1. Introduction
As the world population continues to grow, much more effort and innovation will be needed to increase agricultural production, improve global supply chain, and achieve sustainability. Sustainable Development Goals (SDGs) move towards that direction through supporting sustainable agriculture, empowering small farmers, and tackling climate change. The growing scarcity of soil resources in many parts of the world makes it vital to use and manage them sustainably. Concerning the current extent of land degradation and limited land resources, the need to optimize land use while preserving the ecosystem is crucial.
Sustainable land management requires the application of various techniques combined with local knowledge to achieve sustainability of the agroecosystems [
1,
2,
3]. Soil management along with ecosystem, environmental interactions, and climate change severely affect soil quality and soil organic matter (SOM), which are essential to maintain healthy and productive soils. Developing ecological agricultural practices is the key to maintaining high levels of soil organic matter to keep the soil fertile and thus achieve long-term sustainability. Magdoff and Van Es [
4] aimed to identify and evaluate Soil-Improving Cropping Systems (SICS) that would increase the profitability and sustainability of agriculture. The concept of SICS was applied by Van De Vreken et al. [
5], aiming to identify the interaction between residue management and soil properties on carbon sequestration. Selected SICS using modeling and Europe-wide data were monitored and evaluated for environmental, sociocultural, and economic effects to determine profitability and sustainability [
6]. The results show a small positive impact on environment and soil with variations across Europe, which suggests a better understanding of the local dynamics Europe-wide. In a detailed review, the suggestions for monitoring included only the soil indicators that could be used as point measurements without any spatial information related to crops [
7]. Another synthesis of reviews has analyzed the knowledge gained in literature reviews on SOC changes in long-term experiments (LTEs) and evaluated the results regarding interactions with pedo-climatological factors [
8]. All these studies use point measurements and legacy databases without any spatial analysis of the information and relation to crops, thus inserting errors when extrapolating to the regional scale.
Land suitability evaluation can contribute to agricultural land use optimization, as it used to determine the most appropriate spatial plan for current and future land uses [
9,
10] and constitutes a useful tool to achieve sustainable use of the available resources and limit land degradation. The Food and Agriculture Organization (FAO) approach has been used as a major Land Suitability Analysis (LSA) framework, which gives the basic guidelines in agriculture to carry out a land evaluation process [
11,
12]. Furthermore, LSA can be used to evaluate scenarios of agricultural management practices that contribute to increased productivity, while reducing agricultural inputs, conserving natural resources, and mitigating climate change impacts. Such management practices are those aiming at carbon sequestration in the soil, including manure application and crop residue incorporation, and contributing to reduced fertilizer input, soil quality improvement, and soil erosion amelioration.
LSA methods are classified as traditional, where biophysical characteristics are used to assess crop options using qualitative, quantitative, and parametric methods [
13], as well as modern methods combining GIS and machine learning algorithms [
14], where most of them are multi-criteria decision-making methods. Examples of traditional methods reported for the assessment of several scenarios of agricultural management practices in several crops are the following: (a) Parametric Methods used for wheat; (b) Boolean Logic, Maximum Limiting factor, and Weighed Overlay methods used for rice; (c) Weighted Linear Combination used for rice and soybean; (d) Square root mean used for cereal crops; (e) Expert Knowledge and FAO method used for Chemoriya; (f) Qualitative approaches for maize, potato, and vegetables; and (g) Computer overlay used for canola, soybean, and others. Respectively, the main modern methods used in Land Suitability Assessment to evaluate different scenarios of agricultural management practices are: (a) Analytical Hierarchy Process—AHP used for maize, potato, saffron, rice, grapes, wheat, and sugarcane; (b) fuzzy methods used for groundnut, maize, rice, soybean, sorghum, barley, spinach, wheat, rye, oats, sugarbeet, and more; (c) methods based on the use of crop models (mainly used for cereals); GIS-based Environmental Policy Integrated Climate (EPIC) model, Almagra, ECOCROP, CROPWAT; and (d) machine-learning-related methods (also used for cereals), such as Artificial Neural Networks, TOPSIS, Bayesian Networks (BNs), and Goal programming Species distribution models. In both cases, the thematic factors considered are “Climate”, “Soil and Landscape”, “Socio-Economic”, and “Land Use Land Cover (LULC)” [
14].
The main advantage of modern methods over the traditional is that they succeed in mapping areas with homogenous characteristics considering several variables and, therefore, they can provide solutions to more complicated scenarios. As a result, modern LSA methods compared with traditional methods might use more time-consuming or complex algorithms and procedures [
14], and can subsequently be applied to complex environmental challenges such as land degradation [
15] and landscape attractiveness in highly populated areas [
16]. The main LSA methods are usually classified as: (i) computer-assisted overlay mapping; (ii) artificial intelligence (AI) methods; and (iii) multi-criteria evaluation (MCE) or multi-criteria decision making (MCDM) [
17].
AHP contributes to complicated decisions when dealing with a mix of qualitative and quantitative factors (e.g., [
18,
19,
20]). According to Malczewski [
10], it can be used in two distinctive ways within the GIS environment. Firstly, it can be employed to derive the weights associated with suitability map layers for individual attribute properties. Secondly, the weights can be combined with the attribute map layers in a way similar to the additive Weighted Linear Combination (WLC).
Until now, there has been limited research conducted on how SICS could be integrated into LSA scenarios, except for the approach of an optimal scenario under the improvement of manageable limiting factors such as soil salinity [
21]. In this work, two scenarios were considered to assess and map soil suitability after applying the “Almagra” model: the current situation (CS), where management options were proposed to reduce some limiting factors (soil salinity and sodium saturation), and the optimum scenario (OS) (with a predefined fixed value of 2 dSm
−1 for soil salinity and 5% for sodium saturation). The results show a noticeable increase in the area classified as a highly suitable area for annual and semi-annual crops. Moreover, a marked increase was observed (about 70% for CS and 50% for OS) for perennial crops shifting from the marginal class to moderate suitability class. The results of this work indicate that proper management can contribute to soil suitability in the context of the sustainable land use of a fertile agro-ecosystem [
21]. For Belgium, Vanwindekens et al. [
22] studied the vulnerability of arable cropping and grassland to extreme weather events using a combination of expert knowledge derived from interviews, fuzzy inference systems, and geomatics. The factors underlying agro-ecosystem vulnerability clearly revealed that SICS such as crop rotation, cover crops, grassland species composition, and organic matter added to resilience.
The main objective of the current study is to investigate the use of Land Suitability Analysis (LSA) as a tool to evaluate Soil-Improving Cropping Systems (SICS). The suitability of land was assessed for maize in Flanders (BE), Somogy (HU), and Hengshui (CH) under the current situation, as well as after a hypothetical 100-year application of four scenarios: (a) Conservation Tillage, (b) Crop Cover, (c) Crop Residue Management, and (d) Manure Application.
2. Materials and Methods
The proposed methodology uses the FAO framework for LSA [
11,
12] as a tool to evaluate the level of current land suitability for growing maize and to evaluate the principal limiting factors. The FAO framework differentiates between land suitable for crops (S) and not suitable for crops (N). Suitable land is classified in three suitability classes: (S1) highly suitable, (S2) moderately suitable, which requires several inputs on the land to sustain the crop, and (S3) marginally suitable, which requires significant inputs on the land [
9,
10]. The FAO framework is also used to predict the future land suitability based on scenarios of applying four Soil-Improving Cropping Systems (SICS) for 100 years. The final product is the comparison between the current and future LS to evaluate the potential performance of the SICS.
The main steps in the current work are the following: (1) selection of the crop of interest (maize) in the three sites (Flanders—BE, Somogy—HU, and Hengshui—CH); (2) selection of the categories of Input Parameters (soil, water, climatic, topographic, environmental, and management properties); (3) determination of their threshold values; (4) calculation of the Input Parameters’ (IP) performance via Agricultural Land Use Evaluation System (ALUES software) by mapping land units of each IP into FAO classes; (5) determination of weighing factors for each IP via Analytical Hierarchy Process (AHP) based on local experts’ opinions; (6) calculation of a combined performance for each Category of Input Parameters (CIPs); (7) investigation of the IP as limiting factors (Limiting Factor Analysis); (8) validation of the LSA by comparing the results with the existing LUCAS microdata database; (9) selection of four Soil-Improving Cropping Systems (SICS): (a) Conservation Tillage (CT), (b) Cover Crop (CC), (c) Crop Residue Management (CRM), and (d) Manure Application (MA), which aimed to increase SOM and CEC to enhance soil quality, crop productivity, and carbon sequestration in soil; (10) adopted scenarios of 100-year application of management practices with re-calculation of new values for SOC and CEC; (11) calculation of a combined performance for each SIC scenario; and (12) cross-examination of the LSA classes with the existing LUCAS microdata database. Finally, via LSA maps and data, the comparison between current LS and SICS performance after 100 years of application can provide a projection of the land suitability in the future depending on the different agricultural practices adopted. The method of using LSA provides a quantification of the changes and a visual way to locate the areas most affected (
Figure 1).
2.1. Test Sites
The methodology was tested on three regional test sites, two in Europe and one in China. These were Flanders in Belgium, Somogy in Hungary, and Hengshui in China.
Flanders is located in the northern part of Belgium, sharing borders with France and the Netherlands (
Figure 2a). Geographically, Flanders is mainly flat, and has a small section of coast on the North Sea. The total area is 13,265 km
2 and is densely populated, and the soils are formed mainly on sedimentary and loess deposits. The main arable crops in the area are wheat, potato, maize, grass, and sugar beets. The climate is temperate and is characterized by high precipitation and water scarcity during summer. Winters are mild and rainy according to the Köppen climate classification. The mean annual temperature is 10.5 °C, mean annual lowest and highest temperature are 6.9 and 14.2 °C, respectively, and mean annual precipitation is 852.4 mm.
Somogy is an administrative county that lies in south-western Hungary, on the border with Croatia, with an area of 6064 km
2 (
Figure 2b). The main crops in the area are wheat, potato, maize, alfalfa, and sugar beets. Somogy constitutes two distinct regional units: the northern region, known as Outer Somogy, and the flatter, forested, and southern region, known as Inner Somogy, while Lake Balaton and the Drava River may also be regarded as geographically distinct regions. Clay and sandy soils are characteristic of this region. Peat and swampy forest floors indicate that this area was covered with extensive forests, swamps, and bogs in the past. The proportion of the county’s forest area remains very high (Somogy is the third most forested county), while the proportion of arable land, gardens, and orchards is below the national average. Somogy has a marine west coast and warm summer climate (classification: Cfb). Moving eastwards, the climate becomes more continental.
Hengshui is a prefecture-level city division in southern Hebei province (
Figure 2c). The Hengshui prefecture is located in the lowest part of the province’s terrain, with a total area of 8836.95 km
2. The climate in Hengshui is known as a local steppe climate (classified as BSk by the Köppen–Geiger system). There is little rainfall throughout the year. In Hengshui, the average annual temperature is 14.4 °C, and the average precipitation ranges between 500 and 900 mm, of which 60% is in the summer. The topography of this region is characterized by alluvial plains with low elevation. The main food crops are wheat, corn, grain, and sorghum, while the prevailing cropping pattern is winter wheat in rotation with summer corn. The soil is mostly tidal loam, two-combined soil, and sandy loam, fertile and moderate in texture [
22].
2.2. Input Parameters
Data used in the specific test sites were (a) soil resources properties, (b) climatic properties, and (c) topographic properties. The LSA tool’s input parameters were grouped in the following Categories of Input Parameters (CIPs):
Soil resources properties;
Water resources properties;
Climatic properties;
Topographic properties;
Environmental properties;
Management properties.
Soil data used were taken from SoilGrids (
https://data.isric.org/geonetwork/srv/eng/catalog.search, accessed on 5 August 2021) after integrating values across soil depth up to 60 cm (
Table 1) using Equation (1).
where
is the number of depths,
is the range of k depth, and
is the value of the soil parameter in the
depth.
Climate data were acquired from ERA5-Land, TerraClimate, and WorldClim 2, while terrain data were from the European Digital Elevation Model (EU-DEM) version 1.0 and Shuttle Radar Topography Mission (SRTM).
2.3. LSA Software
LSA was performed using the Agricultural Land Use Evaluation System (ALUES), which is an R programming (open-source statistical software) package developed for evaluating the land suitability of different crops. ALUES assesses suitability by classifying the land units using fuzzy logic. The assessment is based on the standards of the crop requirements from Sys et al. [
23]. The computations of the suitability scores and classes are based on the land use suitability evaluation tool [
24] with extended options for the membership function. The mathematics behind the computations are detailed in
https://cran.r-project.org/package=ALUES accessed on 12 November 2021.
Using ALUES, the task of evaluating land suitability becomes a task to map input characteristics of a land unit into the suitability class of the target factor by checking whether this characteristic is within any of the suitability classes. For example, suppose that an input land unit has terrain with a slope of 1 degree. The suitability classes used are s1—highly suitable, s2—suitable, and s3—marginally suitable. To complete the computation of the suitability scores, ALUES uses fuzzy logic (Triangular, Trapezoidal, and Gaussian functions) to evaluate the suitability of a land unit based on inputs (such as rainfall, temperature, topography, and soil properties). The methods for computing the overall suitability of an area (Minimum, Maximum, and Average) are also included. ALUES is a highly optimized library with core algorithms written in C++.
2.4. Analytical Hierarchy Process
The Analytical Hierarchy Process (AHP) is a multiple criteria decision-making method that was originally developed by Saaty [
25]. The method is widely used in many scientific fields to provide measures of judgement consistency. Using pair-wise comparisons, it successfully derives priorities among criteria and alternatives. The result of the AHP (setting ratings after pair-wise comparisons of the factors) can be evaluated via Consistency Index and Ratio and via Sensitivity Analysis. In a nutshell, the method of AHP simplifies preference ratings among decision criteria. AHP was implemented in the R computational environment, utilizing available resources from the online repository at
https://github.com/gluc/ahp, accessed on 14 September 2021.
2.5. Selection of Threshold Values
Threshold values for available input parameters are presented in
Table 2 and were adopted according to Sys, Van Ranst, Debaveye, and Beernaert [
23]. The table was customized for each site. Five soil parameters were used, which were the same for all sites, namely: CEC, SOC, pH, texture, and coarse fragments. All values were numerical, apart from texture, which was grouped in classes. Concerning topography, detailed Digital Elevation Models were available and provided slope maps that were the most important factor to be considered. For climate, Precipitation and Temperature were used after customization for the local growing cycle.
2.6. SICS Scenarios for 100 Years
To account for SICS scenarios, we adopted four cases of 100-year application of management practices, as seen in
Table 3. Scenarios in the study areas were adopted after Aertsens et al. [
26], Van De Vreken, Gobin, Baken, Van Holm, Verhasselt, Smolders, and Merckx [
5], Nadeu et al. [
27], Vanwindekens, Gobin, Curnel, and Planchon [
22], and personal communication with local experts. These scenarios were incorporated in the LSA procedure by calculating new values for SOC and CEC. The conversion of SOC values to soil organic matter values was achieved using a factor of 1,72, while the value of CEC for soil organic matter was assumed to be 200 cmol/kg.
2.7. Field Data
The LSA results were compared with the location of existing maize fields according to the LUCAS microdata database, which includes 27 locations in Flanders and 28 locations in Somogy. In Hengshui, due to lack of data, the LSA results were only examined by local experts.
4. Discussion and Implications for Land Management
The current study showed that the proposed methodology can produce LS maps to evaluate SICS scenarios by projecting soil characteristics and LS 100 years in the future. To our knowledge, this is the first use of LSA for evaluation of SICS scenarios. The use of LS maps (both limiting factor maps and SICS performance maps) was crucial to locate the exact areas that have the highest potential for improvement. The spatial information that these maps provide can give insights for better land use management and the selection of the proper SICS in the future for a specific region depending on local conditions. As a result, using the proposed methodology, the evaluation of SICS scenarios can be performed seamlessly using LSA as a tool to achieve high-quality results with reduced bias, instead of the previously common practice of evaluating the performance of scenarios in an empirical manner.
Concerning the determination of weighting factors for each input parameter (via AHP), it must be noted that across CIPs and regions (Flanders, Somogy, and Hengshui), soil and climate received higher weighting values than topography. This means that local experts attributed a less important role to topography in all cases. Comparing soil and climate CIPs, the weights received were similar (equal values for Flanders and Hengshui and a bit higher values for climate in Somogy), meaning that local experts attributed an almost equal role to soil and climate across regions. Regarding the socio-economic factors, it should be noted that relative information was not available for all regions; thus, to maintain a balanced result, this information was not included in the analysis. However, it is recommended to include socio-economic factors when available.
The limiting factor analysis recognized climate (precipitation) as the main limiting factor in all three test sites. As maize is a crop that requires a lot of water during summer, both in Flanders and Somogy precipitation turned out to be the main limiting factor, implying that productivity depends on the availability of water for supplementary irrigation. Investments in irrigation are not always justified for every soil type, whereas soil and water conservation measures are [
30]. Climate adaptation through land use changes that include less vulnerable crops can compensate for crop yield loss and lead to utility gains [
30].
The prevailing limiting factors in soil CIP were SOC (Somogy, Hengshui, western part of Flanders) and pH in Flanders. Moreover, soil CIP in Flanders had the highest spatial variability for maize, highlighting the usefulness of the proposed approach. Texture affected the northwestern part of Flanders, and CEC the northwestern and southeastern parts of Flanders, while both restricted LS. Within the soil limiting factors, SOC and CEC were the most important (and Texture and CF were less). However, SICS that add organic matter to the soil certainly improve soil conditions. The LSA–SICS analysis showed that an increase in SOC, and especially CRM in Flanders and CT and CC in Somogy, could enhance suitability, contributing to the greenhouse phenomenon.
The prevailing LS class in Flanders and Somogy is the S3 class (marginally suitable), contrary to Hengshui, where most areas were classified as S2 (moderately suitable). Based on the FAO classes identified, the total absence of the S1 class in all regions (and the limited areas classified as S2) indicates that the conditions for cultivating maize in the area can be considered as not optimal, and it is caused by the limiting factors identified in each site. This, however, does not mean that maize should not be cultivated in these sites but that significant resources should be committed to achieve a high yield, or that the highest potential yield may not be harvested.
Comparing results from current LSA and the examined scenarios of the 100-year application of management practices, all three regions showed an improvement in LS by moving several areas classified as S3 into the S2 class (
Table 4). Land suitability analysis was used as a tool to evaluate the change in the most suitable areas (namely S2 class areas because S1 class is absent) based on the current LS. For Flanders, MA was the SICS with the highest impact on LS (16.6%), while even the least impactful scenario (CC) doubled the S2 area (12.6%). For Somogy, the best scenario was CC (32.8%), where the S2 areas were almost quadrupled in size. For Hengshui, all SICS led to completely improving LS, maximizing the S2 areas to fully cover all the region. In addition to quantifying the results, LSA maps provide a visual way to locate the areas most affected.
The percentage increase in LS score after the application of the best SICS was estimated for each site through per pixel comparison of the SICS with the current LS (
Figure 15). SICS implementation in Flanders provided a mild improvement in maize suitability; the upgrade to the S2 class is small, as climatic conditions have a higher weighing factor and lower scores. CRM in Flanders accounted for the biggest overall increment in maize suitability score. SICS implementation in Somogy had a profound enhancement of LS for maize, as more than 50% of the area moved from S3 to the S2 class. CC as well as CT accounted for the biggest overall increment in maize suitability scores. Class separation was marginal though: less than 10% suitability rise led to 50% of cells upgrading classes. SICS implementation in Hengshui had a significant enhancement of LS for maize. CRM accounted for the biggest overall increment in maize suitability, upgrading at least 1/3 of the area to the S2 class by inducing a score rise no bigger than 12%. In all sites, suitability rise appeared substantial in the areas where organic carbon exhibited the lowest suitability score for maize.
The proposed methodology, by providing future scenarios, proves to be a valuable tool in the hands of national and regional authorities for planning and implementing management measures according to the visions of EU soil strategy, which includes improving soil health, achieving climate neutrality, and becoming resilient to climate change [
31]. Moreover, in the frame of the EC Mission Area—Soil Health and Food [
32], the ability to evaluate SICS scenarios contributes to mapping the most prominent measures and soil management practices against their potential contribution towards the goals identified. Finally, the method can facilitate the achievement of Land Degradation Neutrality (LDN) targets [
33], and especially setting LDN baselines, assessing land degradation trends, and identifying measures addressing the drivers of land degradation.