for a loamy or clayey soil where the surface area is up to three orders of magnitude greater [56,57]. Moreover, the plant species surveyed in

the forest of this study (eucalyptus and pines) are perennial trees with a considerable concentration of resins, waxes or aromatic oils, which have commonly been associated with SWR [58–60]. Under these conditions, it may be advisable to develop proper strategies to reduce SWR, such as a more effective soil management, the addition of clays to increase particle surface area, tillage to break-up and abrade hydrophobic surfaces and the use of chemical wetting agents [29,61]. Moreover, regions with a Mediterranean climate with prolonged dry periods, such as southern Spain, could be particularly affected by SWR and its hydrological impacts, which bring soils within a water content range in which SWR is exhibited [18,62,63].

Conversely, SWR levels surveyed in the other two investigated land uses (IA and AF) are less severe. In the soils previously subjected to agricultural activities, plant natural succession after land abandonment helps to avoid or reduce SWR, contributing to water penetration into the deeper soil layers, thanks to preferential flow paths via plant roots and stem flow [64]. Regarding the plots subjected to IA, agricultural soils are usually considered wettable [29], even though some studies have found that soil management practices can induce water repellency also in cropped areas [30–32]. In general, tilled sites are virtually unaffected by SWR [55]; furthermore, cultivation promotes rainfall infiltration, and, as a consequence, the runoff and erosion rates are significantly reduced [65]. However, although runoff and erosion are mostly reduced after plowing, these processes may increase again over time because of crusting, especially in silty soils [66].

Beside the low SWR levels, the reduced SHC of forestland detected in this study, compared to soils subject to IA, deserves much attention. As a matter of fact, depending on the severity of SWR, such soils may have water infiltration rates lower than would be expected on the basis of their pore size distribution and SWC (e.g., [20,25]). The combination of a reduced wettability (due to high SWR) and low SHC may enhance surface runoff (e.g., [26,67]). With this type of land use, the exposure of soils to the highest and most erosive rainfalls in the experiment area during autumn/winter may aggravate the erosive risks. However, this particular hydrological response of forest sandy soils can be prevented by a proper ground cover and by natural vegetation, which, having a strong influence on soil hydrological properties, reduces soil erosion rates [68].

Also, in the AF of this study, a lower SHC was detected (of the same order of magnitude as FO), which may lead to the degradation of the ecosystem at least during the first 3–5 years after abandonment [65]. This reduced SHC could be mainly due to the finer texture of AF soil compared to IA, but also the land use may have had a role. However, it has been demonstrated that, after land abandonment, the vegetation recovery should improve soil water-retention capacity and hydraulic conductivity, thus increasing infiltration and decreasing runoff and erosion rates [65,69]. In such a way, plant re-growth starts to control the hydrological and erosional soil response [64]. Moreover, the C/N ratio of AF soils is less favorable than in agricultural soils (and this could be due to the absence of fertilizer applications), which subsequently influenced plant succession (depending on the duration of abandonment) [49].

Overall, the worse hydrological response of FO soils compared to IA and AF plots, detected in this study, suggest that, although the forest tree planting by reforestation has been adopted as a viable land management strategy in many parts of the Mediterranean basin, natural scrubland, similar as those of revegetation processes in AF, may be more appropriate with regard to water use efficiency and soil conservation measures, since they assure a lower SWR and a slightly higher SHC [18].

However, it must be understood that the SHC measurements of the tension infiltrometer carried out in this study relate to the unsaturated soil. This parameter may be initially more important than saturated hydraulic conductivity at the time scale of convective rainfall, which is typically short and only lasts 20–60 min, common in many Mediterranean areas [70]. Since this investigation showed that soils of forestland and AF are affected by a lower SHC compared to agricultural areas, these land uses require caution in semi-arid or arid Mediterranean ecosystems, where runoff and soil erosion risks may be high. As a matter of fact, given that in the soils typical of this climatic context (in particular, those of prevalent sandy texture) the Hortonian (that is, infiltration-excess) overland flow type dominates over

concentrated runoff, a reduction of SHC may worsen the hydrological behavior of soils subjected to these land uses, with a possible increase of water runoff and soil erosion [71]. In fact, a lower water infiltration rate increase reduces the water storage of soil during heavy storms, thus increasing the share of precipitation, which generates surface runoff [16,47].

#### **5. Conclusions**

The effects of land use (abandoned farmland, intensive agriculture and natural forest) have been evaluated at the plot scale with particular reference to the physico-chemical properties, water content and repellency, and the hydraulic conductivity of soils. While most soil properties were not significantly different between the land uses, the hydraulic properties of the investigated soils showed specific responses to the different land uses or plant covers. Forest soils showed -high water repellency and low infiltrability, which worsens their hydrological behavior under heavy and frequent storms typical of the Mediterranean landscape. This behavior may increase the risks of soil erosion and pollutant runoff downstream in sloping areas. Abandoned soils previously subjected to agriculture showed a moderate water repellency, but their low hydraulic conductivity can cause serious problems in terms of runoff and soil erosion. However, shrub vegetation recovery, resulting from plant succession, can decrease this concern by increasing soil cover, which may reduce its erodibility. Compared to the forestland and the abandoned land, the agricultural soils were less affected by water repellency and low infiltrability, presumably due to the periodical tillage operations.

Overall, the main conclusion of this study is the important effect of land use on the hydrological characteristics of soils, and indirectly on their different susceptibility to surface runoff and erosion. This suggests paying attention to the specific land use and soil type under the Mediterranean climate (namely, steep sandy soils of forest ecosystems and also in the context of the expected climate changes), which may be affected by high runoff and erosion rates.

**Author Contributions:** Conceptualization, M.E.L.-B.; methodology, M.E.L.-B.; formal analysis, M.E.L.-B.; investigation, P.A.P.-Á., J.S., J.G.-R., and D.M.; resources, J.d.l.H.; data curation, D.A.Z. and M.E.L.-B.; writing—original draft preparation, D.A.Z.; writing—review and editing, D.A.Z. and M.E.L.-B.; supervision, V.Z. and J.B.; project administration, J.d.l.H.; funding acquisition, J.d.l.H.

**Funding:** This research received no external funding.

**Acknowledgments:** This study was supported by funds provided by the University of Castilla-La Mancha to the Forest Ecology Research Group. Also, by the Research Project from the Spanish National Institute for Agricultural and Food Research and Technology (INIA), VIS4FIRE (RTA2017-00042-C05). The authors thanks to Carlos Navarro, Beatriz Ariño and Jose Luis Martínez for the field assistance.

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

#### **References**


© 2019 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 (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Impact of Farmland Abandonment on Water Resources and Soil Conservation in Citrus Plantations in Eastern Spain**

#### **Artemi Cerdà 1, Oren Ackermann 2, Enric Terol <sup>3</sup> and Jesús Rodrigo-Comino 4,\***


Received: 28 March 2019; Accepted: 17 April 2019; Published: 19 April 2019

**Abstract:** Due to the reduction in the prices of oranges on the market and social changes such as the ageing of the population, traditional orange plantation abandonment in the Mediterranean is taking place. Previous research on land abandonment impact on soil and water resources has focused on rainfed agriculture abandonment, but there is no research on irrigated land abandonment. In the Valencia Region—the largest producer of oranges in Europe—abandonment is resulting in a quick vegetation recovery and changes in soil properties, and then in water erosion. Therefore, we performed rainfall simulation experiments (0.28 m2; 38.8 mm h−1) to determine the soil losses in naveline orange plantations with different ages of abandonment (1, 2, 3, 5, 7 and 10 years of abandonment) which will allow for an understanding of the temporal changes in soil and water losses after abandonment. Moreover, these results were also compared with an active plantation (0). The results show that the soils of the active orange plantations have higher runoff discharges and higher erosion rates due to the use of herbicides than the plots after abandonment. Once the soil is abandoned for one year, the plant recovery reaches 33% of the cover and the erosion rate drops one order of magnitude. This is related to the delay in the runoff generation and the increase in infiltration rates. After 2, 3, 5, 7 and 10 years, the soil reduced bulk density, increase in organic matter, plant cover, and soil erosion rates were found negligible. We conclude that the abandonment of orange plantations reduces soil and water losses and can serve as a nature-based solution to restore the soil services, goods, and resources. The reduction in the soil losses was exponential (from 607.4 g m−<sup>2</sup> in the active plot to 7.1 g m−<sup>2</sup> in the 10-year abandoned one) but the water losses were linear (from 77.2 in active plantations till 12.8% in the 10-year abandoned ones).

**Keywords:** agricultural land management; irrigated fields; erosion; abandonment; soil properties

#### **1. Introduction**

Citrus plantations are widespread due to the increase in the fruit demand around the world [1,2]. The traditional citrus production regions such as Valencia face a crisis due to production in other regions [3]. A social and economic change is taking place in traditional citrus production areas due to the ageing of the population, the competition of the large farms managed by companies, and social and economic changes [4]. The small size of traditional farms is also a key concern for the viability of production [5]. Citrus production in Valencia is facing technical modernization in large farms, with

drip irrigation and highly mechanized systems where labor is reduced [6] but the soil and water losses are enhanced (Keesstra et al., 2019). In front, we have traditional small size farms, with flood irrigation and high labor costs that result in low income for the farmers. Moreover, in the Valencia region, the ageing of the farmers and the social changes (to be a farmer is not attractive for the new generation) and low income associated with farming results in farmland abandonment [7].

Land abandonment has been widely researched in the last 50 years [8,9]. This is a consequence of the changes in the economy of developed countries rather than triggered different drivers [10,11]. From an environmental point of view, abandonment results in an increase in the infiltration rates [12] due to vegetation recovery [13]. Soil erosion is also reduced by the effect of the vegetation cover and the recovery of organic matter [14].

The research carried out on recent land abandonment was done in rainfed land. This is due to the fact that the abandonment took place in rainfed land that did not evolve to irrigation [15]. The intensification of the agriculture contributed to increasing the land irrigated; meanwhile, the one that was less productive, usually the rainfed one, was abandoned. The research carried out on irrigated land is recent and focuses mainly on socioeconomic issues [16]. The abandonment provides ecosystems services such as carbon sequestration, water storage, soil properties, and seed bank fate improvements [17]. This response to land abandonment is widespread around the world [18–20].

Research on the abandonment of traditional irrigation farms was restricted until now to the historical, archaeological, and ecological approach [21–23]. There is no information about the changes in soil hydrology and soil erosion once the abandonment takes place in traditional flood irrigated land. This paper is the first approach to understand the effect of irrigated farmland abandonment on soil erosion and runoff discharge. Our main goal is to determine how the soil and water resources change once abandonment takes place and how to evolve along a decade with the use of plots in use or abandoned for 1, 2, 3, 5, 7 and 10 years.

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

#### *2.1. Study Area*

The research presented in this paper was carried out in the Canyoles River watershed, in the fluvial terraces (145 m a.s.l.) in the municipality of Canals (Valencia, Spain), a representative zone of the Mediterranean citrus plantation that was established along the 1960s. The farm properties are small in size in this region, and usually, all the farms are smaller than 1 ha and the average at the study site is 0.45 ha. The active field is flood irrigated and herbicide (Glyphosate) is applied any time the weeds germinate, which results in a bare soil surface. Once abandonment takes place there is a quick recovery of the vegetation. The climate is typically Mediterranean with a mean annual precipitation of 532 mm and a mean annual temperature of 16.2 ◦C. The soil is an Anthrosol [24] and the texture of the soil is clay-loam: 21% clay, 39% silt, and 40% sand. The planting pattern is 5 m × 4 m, the usual pattern for citrus in this agricultural area. The farm was flood-irrigated with freshwater from the Riu de Sants, which flows from the Massís del Caroig aquifer. The slope is negligible due to the land flattering and the effects of the floods on the fields. The soils are basic in pH (7.9). The observed soil profiles were very homogeneous as a consequence of tillage practices for centuries and land levelling. The study area was selected in the Pleistocene fluvial terrace of the Canyoles River (near the hamlet of Aiacor) and show 22% gravel content. The irrigation system (Sèquia de Ranes) flows from the nearby Riu de Sants spring. Irrigation takes place every 2 weeks in the summer and does not take place once the fields are abandoned. The management up until the time of abandonment was similar in all plots: herbicide (Glyphosate (N-(phosphonomethyl)glycine) and inorganic fertilizers (NPK—nitrogen, phosphorus, and potassium—0.8 Mg ha−1 per year). Once the plots are abandoned, neither irrigation nor fertilization takes place.

#### *2.2. Experimental Design*

The experimental design was based on the fact that the study area has suffered since 2000 a progressive abandonment due to the low income in such small parcels, the dropping of the prices of the oranges, and the ageing of the landowners. Then we selected a farm that was active and the neighbouring farms that had been abandoned for 1, 2, 3, 5, 7 and 10 years (Figure 1A). At each farm, we selected 10 representative 1-m<sup>2</sup> plot where sampling and simulated rainfall experiments on 0.28-m<sup>2</sup> plots were carried out (Figure 1B). All the plots have *Citrus sinensis* plantations with Naveline variety and "Carrizo" rootstock.

**Figure 1.** Scheme of the study plots (**A**), localization of the experimental fields (**B**), ring plot in the active farm (**C**), and field campaign (**D**).

#### *2.3. Soil Sampling*

The seven experimental fields show different ages of abandonment. Age 0 is the active field with no weed cover due to the use of herbicides, and 1, 2, 3, 5, 7, and 10 years are the fields that were abandoned along the last decade. The soil sampling took place in August 2014 during the Mediterranean summer drought. Ten soil samples were taken from the top 6 cm of the soil using a 100 cm3 steel cylinder. We measured gravimetric soil water content (SWC) and bulk density (BD). Moreover, soil samples were weighed, oven dried at 105 ◦C during 24 h, and weighed at room temperature to estimate organic matter by the dichromate method [25]. Also, grain size distribution was calculated by the pipette method. Vegetation cover, stones, and soil crusts were estimated using 100 pins evenly distributed in the rainfall simulation plot.

#### *2.4. Rainfall Simulation Experiments*

Seventy experiments (10 repetitions × 7 plots) during 1 h for the active and 1, 2, 3, 5, 7, and 10 years of abandonment were carried out at 38.8 mm h−<sup>1</sup> rainfall intensity on the circular paired ring plot (0.28 m2; Figure 1C,D). These rainfall events are characterised by a return period around 2 years in the eastern Iberian Peninsula and have been used widely in rainfall simulation experiments [26]. All experiments were carried out when the soil moisture was low during the last week of July 2014 with no rainfall events to avoid variability among plots. Runoff was collected each 1-min interval and total water loss was calculated. Then, the runoff coefficient was obtained by means of the percentage of rainfall water flowing through the ring plot. In the laboratory, runoff samples were desiccated (105 ◦C, 24 h) and soil loss was obtained based on the weight basis per area and time (g m2 h<sup>−</sup>1). Furthermore, it was also possible to quantify time to ponding (time required for 40% of the surface to be ponded; Tp, s), time to runoff initiation (Tr, s), and time required by runoff to reach the outlet (Tro, s). Those parameters show how the runoff initiation takes place and how ponding is transformed into runoff and how the runoff reaches the runoff outlet.

#### *2.5. Statistical Analysis*

Descriptive statistics were calculated (mean, median, maximum, and minimum) and used for further statistical analysis. Soil properties were represented in linear graphs in order to show their evolutions during the studied period using Excel software (Windows, Redmond, Washington, DC, USA). The number of points used for each analysis corresponded to all measurements done at each location. Then, the hydrological response was represented as polar graphs dividing by intervals (natural breaks) the obtained results. Finally, soil erosion results were depicted in scatterplots using SigmaPlot 12.0 (Systat, Chicago, IL, USA). Statistical differences were evaluated performing an ANOVA one-way test to check the statistically significant differences among years of abandonment. If the normality test failed (Shapiro–Wilk), a Tukey test was conducted when the homogeneity variance fails (Levene´s test). Finally, a Spearman rank correlation coefficient was carried out in order to assess the possible connection between environmental plot characteristics and soil erosion results.

#### **3. Results**

#### *3.1. Soil Properties*

The soils of the seven study sites affected by different times since abandonment show changes that can shed light into the evolution of the soil properties upon abandonment (Figure 2). Antecedent soil moisture is reduced from 14% to 5.1% in ten years of abandonment. Bulk density does not show significative changes after 1 year of abandonment (from 1.38 to 1.37 g cm<sup>−</sup>3), but after that, the values decrease to 1.15 g cm−<sup>3</sup> in the tenth year of abandonment. On the contrary, organic matter and plant cover quickly increased after the abandonment, showing changes from 0.93 to 1.79% and from 1.1 to 90.3%, respectively. The cover of rock fragments varies among plots, reaching the maximum values in

the plots after 1, 2, and 3 years of abandonment. Finally, the development of soil crusts also shows a drastic decrease from the active plot (83.4%) to the 10 years abandoned plot (3.4%).

**Figure 2.** Variation of soil properties over different time periods since abandonment. (**A**) Antecentt soil moisture; (**B**); Bulk density; (**C**); Organic matter; (**D**); Plants; (**E**) Rock fragments; (**F**) Crust.

#### *3.2. Hydrological Response*

Figure 3 shows the time to ponding, time to runoff generation, and time to runoff in the outlet (seconds) represented in polar graphs. Table 1 shows the results of the Tukey test applied to all the seven study sites. The results show that among plots, there exists a clear statistically significant difference after the abandonment (*p* < 0.001). Time to ponding moves from 25.9 to 238.1 seconds, respectively, from the active plot to the 10 years abandoned one. Time to ponding was much delayed after seven years of abandonment. Runoff was generated after 47.2 seconds in the cultivated plot and after the runoff was delayed by 477.8 seconds. There was a clear trend that showed a progressive delay in the ponding and runoff initiation once the abandonment took place. Moreover, a clear variance among the 10 repetitions in each plot was also relevant, which could be affected by the uneven distribution of the plant recovery. The time to a runoff in the outlet was faster (88.3 s) in the active plot and delayed after the abandoned. The 10-year abandoned plot needed 917 s to achieve runoff at the outlet.

**Figure 3.** Hydrological response depending on the period of abandonment (360◦ = range of values).


**Table 1.** Statistical differences of hydrological responses and soil erosion results among periods of abandonment. Tp: time to ponding; Tr: time to runoff generation; Tro: time to runoff in outlet; Rc: runoff coefficient; SC: sediment concentration; SL: soil loss.

#### *3.3. Runo*ff*, Soil Loss, and Sediment Concentration*

In Figure 4, runoff coefficient, sediment concentration, and soil loss are depicted in box plots. The runoff coefficient shows the highest rates in the active plot, reaching average values of 77.2% of the rainfall, with maximum ones of 96.3%. On the contrary, in the other plots, the average runoff coefficient descends to 71.8, 51, 42.3, 32.3, 22.3, and 12.8% for 1, 2, 3, 5, 7- and 10-years of abandonment, respectively. The highest variability can be observed after the second year of abandonment. In Table 1, it is possible to observe values with non-statistically significant differences in runoff coefficients between 0–1 year and 2–3-years after abandonment. Sediment concentration also shows a drastic decrease from the active plot to the 10-year abandoned field, obtaining values from 14.7 (maximum values of 20.4 g L<sup>−</sup>1) to 1 g L−<sup>1</sup> (maximum values of 1.02 g L<sup>−</sup>1). Table 2 shows that all the plots have significant differences in this parameter. The soil loss parameter also shows a decrease: from 607.4 (active plot), 271.4 (1 year), 150.3 (2 years), 62.8 (3 years), 33.46 (5 years), 19.7 (7 years) till 7.1 g m−<sup>2</sup> (10 years). Table 1 shows that between 1–2 years and 5–7 years of abandonment, no differences can be observed, but there are differences in other years.

**Figure 4.** Evolution of the runoff coefficient, sediment concentration, and soil loss along the abandonment time from active plantations to 10-years old abandonment. Dotted line: average (*n* = 10). (**a**) Runoff coefficient; (**b**) Sediment concentration; (**c**) Soil loss.

**Table 2.** Spearman rank coefficient between environmental factors, hydrological responses, and soil erosion. SM: antecedent soil moisture; BD: bulk density; OM: organic matter; Plants: plant cover; Rock: rock fragment cover; Crust: soil crust; Tp: time to ponding; Tr: time to runoff generation; Tro: time to runoff in outlet; Rc: runoff coefficient; SC: sediment concentration; SL: soil loss.


In Table 2, it can be observed that only the rock fragment cover does not show a high correlation with the hydrological responses and soil erosion results. On the other hand, the highest Spearman rank coefficient values were obtained between the development of a high plant cover and the positive relationship with Tp (0.95), Tr (0.95), and Tro (0.95), and the negative one with Rc (−0.87), SSC (−0.88), and SL (−0.84). Moreover, high positive correlations were observed with the OM contents. On the contrary, higher values of antecedent soil moisture, bulk density, and soil crust were highly correlated with Tp, Tro, and Tro, and Rc, Sc, and SL.

#### **4. Discussion**

The use of developing countries as food producers, an increase in agricultural productivity, and the use of fossil fuels that reduce the need for animal traction and then for food production for them has reduced the need to use part of the agriculture land [27,28]. This abandonment has environmental consequences such as changes in water resources, vegetation cover and floristic composition, fauna, and soil properties, which will affect also earth system functioning as they will control the biogeochemical cycles [29–33]. Agriculture land abandonment is not a new issue from a social, economic, and environmental point of view [34], but what is new is the agriculture technology that was developed after War World II which has led to the widespread abandonment and contributed to the restoration of the vegetation and fauna and the reduction in the soil losses in many parts of the world [35,36].

The research on the impact of abandonment of agriculture on environmental issues focused on rainfed land [37], as the irrigated land used to be intensified [38,39]. Our research contributes to new data generated in flood irrigated land that is being abandoned due to socioeconomic changes. The findings demonstrated that the soils are able to recover plant cover and soil organic matter once abandoned as other authors demonstrated in other types of agricultural land [17,40]. The changes in plant cover will affect fauna such as has been demonstrated by other colleagues [41–43] and soil properties. The non-utilisation of irrigation has generated soils which contain a lower amount of water (antecedent soil moisture), which could affect some other pedological processes such as organic matter development or aggregate stability [44,45]. Considering these changes, the diminution of soil bulk density is also a relevant factor that could affect soil erosion activation [46]. As Bienes et al. [47] observed in Central Spain, significant changes in bulk density used to appear after the abandonment took place due to plant recolonization because of the root development and organic matter increase. In our study area, weeds, grass, and small shrubs were distributed in little patches that directly affect these pedological changes.

The plant cover (weeds) was mainly *Paretaria Judaica*, *Conyza sumatrensis*, *Amarathus hibridus* L., and *Chenopodium album*. Once abandoned, the cover of *Paretaria Judaica* increased, but other plants appeared such as *Avena fatua*, *Asparragus* sp., and *Rubus ulmifolius* that finally became the dominant species.

Once the land was abandoned, we also observed a significant difference in the rock fragment cover. Possibly, the surface wash and tillage could remove the fine material and more rock fragments were visible on the surface [48,49]. However, after the third year of abandonment, the vegetation cover reduced neither the crust nor the rock fragments in the soil surface. This would also be an interesting topic for research in the future because of its direct impacts on soil erosion processes and biogeological systems.

Our research confirms that after the abandonment there was a sudden reduction in overland flows that is shown in delayed ponding and runoff generation. This means that more water infiltrates. However, our measurements also demonstrate that the amount of water flowing through and on the soil is also reduced after the abandonment takes place. This can be due to the reduction in the effective rainfall as the interception increases with the growth of the vegetation and the development of a litter layer [50,51]. The impact of the litter was researched by the pioneer study of Facelli and Pickett [52] and found that litter plays a key role in the water balance. Rainfall interception in abandoned fields is related to the vegetation recovery, and the higher the biomass means a lower the net precipitation on the soil [53].

The loss of water reaching the soil after the abandonment of land changes the hydrological cycle, such as Hou et al. [54] found in the Loess hillslopes in China, Šraj et al. [50] in Slovenia, and Otero et al. [55] in Catalonia (Spain), where a loss of stream biodiversity and water availability was found. Those findings are highlighted by García-Ruiz and Lana-Renault [13] along with their review on the hydrological and erosive consequences of farmland abandonment in Europe. The impact of abandonment shows less water availability and more water used by the vegetation. This results in a river discharge reduction such as was demonstrated in the Central Spanish Pyrenees by Beguería et al. [56] or in Slovenia where the Dragonja River reduced the sediment delivery due to the loss of the runoff discharge as a consequence of the natural afforestation [57]. The use of water by the vegetation (transpiration) also resulted in the loss of water available by plants such as Moreira et al. [58] found in the Amazon on abandoned pastures, Rambousková et al. [59] in the abandoned fields in the Bohemian Karst, and Farrick and Price [60] in the Sphagnum restoration in peatlands.

Our research at the Canals municipality demonstrated that the reduction in the water yield was efficient as a consequence of the abandonment (from 77.2% in the active orange plantation to 12.8% in the 10-years abandoned plots), but the reduction shows a linear correlation (0.91; Figure 4a). However, for the sediment concentration and soil loss, the reduction followed a negative exponential trend with an adjustment of 0.97 and 0.95, respectively (Figure 4b,c). This trend in the reduction of soil and water losses were also found in previous literature. Ruecker et al. [61]; Koulouri and Giourga [62], and Lesschen et al. [63] found this trend in abandoned Mediterranean terraces, Cerdà [12] in southeastern Spain in a semi-arid landscape, Liu et al. [64] in the Loess Plateau, and Arnáez et al. [65] in the La Rioja region, wherein the Camero Viejo district they found that abandonment controlled the soil erosion rates and landscape evolution. Land abandonment is a recurrent topic in the Mediterranean, and Portuguese examples confirm our findings here: control of soil losses by vegetation recovery [66].

Abandonment could be seen as a nature-based solution such as the use of straw in forest fire affected land [66,67] to the high erosion rates found in agricultural land [68], and could be used as a strategy to balance the carbon cycle [69] and rehabilitate the soils under the millennia-old use of agriculture [70]. The research carried out at the Canals Municipality shows that abandonment could be positive from an environmental point of view, but there is also the risk of a forest fire as the vegetation could be very flammable during the Mediterranean summer drought [71]. Thus, the next essential step of research is to find the optimal management of the abandoned orchards, and maybe it can be used as a recreation area for nearby urban citizens [72]. This could be a successful solution in the Mediterranean where agriculture land abandonment in the last decade also takes place in the surroundings of city areas due to the economic crisis [69] and for which similar responses have been shown in other developed countries [73]. How urban changes occur is related in one way to the environmental conditions [74] but also to the cultural impact of humans [75].

Once the fields are abandoned, the vegetation recovery takes places as a mosaic of plants and this response results in an increase in the spatial variability such as other authors found along climatological gradients [76]. We found this increase in the spatial distribution due to the formation of tussocks, or bare and plant covered patches, which is a clear factor in soil erosion, but with more intensity after the second year of abandonment [77]. Our findings are based on a local survey and should be used to supply basic information to develop proper models of water balance and soil erosion [78] that will shed light onto the effect of other management systems such as organic farming, land abandonment or cover crops and mulches [79,80].

#### **5. Conclusions**

Citrus plantation abandonment results in a recovery of the vegetation cover and soil organic matter, and a reduction in the soil bulk density drought and soil moisture. Plant recovery is the key factor that determined a linear reduction in the runoff discharge (from 72 till 13% of the rainfall) over ten years of abandonment. The soil losses dropped from 607.4 g m−<sup>2</sup> in the active plot to 7.1 g m−<sup>2</sup> in the 10 years after abandonment took place. We conclude that the abandonment of orange plantations

could reduce soil and water losses if there is a proper plant recovery, which allows it to be considered as a potential nature-based solution to restore the soil services, goods, and resources.

**Author Contributions:** A.C.: investigation, methodology, data acquisition, data preparation, writing, supervising. O.A.: data analysis, data preparation, writing. E.T.: investigation, methodology, data acquisition, data preparation, writing. J.R.C.: data analysis, data preparation, methodology, writing.

**Funding:** This paper is part of the results of research projects GL2008-02879/BTE, LEDDRA 243857 and RECARE-FP7 (ENV.2013.6.2-4).

**Acknowledgments:** During the field work the music of Raimon "Diguem no" was our inspiration and during the data analysis and writing the histories of "El Comandante Lara". Moreover, we want to thank the guest editors and reviewers for their careful review process.

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

#### **References**


© 2019 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 (http://creativecommons.org/licenses/by/4.0/).
