Keywords

1 Introduction

India possesses 2.4% of world’s geographical area, 4% of world’s renewable water resources and holds 18% of world’s population [1]. India’s total food grain production during 1999–2000 was 208MT, during 2006–2007 it was 209.2 MT and it needs to be raised to 350 MT by 2050 [2, 3] to feed the expected population of 1390M. On the contrary, since 1995–1996, the average size land holding has decreased from 1.41 hectares to 1.15 hectares which accounts for a decrease in cultivable land at the rate of 30,000 ha/year. With an increase in population, decrease in percentage of farmers and with problems like drought and floods in various states, one can fathom the seriousness of the issue [4].

Madhya Pradesh (M.P.) is the second largest state of India. Its population growth rate is 20.4%, higher than that of India (17.6%) [5]. According to Directorate of Census Operation of M.P., District Dewas has tenth largest urban population and fourteenth largest urban area of M.P. [6]. Lowering groundwater levels, excessive fluoride content in groundwater, late commencing and erratic rainfall are regular scenes in Dewas. According to Central Ground Water Board’s Aquifer Mapping Report, only 50% of district’s net sown area is actually irrigated [7]. According to District Census Handbook, Dewas (2011), there has been 19.53% increase in its population between 2001 and 2011 [8]. The rate and the extent of urbanization in Dewas may challenge its agriculture scene in future.

A land regime once urbanized cannot be reclaimed again for agriculture. Post urbanization, agriculture can only be practiced on what is left open, irrespective of its appropriateness. Wastelands in India assume 67 Mha of the land use pattern, out of this cultivable wasteland constitute 20%. Clearly, there is a mismatch between land aptitude and land use. Land use should be in conjunction with its inherent capability to achieve long sustained agricultural productivity. In order to increase crop output and achieve food security, crops must be grown in areas where they are best suited. To generate the maximum agricultural production, land regimes must be utilized in an environmentally compassionate, socially justified and economically feasible manner. It is often observed that the land is either overused or underused or improperly used without examining its potential and constraints for that intended purpose. In order to increase/maintain the productivity per land unit, crops must be grown in areas where they are best suited. Remote Sensing and GIS-based Land Suitability Assessments (LSAs) should be performed before every major land-use-related decision to assure respectable use of natural resources and to control any unsustainable shifts and transformations.

LSA is an inter-disciplinary approach defined by FAO [9] as ‘The analysis of land performance when used for a specified purpose, involving the execution and interpretation of anatomization-based surveys and studies of land forms, soils, vegetation, climate and other characteristics of land in order to identify and make a comparison of promising kinds of land use in terms of suitability to the objectives of the evaluation’. Land is categorized into spatially distributed agriculture potential zones based on soil and terrain characteristics and hydroclimatic constraints [10].

Through Analytical Hierarchy Process (AHP), dissimilar inputs from these parameters are combined and processed to form a single index of evaluation [11, 12]. AHP can integrate huge amount of heterogeneous data and deduce the relative weights of the parameters [13,14,15] facilitating the selection of best alternative amongst multiple options [16]. In this study, we combine the concept of AHP with geospatial modelling and create a decision support system to classify the district’s land regimes and identify the potential land parcels for agriculture. In order to stop conversion of agriculture land for non-agricultural purposes, the government has formulated the National Policy for Farmers 2007 (NPF 2007) [17] and the National Rehabilitation and Resettlement Policy 2007 (NRRP 2007) [18]. The present study aims to make the implementation of these policies easier and more accurate.

2 Study Area

Dewas lies between 20°17′N and 23°20′N latitude and 75°54′E and 77°08′E longitude, falling under Survey of India Toposheets 46M, 46N, 55A, 55B and 55F and UTM zone WGS1984 43N [19]. Dewas covers an area of 7020.84 km2 with a population of 1,563,715 [8]. It is situated at an altitude of 555.5 m above mean sea level. It falls under Agro-Ecological Sub-Region 5.2 (hot semi-arid moist eco-region) [20]. The district has six administrative blocks: Tonk Khurd, Dewas, Sonkach, Kannod, Khategaon and Bagli as shown in Fig. 1. There are three national highways (NH-3, NH-86, NH-59A). The average annual rainfall is 1083 mm. Narmada, Kali Sindh and Kshipra are the three main rivers. There are four physiographic regions, such as Dewas Plateau, Vindhyan Range, Kali Sindh Basin and Middle Narmada Valley. Hydrogeology is largely Basalt. Black cotton soil is the major soil type. Groundwater contains alkaline earth–bicarbonate. There are two major drainage basins: The Ganga (North) and The Narmada (South). Groundwater is ‘alkaline earth–bicarbonate (Sangemini)’ and accounts for 82% of irrigation. The principal crops are soya bean, wheat, groundnut and cotton [21].

Fig. 1
figure 1

Study area

Full size image

3 Methodology

3.1 Preparation of Parameter Maps

The District Administrative Boundary shapefile is obtained from DivaGIS [22]. The units of the parameters are shown separately in Table 4. All the maps are projected (geometric correction) to World Geodetic System (WGS) 1984 datum and Universal Transverse Mercator (UTM) Zone 43N (Fig. 2).

Fig. 2
figure 2

Flow chart of land suitability assessment methodology

Full size image

The digital elevation model given in Fig. 3 is prepared using ASTER GDEM V2 product [23] for 21 January 2018. The DEM is a [1 × 1] degree tile and has a spatial resolution of 30 m. The slope map given in Fig. 4 is derived from the DEM using ‘Slope’ Spatial Analyst tool.

Fig. 3
figure 3

Digital elevation model

Full size image
Fig. 4
figure 4

Slope map

Full size image

The drainage density map is prepared using ‘ArcHydro Tools’ extension [24] for ArcGIS. Consecutive terrain processing operations, fill sink < flow direction < flow accumulation < stream definition < stream segmentation < stream order, are performed on the DEM to obtain the drainage map given in Fig. 5. ‘Line Density’ tool is applied on the drainage map to obtain drainage density map given in Fig. 6. Efficient drainage prevents cases of waterlogging or excessive infiltration.

Fig. 5
figure 5

Drainage map

Full size image
Fig. 6
figure 6

Drainage density map

Full size image

The multi-temporal (*MTL.txt) Landsat-8 OLI+ image for 26 February 2018 (Path: 146/Row: 44) given in Fig. 7 obtained from USGS Earth Explorer [25] is subjected to automatic cloud detection and haze removal (radiometric correction) using PCI Geomatica. Iso-cluster unsupervised classification is performed in ArcGIS to prepare the land use/land cover map given in Fig. 8 which is then verified from Google Earth [26].

Fig. 7
figure 7

Landsat-8 OLI+ FCC image

Full size image
Fig. 8
figure 8

Land use/land cover map

Full size image

Net Differential Vegetation Index (NDVI). Map shown in Fig. 9 is prepared from Landsat-8 Band-4 and Band-5 in ‘Raster Calculator’ as shown in Eq. 1.

Fig. 9
figure 9

NDVI map

Full size image
$$\, {{\left( {{\text{Band}}_{5} - {\text{Band}}_{4} } \right)} \mathord{\left/ {\vphantom {{\left( {{\text{Band}}_{5} - {\text{Band}}_{4} } \right)} {\left( {{\text{Band}}_{5} + {\text{Band}}_{4} } \right)}}} \right. \kern-0pt} {\left( {{\text{Band}}_{5} + {\text{Band}}_{4} } \right)}}$$
(1)

NDVI is used for estimating vegetation health from remote sensing data. NDVI is a numerical indicator that uses red and NIR bands of the electromagnetic spectrum to detect the regions of live green vegetation (regions with greener vegetation are good for agriculture practices). According to USGS’s band designations for Landsat-8 OLI+ satellite image given in Table 1, healthy vegetation (chlorophyll) reflects near-infrared light and absorbs red light.

Table 1 Band designations for Landsat-8 OLI+ satellite
Full size table

Normalized Difference Water Index (NDWI). Map shown in Fig. 10 is prepared from Landsat-8 Band-5 and Band-6 in ‘Raster Calculator’ as shown in Eq. 2

Fig. 10
figure 10

NDWI map

Full size image
$${{\left( {{\text{Band}}_{6} {-}{\text{Band}}_{5} } \right)} \mathord{\left/ {\vphantom {{\left( {{\text{Band}}_{6} {-}{\text{Band}}_{5} } \right)} {\left( {{\text{Band}}_{6} + {\text{Band}}_{5} } \right)}}} \right. \kern-0pt} {\left( {{\text{Band}}_{6} + {\text{Band}}_{5} } \right)}} \,$$
(2)

NDWI detects the changes in liquid water content of spongy mesophyll of vegetation canopies. SWIR shows changes in the water content of mesophyll, whereas NIR shows changes in leaf internal structure and its dry matter content, and hence the equation.

The road proximity map shown in Fig. 11 is developed from the road network shapefile using ‘Euclidean Distance’ Spatial Analyst Tool. The road network shows a line layer of road systems like national highways, state highways and major roads. The river proximity map shown in Fig. 12 and well proximity map shown in Fig. 13 are prepared in the same manner and show a line layer of river systems (Narmada, Kali Sindh and Kshipra) and a point layer of wells, respectively. The road shapefile and river shapefile are obtained from OpenStreetMap via Geofabrik [28]. The proximity maps are verified from 1:50,000 Survey of India (SOI) Toposheets (46M, 46N, 55A, 55B and 55F) [29].

Fig. 11
figure 11

Road proximity map

Full size image
Fig. 12
figure 12

River proximity map

Full size image
Fig. 13
figure 13

Well proximity map

Full size image

Soil Cation Exchange Capacity (CEC) map shown in Fig. 14, Soil Organic Carbon Content (OCC) map shown in Fig. 15, Soil pH map shown in Fig. 16 and Soil Depth map shown in Fig. 17 were available from ISRIC SoilGrids [30]. CEC is the measure of negatively charged content present in soil that holds ions like calcium (Ca2+), magnesium (Mg2+) and potassium (K+) through electrostatic forces and stops them from being leached down the soil profile. It is the nutrient retention capacity of the roots [31].

Fig. 14
figure 14

Soil cation exchange capacity map

Full size image
Fig. 15
figure 15

Organic soil carbon content map

Full size image
Fig. 16
figure 16

Soil pH map

Full size image
Fig. 17
figure 17

Soil depth map

Full size image

OCC is a direct measure of soil fertility. OCC affects water holding capacity, nutrient availability and maintains soil structure. Soil pH is the degree of acidity/alkalinity of soil. For majority of crops, the optimal range of pH is from 5.8 to 6.8. Crops rarely survive extreme pH environments. Soil depth basically depends on the nature of the parent material on which soils were formed. For a well-drained friable soil, the maximum root depth of crops can be 200 cm. Lower depth of soil restricts plant rooting, hinders overall growth and ultimately the yield.

Soil Electrical Conductivity (EC) map shown in Fig. 18 is prepared using the remote sensing data of 202 soil sampling sites obtained from KrishiKosh [32]. The data is imported into ArcGIS as a point shapefile, followed by subjection to Inverse Distance Weighted (IDW) Interpolation tool. EC is a direct measure of salinity of soil. Plants are detrimentally affected, both physically and chemically, by excess salts in some soils and by high levels of exchangeable sodium in others. EC depends on the amount of moisture held by soil particles [33].

Fig. 18
figure 18

Soil electrical conductivity map

Full size image

The location of wells and groundwater level data are obtained from India-WRIS [34] and CGWB’s Aquifer Mapping Report [7]. A point shapefile containing the coordinates of the wells and their respective water level depths is created. The groundwater depth map given in Fig. 19 is prepared using IDW tool. The annual average rainfall map given in Fig. 20 is prepared in the same manner using rainfall data (2004–2014) from Global Weather Data for SWAT [35].

Fig. 19
figure 19

Groundwater level map

Full size image
Fig. 20
figure 20

Annual rainfall distribution map

Full size image

The soil texture map given in Fig. 21 is prepared in QGIS Desktop 2.18 using ‘Soil Texture’ plug-in [36]. Soil texture is determined by the relative proportions of sand (0.05–2 mm), silt (0.002–0.05 mm) and clay (<0.002 mm) [37]. The inputs of sand map and clay map are given and the output is generated based on the rules of soil texture triangle. The geology map given in Fig. 22 is digitized from its photo image [7] using ArcScan tool and converted to raster format using ‘Polyline to Raster’ conversion tool.

Fig. 21
figure 21

Soil texture map

Full size image
Fig. 22
figure 22

Geology map

Full size image

3.2 Analytical Hierarchy Process

To assign a relative degree of importance (weight) to every parameter based on a fundamental scale [14], AHP priority calculator is used [38]. The scale ranges from 1 (equal significance) to 9 (extreme significance) as shown in Table 2. Comparisons are performed for different pairings of 18 parameters using Pairwise Comparison Matrix (PWCM) as shown in Table 3. The total number of pairwise comparisons is calculated as shown in Eq. 3:

Table 2 The fundamental scale for pairwise comparison matrix [14]
Full size table
Table 3 Pairwise comparison matrix for AHP-MCDM
Full size table
$${\text{PC}} = \left( {n\left( {n - 1} \right)/2} \right)$$
(3)

where ‘n’ is the number of parameters. Hence, for n = 18, PC = 153. Consistency ratio, an indicator of the degree of consistency/efficiency of AHP analysis, is calculated as shown in Eq. 4:

$${\text{CR}} = \left( {{{\text{CI}} \mathord{\left/ {\vphantom {{\text{CI}} {\text{RI}}}} \right. \kern-0pt} {\text{RI}}}} \right)$$
(4)

where RI is the random index. RI depends on the order of the matrix as shown in Table 4 [13]. For n = 18, RI is 1.62 [14]. CI is the consistency index. It is expressed as shown in Eq. 5:

Table 4 Average RI for corresponding matrix size [14, 43, 44]
Full size table
$${\text{CI}} = {{\left( {\lambda_{\hbox{max} } {-} \, n} \right)} \mathord{\left/ {\vphantom {{\left( {\lambda_{\hbox{max} } {-} \, n} \right)} {\left( {n \, {-} \, 1} \right)}}} \right. \kern-0pt} {\left( {n \, {-} \, 1} \right)}}$$
(5)

where ‘λmax’ is the principle eigenvector and ‘n’ is the order of the matrix [12]. If CR < 0.10, it means that an acceptable level of consistency is achieved [13], the weights calculated are meaningful and valid [39]. In this case, CR is 0.04. The synthesized matrix is shown in Table 5. The ranking of parameters is given in Table 6. The results of AHP analysis are given in Table 7.

Table 5 Synthesized matrix for AHP-MCDM
Full size table
Table 6 Ranking of parameters
Full size table
Table 7 Results of AHP analysis
Full size table

AHP-MCE calculates the ‘weight’ of every parameter by calculating the eigenvalue corresponding to the highest eigenvector of the matrix and normalizing the sum of all the parameters to unity [40, 41]. In this case, eigenvector solution is obtained after six iterations. The weights, hence, computed from AHP analysis are converted into percent for further processing in Weighted Overlay Analysis (WOA) in ArcGIS ModelBuilder. ModelBuilder is a dynamic, interactive and heavily customizable geoprocessing application of ArcGIS. It is used for data integration and data analysis. It can be regarded as a visual programming language for generating workflows. ModelBuilder is used to create, edit and manage custom-based models for a diverse stream of applications. Models are automated GIS workflows that string together sequences of geoprocessing tools that need GIS data as the initial input [42].

3.3 Weighted Overlay Analysis

Reclassification/Standardization: Initially, all the parameters have diverse and dissimilar units. Standardization converts the measurement to uniform units [39] for further processing in WOA. The parameters lose their dimensions [45]. The sub-criteria are brought to a common suitability scale ranging from 1 to 10 as shown in Table 8 in order to eliminate their inherent vagueness.

Table 8 Weights of main criteria and scores of sub-criteria
Full size table

Overlaying: The reclassified raster layers are overlaid by multiplying the values of their sub-criteria by their corresponding weight value [40] as shown in Eq. 6. The combined weight of all input maps should be

$${\text{LSS}} = \sum\limits_{i = 1}^{n} {W_{i} X_{i} }$$
(6)

where LSS = l and suitability score, Wi = weight of the suitability parameter, Xi = score of the sub-criteria of the suitability parameter, n = number of suitability parameters.

Addition: All the resulting cell values are finally added. The area falling under each suitability class is determined using ‘Zonal Geometry as Table’ Spatial Analyst Tool.

4 Results and Discussions

The land suitability assessment model is given in Fig. 23. The final land suitability map for agriculture is given in Fig. 24. The analysis designates 17.8% (1,24,970 ha) area as highly suitable, 57.3% (4,02,294 ha) area as moderately suitable and 24.9% (1,74,818 ha) area as marginally suitable for agriculture. A niche-based multiparametric LSA was performed in accordance with FAO (1976). The assessment faced internal variability and complexity with abundance of highly localized features. Weighted overlay analysis allowed strategic delineation of suitability regimes by integrating the diverse spatial layers encompassing 18 different environmental themes, terrain specifics and hydroclimatic vistas. The relative degree of importance between parameters was realized after considering the subjective opinions of farmers and local experts which made the results more original at grassroot level. The study benefitted from online availability of remote sensing data.

Fig. 23
figure 23

Land suitability assessment model

Full size image
Fig. 24
figure 24

Land suitability map for agriculture

Full size image

5 Conclusion

The aim of this research was to create a comprehensive guideline integrating AHP with ArcGIS ModelBuilder that can classify the district into land parcels based on their suitability for agriculture. In highly suitable areas, large-scale agriculture can be practiced and appreciable production can be expected. In moderately suitable areas, irrigation method should be chosen carefully and soil management must be performed. In marginally suitable areas, farming will be risky and should be limited. The results of this study help us identify the inherent potentials and constraints of land for agriculture whilst accounting for the rigidity of the existing land use pattern.