Land Suitability Assessment for Soybean ( Glycine max ( L . ) Merr . ) Production in Kabwe District , Central Zambia

Soybean (Glycine max (L.) Merr.), is a high value crop that can generate income for households. As a legume, soybean is incorporated in cropping systems to improve soil fertility. Soybean productivity is however limited by factors including declined soil fertility, climate change and partly due to inadequate land suitability information. This study aimed at identifying suitable land for soybean production in Kabwe district. Data layers of selected attributes relevant to soybean production were generated with slope and wetness data layers extracted from the digital elevation model (DEM). Elevation was used as a proxy for climate (rainfall and temperature) and was generated by reclassifying the elevation grid into elevation classes. Data layers for soil reaction (pH), soil organic carbon, phosphorus and texture were generated by inverse distance weighting interpolation method based on soil point data. A distance to roads layer was created using the euclidean distance tool. A spatial process model based on multi-criteria evaluation was used to integrate data layers in a weighted sum overlay to generate a soybean suitability map, whose quality was assessed using an error matrix. Results showed that 15.07% of the investigated area was highly suitable for soybean production, whereas 26.53% was suitable and 25.18% was moderately suitable. The other 20.57% was marginally suitable, 10.74% was currently not suitable and 1.92% was permanently not suitable. Based on ground truth data, the overall classification accuracy of the suitability map was 65%. The map was therefore good enough for use as a guide in selecting suitable sites for soybean production.


Introduction
Soybean (Glycine max.(L.) Merr.), is a high value crop whose grain consists 40 percent protein, 20 percent oil and 34 percent carbohydrates.It is used in the production of oil, human food and stock-feed.Soybean is a leguminous plant mostly grown in the temperate and sub-tropical regions (Muliokela, 1995).It is usually incorporated in cropping systems to improve soil fertility and increase yields.This leads to increased crop production, further resulting in increased food security and income generation (Lubungu et al., 2013;Tembo & Sitko, 2013).There is potential for soybean production in Zambia, however productivity is usually limited by many factors that include climate change, decline in soil fertility, low availability of improved seed and low usage of microbial inoculum.Other factors limiting soybean productivity include limited availability of inputs, expertise and market opportunities.In 2014/2015 farming season, the area planted to soybean in Zambia increased by 11.2%, however, its national yield rate declined by 5% due to factors such as the late on set of the rains, poor crop management practices and low soil fertility (NAIS, 2015;MAL, 2016).
To address the challenges of soybean productivity, technologies such as improved seed varieties, crop rotations and tillage practices, inoculation, use of herbicides, pesticides, fungicides and fertilizers have been introduced at most farmers fields.While these technologies are intended to improve soybean productivity, the availability of soil suitability information when selecting production sites is also key for increased productivity (Sullivan, 2003;Adornado & Yoshida, 2008).Land suitability assessment for soybean production is therefore relevant, otherwise implementing new and improved technologies will have no positive impact on soybean production.areas to grow crops as well as identify the main limiting factors for agriculture production.The systems used in assessing land suitability allow for the integration of different criteria that affect suitability.These include soil, rainfall, temperature, topography and other non-biophysical factors such as social, economic and political governance aspects.Land suitability assessment is therefore a complex process and such complexities lead to the simultaneous use of several decision support tools such as Geographic Information Systems (GIS) and multi-criteria decision analysis (MCDA), which are inevitable in determining suitable land for crop production (Malczewski, 2004(Malczewski, , 2006;;Mendas & Delali, 2012).
The process of land suitability analysis for crop production using a GIS based multi-criteria approach first involve the selection of criteria relevant for growing a particular crop.Data layers are generated for each of the selected criteria and thereafter, relative weights are assigned to each data layer.The weighted data layers are used as inputs in a weighted overlay analysis, culminating into land suitability maps (Bolstad, 2005;Perveen et al., 2013).
Traditionally in Zambia, land suitability assessment for crop production is conducted through surveys.The Zambia Semi-Quantified Land Evaluation System for rainfed agriculture is used to assess land suitability.This system incorporates physical and chemical properties of the soil, climatic conditions of the location and the farmer's ability to manage the land for crop production.The system further predicts the potential yield of the crop cultivated (Kalima & Veldkamp, 1985;Veldkamp, 1987;Woode, 1988).According to Chinene (1991), this system is very good for recognizing the limitations of land.Prior to this system, the Land Capability Classification System (LCCS) was used to assess land suitability for rainfed crops, only that it emphasizes soil physical properties more than the chemical status of the soil.A study by Clayton (1974) applied the LCCS in which the extent of the soil suitable for Virginia tobacco and maize were assessed.Woode (1979) also applied the LCCS to assess the suitable area for the rural reconstruction center.Other studies that applied the LCCS include that of English (1982) and Sokotela (1982).
Despite the land evaluation studies that have been done in Zambia, the information in the final reports is not available to most farmers, to assist them in decision making concerning site selection.At times, the land suitability information available does not effectively represent the conditions of most farmers' fields because the suitability assessments were done on a large scale.Farmers are therefore left to plant crops anywhere without considering suitability issues, further leading to low crop productivity.The use of fast methods to map land suitability for crop production is also important considering the effects of climate change on some land qualities (Elsheikh et al., 2013).A GIS based multi-criteria approach in assessing land suitability for soybean production therefore becomes key.Kalima and Veldkamp (1985) also recommended a computerized system in land evaluation to reduce on the number of calculations which were done manually.
This study was therefore aimed at evaluating land suitability for rainfed soybean production in Kabwe district by integrating relevant variables of soil, climate and topography using a GIS based multi-criteria approach.The study aimed at generating a suitability map that could be used as a guide in decision making for selecting suitable sites for soybean production in the area.

Materials and Methods
The study was carried out in Kabwe district located in central Zambia.Attributes relevant for soybean production were identified from the literature search and reviews, and these include slope, drainage, soil reaction, rainfall, temperature, soil organic carbon, phosphorus and texture.Accessibility was also included since it determines the ease of crop movement and market access.
Data layers for slope and drainage (wetness) were extracted from the digital elevation model (DEM) using appropriate algorithms in ArcGIS.Elevation was used as a proxy for climate (rainfall and temperature) and was generated by reclassifying the elevation grid into elevation classes.The distance dataset was generated using the euclidean distance tool.Data sets for soil parameters were generated by inverse distance weighting (IDW) based on soil samples collected from the field.A spatial process model based on multi-criteria evaluation was used to integrate selected spatial attributes in a weighted sum overlay to generate a soybean suitability map.The quality of the suitability map generated was assessed using an error matrix.The following sections describe in detail what was involved at every stage of the methodology.

Description of the Study Area
Kabwe district (Figure 1) is located approximately 130 km north of Lusaka.It lies between Latitude 14 o 28′0″ S and Longitude 028 o 25′5″ E. The district covers an area of about 1,570 km 2 and it is connected to Lusaka and other surrounding districts by rail line and the Great North Road.

Data Sources
Data layers for the selected attributes relevant to soybean production were prepared from secondary and primary data sources.The sources of data sets include topographic maps obtained from Ministry of Lands and Natural Resources (MLNR), the national soil map of Zambia obtained from the Soil Survey Section of the Zambia Agriculture Research Institute (ZARI), and a 90 m resolution Shuttle Radar Topography Mission (SRTM) downloaded from the United States Geological Survey (USGS) website (http://earthexplorer.usgs.gov).The SRTM DEM was covering an area between Latitude 14°25′42″ S and Longitude 028°27′05″ E. Soil samples were randomly collected in the study area as primary data in order to generate data layers for selected soil attributes.A GIS software was used to process the data layers for use in the land suitability process model.

Preprocessing of Digital Elevation Model
The DEM was first loaded into ArcMap and projected to World Geodetic System 1984 (WGS, 1984) UTM Zone 35S coordinates.It was then preprocessed to remove sinks related to imperfections in the data before extracting slope and drainage.A preprocessing method by Chabala et al. (2013) was followed where the flow direction was first extracted using the flow direction tool.A raster of sinks was then created by enabling the sink tool with flow direction used as the input flow direction raster in the table.This was followed by the creation of a sink area raster by enabling the watershed tool.At this stage, the extracted flow direction was the input flow direction raster and the sinks as the feature pour point data or input raster.The zonal statistics tool was then used to create a raster of the minimum elevation in the watershed of each sink, with sink area as the input raster or feature zone data.The DEM was used as the input value raster and sink minimum as the output raster.When creating the sink minimum, statistics type was set to minimum and value as the zone field.The zonal fill tool was then used to create a raster of the maximum elevation with sink area as the input zone raster and the DEM as the input weight raster.Further, the minus tool was then used to subtract the minimum value from the maximum value to find the sink depth, with sink maximum as the first input raster and the sink minimum as second input raster.Finally the fill tool was applied to the DEM with the Z value set to the sink depth to generate a DEM with all the sinks in it filled.

Process Model for Soybean Suitability Mapping
The overall flow of the methodology followed in this study as presented in the process model (Figure 2) involved the selection of attributes relevant to soybean production, generation of data layers, weighting and overlaying of the data layers to generate a land suitability map for soybean production.The data sets for the soil variables were generated with inverse distance weighting interpolation method, based on soil samples collected from the field.Data sets for topographic variables were generated from the DEM using various algorithms in ArcGIS.Elevation dataset was used as proxy for climatic variables (rainfall and temperature) and it was generated by reclassifying the DEM.The distance dataset representing accessibility was generated using the euclidean distance tool.
Each data layer was reclassified into suitability classes based on soybean requirement.The suitability levels assigned to each dataset were defined based on literature, expert knowledge, observation and practical experiences.The suitability levels were ranked as highly suitable, suitable, moderately suitable, marginally suitable, currently not suitable and permanently not suitable based on the suitability classification structure of the Food and Agriculture Organisation [FAO] (FAO, 1976(FAO, , 1983(FAO, , 2007)).The datasets generated were finally combined using the weighted sum overlay to generate the land suitability map for rainfed soybean production.

Dista
In addition affecting l area is to t was create euclidean feature sou to a source from each the roads r

Land C
In order to performed n was tool.
The main limiting factors were identified to help small scale farmers develop crop management practices that will help increase soybean productivity (Rabia, 2012).

Generating the Land Suitability Map for Soybean Production
The reclassified and weighted data layers were combined using the weighted sum overlay tool.Weighted overlay is a method for applying a common scale of values to a number of different input data to create an integrated analysis (Malczewski, 2004).The principle in weighted sum overlay is that, the cell values or rating of suitability class of each input are multiplied by the raster's weight, which is the criteria weights.The resulting cell values are added to produce the final output raster or suitability map.In this study, the resultant of the weighted sum overlay was a land suitability map for soybean production.The map classified the study area into six land suitability classes for soybean production.These classes were identified as highly suitable, suitable, moderately suitable, marginally suitable, currently not suitable and permanently not suitable.

Validation of the Suitability Map
Ground truth data was used to validate the land suitability map.A field visit was undertaken to farms where soybean production is practiced.Information on land suitability and soybean production was gathered by interviewing the farmers in the area while taking note of the physical attributes of each farm visited.The data collected was used to construct an error matrix which was used to calculate the map accuracy by comparing the number of correctly verified points in the field and the predicted data on the generated land suitability map.

Spatial Variation of Selected Attributes Relevant for Mapping Soybean Land Suitability
3.1.1Slope, Drainage and Elevation (Proxy for Climate) The spatial variation showed that 92.62% of the land in Kabwe is almost flat characterized by slope ranging from 0% to 3%.This area was rated suitable for soybean production with regards to slope.Steep slope greater than 5% was also recorded near hilly areas covering 1.45% of the total investigated area (Figure 3).Similar results were obtained by Woode and Mwenda (1988) where the area in the extreme north of Kabwe was found to be steep and therefore recommended land levelling in that area for successful crop production.
Results further showed that, there was less soil moisture in the area lying on higher points which was characterized by high percent slope.This was due to less accumulation.At these points, water usually flow out to other points, therefore soybean production may be hindered in this area.Soybean only requires enough soil moisture to effect growth.In terms of soil moisture, 44.76% of the total area was classified as suitable, moderately suitable and marginally suitable with enough soil moisture to support soybean growth.The area with high accumulation and with no accumulation may get flooded or extremely dry respectively and was therefore classified as not suitable for soybean production.Soybean growth is hindered in this area unless soil moisture management is employed.High accumulation area with wetness index greater than 19.64 was classified as rivers and streams covering 3.40% of the total area (Figure 4).
Four elevation classes were generated according to soybean suitability with 50.99% of the investigated area rated as suitable and 33.66% as moderately suitable for soybean production with regards to weather conditions experienced.Relating elevation to climate (rainfall and temperature), results showed that the area with elevation between 1195 m and 1337 m was associated with severe weather conditions, covering 15.34% of the total area (Figure 5).This area was mostly lying on higher points characterized by hills, and therefore rated as currently not suitable for soybean production.The rest of the area covering 84.66% of the total area was characterized by favorable to moderate weather conditions with elevation ranging between 1110 m and 1195 m.

Soil Attributes
The spatial variation showed that soil reaction was a limiting factor to soybean production with most of the area characterized by soils with low pH values.The soil reaction mapping by Mambo and Phiri (2004) also showed that Kabwe district area is characterized by soils with low pH values.Soybean growth requires soil pH range of 5.6 to 7, and results from this study show that only 0.62% of the investigated area was characterized by soils in this pH range indicating slightly acidic soil.Results further showed that 73.88% of the area was characterized by strongly acidic soils with pH values ranging from 4.51 to 5.0 (Figure 6).This area was rated as marginally suitable for soybean production in terms of soil reaction (pH), requiring application of agricultural lime to improve soybean yields.The other 18.43% of the area was characterized by very strongly acidic soils having pH values ranging from 4.02 to 4.5.This area was rated as currently not suitable for soybean production.
Results further showed that, only 19.55% of the total area was characterized by soils having moderate to high amounts of organic carbon (Figure 7).The rest of the area was characterized by less than 1% of organic carbon in the soil and this may affect the fertility of the soil.It is therefore important that management practices to improve organic matter in the soil are employed to enhance soybean production in Kabwe.
A total of 86.15% of the area was characterized by soils having less than 10.85 ppm of available phosphorus (Figure 8).Similar results were obtained by Sokotela (1982) and Banda et al. (1986) showing that most of the soil in Kabwe is characterized by less amounts of available phosphorus below the critical limit of 10 ppm.Agriculture crops including soybean requires about 10 ppm of phosphorus or more for effective growth.In this study, only 13.85% of the total area was characterized by soils having more than 10.85 ppm of available phosphorus.The lack of phosphorus in crops may prevent other nutrients from being acquired because phosphorus has a stimulating effect on root growth since it is usually concentrated in the root tips of most plants (Linda et al., 2015).
Spatial variation of sol texture showed that Kabwe area was characterized by soils having clay proportions ranging between 9.16% and 51.67% (Figure 9).Generally, the best agriculture soils for crop production are those containing 10 to 20 percent clay.Soybean on the other hand requires deep and well drained soils varying in texture from sandy, sandy loams to clay loams.Dugje et al. (2009) also noted that the best soil for optimum soybean production is a loose, well-drained loam with less clay fractions.
The soil texture identified in the area consists of sandy (S), loam (L), sandy loam (SL) and silty loam (SiL) soils with concentration of clay proportions ranging between 9.16% and 23.17% (Figure 9).These soils were rated as suitable for soybean production, covering 72.07% of the total area.The other area was characterized by soil having clay proportions ranging between 23.18% and 51.67%, rated as moderately suitable for soybean production.This area comprised of medium textured soils, moderately fine textured to fine textured soils and clayey soils, including clay loam (CL), sandy clay loam (SCL), silt loam (SiL) and silt clay (SiC) soils.The soils with more clay content were mostly those distributed around water areas such as on the river banks.
It was also shown that must of the area was accessible with 50.52% of the total area located close to access roads within a distance of 1.53 km.Hence the ease of transportation of inputs and access to the market.

Suitability Analysis for Soybean Production
Based on the weighted attributes, the suitability map for soybean was generated showing the different levels of suitability (Figure 11).Results for the qualitative land suitability analysis showed that 87.35% of the total area was suitable for soybean production, of which 15.07% was highly suitable having no limitation, 26.53% was suitable with minor limitations and 25.18% was moderately suitable with moderate limitations (Figure 11; Table 2).The other 20.57% was marginally suitable having moderately severe limitations.Results further showed that 12.66% of the area was not suitable for soybean production of which 10.74% was currently not suitable with severe limitations and 1.92% was permanently not suitable with extremely severe limitations.
A land capability assessment showed that the area in Kabwe has some limitations owing to different factors including soil reaction, low levels of available phosphorus low SOC, and to a lesser extent owing to slope.This indicated that there was no land that could be classified as highly suitable with no limitations or with insignificant limitations.However, results from this study, showed that 15.07% of the area was highly suitable contrary to the results obtained in a land capability assessment.This was attributed to the fact that only the biophysical factors were considered in land capability assessment while suitability assessment included accessibility as a non-biophysical factor.The non-biophysical factors such as the social economic factors have an influence on the suitability of an area for a particular utilization, and are therefore considered in suitability mapping (FAO, 1976;Grose, 1999).

Conclusion
Based on the findings of this study, it can be concluded that Kabwe has great potential for soybean production.
The generated land suitability map, showed that 87.35% of the total area in Kabwe is suitable for soybean production.The other 12.66% is not suitable owing to different limiting factors.The main limitations to soybean production in Kabwe as identified in the land capability analysis are soil reaction and available phosphorus.Therefore, application of phosphate fertilizers to address low available phosphorus and agriculture lime to correct soil acidity is necessary to enhance soybean production in the area.It can further be concluded that the generated land suitability map with overall classification accuracy of 65% can rightly be used as a guide in decision making for soya bean production.This study therefore demonstrated that a GIS based multi-criteria analysis can be considered as an important tool for delineating land suitability for soybean production.
It is recommended that future studies can apply this method for mapping land suitability with addition of other parameters that were not included in this study.The method can also be used to map land suitability for other crops such as cereals.Field trials can also be set up in the various suitability classes as identified in the study.This will allow for testing of the various management options for optimal soybean production in each of the suitability classes.

Figure
Figure 2. Pro

Table 2 .
A