Agricultural Systems 196 (2022) 103343

Contents lists available at ScienceDirect

Agricultural Systems
journal homepage: www.elsevier.com/locate/agsy

Coupling geographic information system integrated fuzzy logic-analytical

hierarchy process with global and machine learning based sensitivity
analysis for agricultural suitability mapping
Swapan Talukdar a, b, Mohd Waseem Naikoo b, Javed Mallick c, Bushra Praveen d, Shahfahad b,
Pritee Sharma d, Abu Reza Md. Towfiqul Islam e, Swades Pal f, Atiqur Rahman b, *
a
Department of Geography, University of Gour Banga, Mokdumpur, Malda, West Bengal, India
b
Department of Geography, Faculty of Natural Sciences, Jamia Millia Islamia, New Delhi 110025, India
c
Department of Civil Engineering, College of Engineering, King Khalid University, Abha, Saudi Arabia
d
School of Humanities and Social Sciences, Indian Institute of Technology Indore, Simrol, Indore 453552, India
e
Department of Disaster management, Begum Rokeya University, Rangpur 5400, Bangladesh
f
Department of Geography, University of GourBanga, Mokdumpur, Malda, West Bengal 732103, India

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• Agriculture land suitability (ALS) map-

ping for the country has not been yet
done due to data scarcity and lack of
technology.
• e developed and applied fuzzy-AHP and
fuzzy logic-based ALS models, as well as
several sensitivity analyses.
• 20.2% and 19.78% area were identified
as highly suitable and permanently un-
suitable for agriculture.
• The integrated fuzzy-AHP based ALS
model was found to be the best model
for ALS mapping.
• A robust ALS model could aid agricul-
tural decisions that support smart agri-
cultural practices with greater
production.

A R T I C L E I N F O A B S T R A C T

Editor: Dr. Jagadish Timsina CONTEXT: India’s increasing population growth and unsystematic land cover transformation have led to land
degradation and a decline in agricultural production. To achieve optimum advantage from the land, proper
Keywords: exploitation of its resources is necessary. Remote sensing, advanced fuzzy logic, and multi-criteria decision-
Agricultural suitability model making like analytical hierarchy process (AHP) integrated agricultural land suitability analysis (ALAS) may
Fuzzy logic
facilitate identifying and formulating effective agricultural management strategies required for smart agriculture.
Analytical hierarchy process
OBJECTIVES: The present study was conducted to construct India’s robust agricultural suitability model by
Remote sensing
Sensitivity analysis developing hybrid fuzzy logic and the AHP based model.
Machine learning algorithms METHODS: Fourteen topographical, climatological, soil-related, land-use, and land-cover-related factors were
prepared and employed to model agricultural suitability. Agricultural suitability models predicted multi-

* Corresponding author.
E-mail addresses: jmallick@kku.edu.sa (J. Mallick), psharma@iiti.ac.in (P. Sharma), towfiq_dm@brur.ac.bd (A.R.Md.T. Islam), arahman2@jmi.ac.in (A. Rahman).

https://doi.org/10.1016/j.agsy.2021.103343
Received 12 June 2021; Received in revised form 29 November 2021; Accepted 30 November 2021
Available online 8 December 2021
0308-521X/© 2021 Published by Elsevier Ltd.
S. Talukdar et al. Agricultural Systems 196 (2022) 103343

parameters based agricultural suitable zones for the entire country using three fuzzy operators (AND, Gamma
0.8, Gamma 0.9) and a hybrid fuzzy-AHP model. Sensitivity analysis was conducted to test the models’ reliability
using Moris technique-based global sensitivity analysis, random forest (RF), and correlation coefficient. The best
agricultural suitable model was compared with the production of major crops in India.
RESULTS AND CONCLUSIONS: Results showed that 19.8% of the study area was permanently not suitable in the
northernmost region, 19.7% was currently not suitable in the northernmost region, while 20.1% and 20.2% areas
were predicted as moderately suitable and highly suitable zones, respectively. The rainfall, elevation, slopes,
evapotranspiration, and aridity index had a prime influence on the output of the agricultural suitability model.
SIGNIFICANCE: The adopted method and its application processes can analyze agricultural land suitability and
recommend optimal farming methods. It is also comprehended as a promising option for meeting food, nutrition,
energy, and job demands while still protecting our threatened environment.

1. Introduction Choosing a precise and efficient algorithm for determining agricul-

Agricultural food products remain in perpetual demand (OECD/FAO,
2019) because of the increasing population (Sands et al., 2014;
Mozumdar, 2012), rapid urbanization and urban growth (UN, 2018),
declining productivity of the agricultural land (Sands et al., 2020),
climate change (Fukase and Martin, 2020; Talukder et al., 2020;
Anderson et al., 2020; Leisner, 2020). As a result, world food demand is
projected to rise by 70% by 2050 (Bocchiola et al., 2019; Schneider
et al., 2011). This rising demand for agricultural goods has resulted in
the exhaustion of global land resources in recent decades (Lambin and
Meyfroidt, 2011), causing not only agro-ecological problems (Hath-
away, 2016) but also jeopardizes agricultural sustainability (Hunter
et al., 2017). Planning through agricultural suitability assessment is of
paramount importance to address these issues and optimally utilize the
land resources (Song and Zhang, 2021; Ahmed et al., 2016: Yohannes
and Soromessa, 2018).
Land suitability appraisal is the evaluation and aggregation of the
suitability of particular areas of land for defined uses (Liu et al., 2006). It
is a tool for deciding the factors that inhibit a crop from growing
(Halder, 2013; Chozom and Nimasow, 2021). Land suitability evalua-
tion involves qualitative valuations of topography, vegetation, climate,
hydrology, and soil properties and quantitative valuations that rely on
yield estimates (Mosleh et al., 2017; El Baroudy, 2016). Typically, this
land suitability appraisal is performed separately for each crop type
(Herzberg et al., 2019).
One of the most fundamental aspects of the agricultural suitability
evaluation process is the selection of criteria (Tercan and Dereli, 2020;
Zolekar and Bhagat, 2015). For agricultural suitability zonation, Pilevar
et al. (2020) used climatic factors such as temperature, topographic
factors such as slope and elevation, and soil characteristics such as soil
texture, soil PH, and electric conductivity. However, Seng et al. (2009)
showed that soil’s alkalinity, acidity, water storage profile, and water-
logging characteristics are essential for mapping agricultural suit-
ability. Similarly, Akıncı et al. (2013) show that the soil classification
category, land capacity level and subclass, height, slope, soil density,
rockiness, and stoniness are essential factors to consider when assessing
land suitability. Seyedmohammadi et al. (2019) have used climatic
characteristics such as mean daily maximum and minimum tempera-
tures for the coldest month, as well as the mean temperature at various
stages of crop development; soil characteristics such as depth, gypsum,
and calcium carbonate content, pH, electrical conductivity, exchange-
able sodium percentage, and topographic characteristics of slope for
agricultural suitability. Finally, for Land suitability evaluation, Sahoo
et al. (2018) consider various geological and hydrometeorological
characteristics, such as rainfall, ET, NDVI, LULC, soil, soil moisture,
groundwater level, geology, slope, and elevation. The present study
employs climatic characteristics (rainfall, temperature, wind speed, ET,
and aridity), topographic characteristics (slope, aspect, elevation, and
topographic roughness index), soil characteristics (soil quality, soil
composition, soil erosion, and the amount of soil organic carbon), and
finally, LULC parameters, considering the findings of the previous
research to determine land suitability.
tural land suitability analysis (ALSA) significantly influences present
and future land use planning (Gardner et al., 2021; Vasu et al., 2018).
This research categorizes ALSA as the modern system (De la Rosa and
Van Diepen, 2002). Conventional ALSA approaches are qualitative,
quantitative, and parametric, primarily based on several edaphic bio-
physical variables. Modern ALSA frameworks provide novel approaches
for dealing with MCE, GIS, and remote sensing for vast amounts of data
and Decision Support Systems (DSS) (Gardner et al., 2021; Gilliams
et al., 2005) the ALSA operations are incorporated. GIS-based land-use
suitability analysis can be classified into three groups, as per Malczewski
(2004): (i) computer-assisted overlay modeling, (ii) multi-criteria
assessment methods, and (iii) soft computing or geo-computation, also
known as Artificial Intelligence (AI) methods. GIS-based MCE could be
described as combining and transforming spatial data or input into a
decision map or output (Malczewski, 1999). The MCE procedures pro-
vide a link between the input and output maps. GIS-MCE could show a
rational and objective approach for deciding in agriculture (Feizizadeh
and Blaschke, 2013). The procedures include using spatial data, the
decision maker’s interests and data preference manipulation according
to established decision rules (Akıncı et al., 2013).
In addition, combining MCE techniques with artificial intelligence
techniques, such as combining AHP and fuzzy logic, can increase
modeling efficiency because MCE and AHP techniques have flaws,
particularly with pair-wise comparisons, where certainty is lacking
(Huang et al., 2008). Even though it integrates expert intelligence, the
AHP approach does not depict human insight. In order to overcome
these shortcomings, the FAHP approach was proposed. Fuzzy logic ap-
plications are constantly expanding (Özkan et al., 2020), and its
convergence with AHP is proving to be an effective technique with
improved precision for land suitability evaluations (Tashayo et al.,
2020). The approach has been applied to a variety of issues, including
site selection (Ertuğrul and Karakaşoğlu, 2008), tourism management
(Wang et al., 2013), machine-tool selection (Durán and Aguilo, 2008),
supply chain management (Jakhar and Barua, 2014), arms selection
(Dağdeviren et al., 2009) and energy systems management (Durán and
Aguilo, 2008). FAHP methods have been used in research, but only a few
reports have documented the method’s use in agricultural suitability
modeling (Salvacion, 2019; Tashayo et al., 2020; Nabati et al., 2020).
Therefore, in the present study, different operators of fuzzy logic and
MCE integrated fuzzy logic (AHP-fuzzy logic) were performed to map
the agriculture land suitability mapping.
Previous studies show that much researchers have been carried out
on the land suitability assessment using statistical models in different
parts of the world. In addition, we have introduced the FAHP approach
and its reliability tests to model agriculture suitability at the national-
level. Therefore, some studies in India have been conducted in partic-
ular regions, but very few studies have considered national-level
modeling. Therefore, this is the novelty of the present paper. Agricul-
ture suitability mapping has also not been carried out in India. Because
of the research void, the present study aims to plan agriculture suit-
ability mapping for the entire region. The objectives are to (i) model
agricultural suitability for entire India by integrating fuzzy AHP model

2
S. Talukdar et al. Agricultural Systems 196 (2022) 103343

and several fuzzy operators, and (ii) test the reliability and robustness criteria for agricultural land suitability in this analysis. We derived the
using global sensitivity analysis based on the Morris technique. The above parameters from various sources with varying resolutions. Table 1
significant novelties are: contains detailed information on the data sources. Using resampling
techniques eliminated the resolution effects. We employed a variety of
• General: The work increases the robustness of expertise by designing resampling approaches in this work, including nearest neighbor and
and applying methods to a previously unstudied field on agriculture cubic convolution. The nearest neighbor approach has been employed
suitability mapping for the entire country. for discrete datasets, such as land use, land cover, soil types, and nutrient
• Regional: Agriculture suitability mapping has not yet been quality. It is used for discrete datasets since the values of the cells will
completed in the entire country. As agriculture is a significant pillar not change. In the present research, cubic convolution was utilized for
of the Indian economy, this work is important in proposing man-
agement plans. The results of this work will be a useful framework to
enhance land and agricultural management for earth scientists, Table 1
government authorities, and stakeholders. Details of data sources of the parameters used for agriculture suitability
• Methodical: Several fuzzy operators, such as AND, OR, and GAMMA, modeling.
were used in the present study, along with integrated AHP-based Data types Sources Resolution
fuzzy logic for agriculture suitability mapping, which has not yet Rainfall IMD –
been extended to agriculture land suitability mapping. Even, time- Temperature IMD –
variant agricultural suitability model has been generated with AHP Wind speed
Giovanni (https://giovanni.gsfc.nasa.gov/gio
1◦
based fuzzy logic model, which has not been studied so far to explore vanni/)
Giovanni (https://giovanni.gsfc.nasa.gov/gio
the impact of time-variant parameters like rainfall and temperature. Evapotranspiration
vanni/)
1◦
Also, we validated the agricultural suitability models with proxy data Soil organic carbon stock in t/ha for 0-30 cm
like gridded major crop production data, which has not been studied. Soil organic carbon depth intervals. (https://maps.isric.org/maps 250 m
Furthermore, to the best of the authors’ knowledge, this is the first erv?map=/map/ocs.map)
Reference soil group (2006), soil grid (htt
research in which sensitivity analysis using the Morris approach was Soil Types –
ps://soilgrids.org/)
used to explore the relationship between parameters and models. Soil Nutrient
Fischer et al., 2008 250 m
Qualities
2. Data sources and methods Soil erosion ISRIC, world soil information 250 m
DEM DIVA GIS 90 m
Aridity ISRIC, world soil information 250 m
2.1. Data sources and rationale for selecting the data Oak Ridge National Laboratory (ORNL)
Distributed Active Archive Center (DAAC)
Land use land cover 100 m
This study was conducted using data from all states of India (Fig. 1). (https://daac.ornl.gov/cgi-bin/dsviewer.pl?
Rainfall, temperature, wind speed, evapotranspiration, soil organic ds_id=1336)

carbon, soil types, soil nutrient qualities, gross soil erosion, elevation,
aspect, slope, aridity index, and land use land cover were chosen as

Fig. 1. Location of the study area, India, and distribution of the thirty-four meteorological sub-division.

3
S. Talukdar et al.
continuous datasets because it calculates a cell’s new value by fitting a
smooth curve through the 16 closest input cell centers. It has less geo-
metric distortion than the raster produced by the nearest neighbor
resampling method. We utilized elevation as an essential parameter with
a lower resolution (90 m) for resampling and converted all values to 90
m. Different resolution effects have been removed in this way. In this
analysis, all parameters were resampled at 100 m, such as land use and
land cover.
The method for determining land suitability has been extended by
emphasizing many essential factors that affect crop production. A
multitude of variables must be considered in order to produce accurate
and robust, sustainable agriculture management. Therefore, to help
policies and land use planning strategies for adequate crop production,
various factors are integrated. As a result, a suitability research tech-
nique was designed to distinguish each conditioning factor based on the
associated positive and negative effects in terms of crop yield. Climate,
topography, and soil characteristics were defined in the current study
based on their effect on crop yield. Given these factors, the suitability of
land was achieved in agricultural mapping. The reasons for selecting the
variables for agricultural land suitability modeling are detailed below:
Climate change has had a more negative effect on crop production
than positive effects across a large variety of regions, according to the
Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment
Report (AR5) (IPCC, 2014). Climate change would significantly affect
water availability and supply by altering rainfall, evaporation, runoff,
soil moisture storage, and causing significant temperature volatility
(Olesen and Bindi, 2002). When the average precipitation is somewhat
smaller or higher than the optimal, major problems may occur from
drowning to poorer productivity. Aside from the impact of high and low
rainfall on cropping trends, events such as coastal floods are expected to
decrease the amount of land suitable for cultivation. Farmers are still
struggling to respond to these environmental changes, although nearly
all crops are seasonal and dependent on rainfall (Singh et al., 2014).
Droughts, floods, erratic precipitation patterns, heat waves, and other
severe events have risen because of the extreme rise in temperature
worldwide. According to a study released by the United Nations Envi-
ronment Programme in 2017, drought and desertification have resulted
in the abandonment of 500 million hectares of farmland (Arora, 2019).
The direction and velocity of the wind have a significant impact on crop
development. As a result, decent crop yields cause optimum wind speed
(Zabihi et al., 2015). Evapotranspiration is an essential element of the
soil water equilibrium, and it plays a significant role in assessing agri-
cultural yield potential. Crop productivity will likely decline as the crop
growth cycle is shortened because of rising temperatures, lack of mois-
ture, and soil water scarcity (Bhatt and Hossain, 2019). Therefore, to
propose a reliable land suitability model for intelligent agriculture
practice, it is necessary to know the spatial pattern of
evapotranspiration.
The evaluation of topographical criteria offers knowledge about the
land’s limits for agricultural development. Topographic indicators, such
as elevation, slope, aspect, index of topographical ruggedness, are
important factors that affect the crop substantially. Elevation variations
have a bearing on soils, microclimatic impacts, and other processes that
can influence land suitability (Yi and Wang, 2013). Food crop yield was
inversely proportional to elevation (Minda et al., 2018). Crops and
pulses such as rice, jute, wheat, and maize could be at risk as the altitude
rises since low altitude areas are suitable for these food and commercial
crops. While few fruits are suitable for high altitude, plain regions are
widely recommended for food and commercial crops to feed a large
population. Since the degree of sunlight intensity is influenced by the
aspect, the southern and western aspects are commonly considered the
most suitable for agriculture (Akıncı et al., 2013). The slope has a major
impact on plant structure as well as soil erosion. The slope is essential for
both the surface and the internal soil water drainage since both features
perform a significant role in crop growth (Yi and Wang, 2013). For these
reasons, topographical criteria should be taken into account when
4
Agricultural Systems 196 (2022) 103343
developing innovative agriculture management plans.
For agricultural development and long-term soil usage, the physical
properties of the soil are critical. The ability of the roots to absorb soil
solution, and the ability of the soil to provide it to the roots, determines
the quantity and rate of water, oxygen, and nutrient absorption by
plants, both of which have a significant impact on the health of vege-
tation, including crops. As a result, when proposing agricultural and
irrigation management policies, soil properties such as soil types,
nutrient conditions, soil organic carbon, and soil degradation should be
considered. As a result, when proposing agricultural management pol-
icies, the spatial pattern of the soil erosion region should be taken into
account.
The Aridity index is a numerical indicator of the degree of water
scarcity, which is most frequently related to natural vegetation and crop
distribution. As a result, it should be considered when modeling land
suitability for agriculture. Land availability can be determined by un-
derstanding existing land use. Land use data aids in determining an
area’s viability for a specific cropping method (Akpoti et al., 2019). As a
result, it must be taken into account before determining suitability for
agriculture. Based on these discussions, we have chosen the mentioned
parameters for agriculture suitability modeling.
2.2. ALAS modeling
2.2.1. Data filtering
In the present study, before proceeding with modeling, data filtering
has been performed. Data filtering or feature selection is essential before
modeling because redundant data can produce erroneous results.
Therefore, in the present study, we used multicollinearity analysis for
data filtering. Some approaches for measuring multicollinearity include
variance inflation factors (VIF), tolerance (TOL), linear support vector
machine, and chi-square (Talukder et al., 2020). In this study, tolerance
and VIF were employed to assess multicollinearity among the variables.
It is worth mentioning that variables with collinearity concerns should
be removed before modeling to achieve high accuracy results. The
greater the collinearity, the larger the VIF. The coefficient of determi-
nation was determined using rice production data (a proxy for existing
agricultural appropriate regions) as response variables and suitability
indicators as predictor factors in a linear regression analysis to evaluate
multicollinearity.
2.2.2. Methods for fuzzy logic and FAHP
The following steps have been performed to implement the fuzzy
logic and FAHP for generating ALSA.
Step-1: Development of hierarchical structure
In this phase, a literary survey and expert opinion relevant to the
field of research are used to implement the AHP process. The next step is
the creation of a matrix of judgment through comparison of pairs and
evaluations by decision-makers and experts. The matrix was checked for
consistency values. Following this, the consistency of the matrix de-
cisions in the pair-wise evaluation is verified. If the consistency test fails,
the initial values of the pair-wise comparison matrix must be changed. In
the present study, the CI and other evaluating factors are consistency
1%, CI is 0.01316, RI is 0.0146, CI/RI is 0.9.
Step-2: Determination of degree of membership and computation of fuzzy
evaluation matrix
The fuzzy membership function (MF) is determined by the fuzzy set
theory based on the spatial relationships of the conditioning variables in
fuzzy logic and decision-making. In contrast to boolean logic, fuzzy logic
has a membership value ranging from 0 to 1, while boolean logic has a
membership value of 0 or 1. (Qiu et al., 2014). The 0 values indicate that
the desired value is not a member of that set, while the value 1 indicates
that the desired value is a member of that set. Other values range from
0 to 1, depending on their level of membership. One of the biggest
challenges of fuzzy research is that the type of membership and its
criteria are not determined using the optimum approach. These
S. Talukdar et al. Agricultural Systems 196 (2022) 103343

membership functions are usually chosen based on decision-makers’
interests in the study field (Shahabi et al., 2015).
The variables in fuzzy set theory can be given a membership value
ranging from 0 to 1, which indicates the degree of MF. The linear
membership function (MF) is used in ArcGIS software to establish the
fuzzy map of each parameter. For all parameters (Supplementary
Figs. 1–4), the maximum and minimum values of the membership
functions are evaluated; for example, the MF value for a DEM less than
100 m is 1, the value for a DEM greater than 5000 m is MF = 0, and MF is
computed between 0 and 1 for DEM values between 100 and > 7000 m.
In the same way, the magnitude for other parameters is determined by
the minimum and maximum MF values. For each parameter between
0 and 1, membership functions were specified based on Table 2. As a
result, all agriculture land suitability conditioning variables were scaled
from 0 (lesser suitable) to 1 (highly suitable).
Step-3 final mapping
The final agricultural suitability mapping in this study was accom-
plished in two ways: (1) the agriculture suitable conditioning parame-
ters were fuzzified using membership functions, and then different fuzzy
operators, such as ‘AND’, ‘GAMMA 0.8’, and ‘GAMMA 0.9’, were used to
integrate all the fuzzified parameters and generate final agriculture
suitability maps in ArcGIS 10.8 software. (2) In ArcGIS software, the
weights for all parameters obtained from FAHP algorithms were easily
integrated using the weighted sum technique.
2.3. Sensitivity analysis
The sensitivity analysis was carried out in this study because ground
truth evidence for validating the land suitability model for agriculture is
sparse and difficult to come by. As a result, a sensitivity analysis was
performed to express the robustness and precision of the suitability
models. Sensitivity analysis was performed in three different ways in
this study: (1) global sensitivity analysis using the Morris model, (ii)
random forest-dependent sensitivity analysis, and (iii) Pearson’s corre-
lation coefficient.
2.3.1. Global sensitivity analysis using Morris model
Morris (1991) devised the Morris one-at-a-time approach (MOAT)
for parameter screening as a global sensitivity analysis method. All input
parameters’ overall effect and interaction effect are computed using the
mean m and standard deviations of the gradients of all input parameters
sampled from the r MOAT path as the theoretical basis of this process.
All input parameters were modified by the same relative amount using
this global sensitivity process. Morris’ procedure differs from conven-
tional OAT research because it considers adjusting the variable in
question between two model simulations (Morris, 1991; Saltelli et al.,
2000).
2.3.2. Random forest-based feature selection technique
Breiman’s Random Forest (RF) (Breiman, 2001) is one of the most
commonly used efficient ensemble supervised algorithms. The regres-
sion problem, classification, and unsupervised learning can all be solved
with this algorithm. It has been widely used in various fields, including
natural hazard modeling, hydrology, LULC classification, and finance
(Salam and Islam, 2020; Chen et al., 2020; Pal et al., 2020). In the
present study, RF offers two distinct important metrics for ordering
variables and variable choice, mean decrease accuracy (MDA) and mean
decrease Gini (MDG). When the values of a variable become randomly
permuted relative to the original data, MDA evaluates the significance of
the variable by evaluating the change in prediction accuracy. MDG is the
total of all Gini impurity reductions caused by a particular variable
(when that variable is used to generate a split in the Random Forest),
normalized by the number of trees.
This approach was used in the present analysis to determine the
weight or control of the various parameters in describing the expected
agriculture land suitability models. We computed the variable’s influ-
ence on different agriculture suitability models. Therefore, the agricul-
ture suitability models have been treated as target variables. To
implement the RF model, we set the following optimized parameters:
seed-5, number of trees- 500, number of variables tried at each split- 3,
iteration- 200, out of bag estimate error rate: 11.32%. To the best of the
authors’ knowledge, this is the first work in the study field to use RF-
based sensitivity analysis to identify the model’s most sensitive
parameters.
2.3.3. Pearson correlation-based sensitivity analysis
The Pearson correlation coefficient was used to assess the relation-
ship between ten parameters and land suitability approaches. We used
SPSS (version 22) tools to conduct correlation coefficient analysis on ten
parameters and land suitability models in this analysis.
The entire work has been summarized in Fig. 2.
3. Results
3.1. Data filtering analysis
A multicollinearity test was used in this work to see whether there
was a connection between the conditioning factors using variance
inflation factors (VIF) and the tolerances technique for selecting agri-
culture suitability conditioning factors (Table 2). VIF > 10 or tolerance

Table 2
Diagnosis of multicollinearity for agricultural suitability conditioning factors.
Coefficients

Model Unstandardized Coefficients Standardized Coefficients t Sig. Collinearity Statistics

B Std. Error Beta Tolerance VIF

1 (Constant) −0.194 0.745 −0.261 0.794

Elevation 10.356 0.847 0.381 12.227 0.000 0.112 8.941
Slope 6.039 0.399 0.249 15.128 0.000 0.402 2.485
Aspect 4.699 0.286 0.177 16.451 0.000 0.944 1.060
TRI 5.443 0.670 0.195 8.123 0.000 0.188 5.331
Rainfall 8.666 0.392 0.320 22.108 0.000 0.518 1.931
Temperature 4.003 0.962 0.080 4.160 0.000 0.292 3.420
Windspeed 3.599 0.469 0.125 7.670 0.000 0.411 2.431
Evapotranspiration 3.066 0.514 0.102 5.968 0.000 0.371 2.695
Soil qualities 0.126 0.352 0.005 0.358 0.721 0.597 1.674
SOM 0.505 0.460 0.012 1.098 0.273 0.876 1.141
Soil types 0.320 0.377 0.011 0.850 0.396 0.708 1.412
Soil erosion rate −0.554 0.366 −0.019 −1.513 0.131 0.660 1.516
LULC types 0.618 0.368 0.024 1.677 0.094 0.537 1.864
Aridity index 0.260 0.481 0.009 0.541 0.589 0.412 2.426

a: Dependent Variable: Rice production

5
S. Talukdar et al.
Fig. 2. Hierarchical structure of the agriculutural suitable models of the present work.
0.1 indicates a multicollinearity concern in conditioning factors (Mallick
et al., 2021). According to the multicollinearity test findings, no
collinearity among the 14 conditioning factors affecting the agriculture
suitability model (Table 2).
3.2. Description of data layers
3.2.1. Climatic parameters
The study region has a humid to desert environment with a south-
west monsoon system. Rainfall is an essential factor in determining the
suitability of agricultural land. We used ArcGIS 10.2 software to
generate a rainfall map using rainfall data from an Indian meteorolog-
ical station and the kriging interpolation method. In the studied area,
yearly rainfall ranges from 0 to 345 mm (Supplementary Fig. 1a).
Another important factor that influences crop output is temperature.
The yearly temperature ranges from 30.2 degrees Celsius to −18.3 de-
grees Celsius (Supplementary Fig. 1b). Compared to other locations, the
northernmost area observes greater temperatures. Evapotranspiration
6
Agricultural Systems 196 (2022) 103343
affects agricultural water resource management as well. The northern
and eastern portions of the nation had lower evapotranspiration,
whereas the central and northwestern areas had more evapotranspira-
tion (Supplementary Fig. 1c). Another critical factor in determining land
suitability is wind speed. The wind speed ranges from 0.5 to 10.3 knots
per second yearly (Supplementary Fig. 1d). Wind speeds were found to
be greater and lower in the research area’s northern and central parts.
3.2.2. Topographic parameters
Elevation and slope were recognized as critical topographic charac-
teristics in this study. The topographic factor has a significant influence
on farming mechanization. The elevation factor affects climatic data,
such as precipitation and temperature. Crop growth and dispersion are
greatly influenced by elevation above sea level. The greater height is
found in the northern and certain parts of the eastern hilly regions,
whereas the rest of the Deccan plateau is at a lower elevation (Supple-
mentary Fig. 2a). The elevation varies from 800 to 900 m above mean
sea level (AMSL) and accounts for more than a third of the research area.
S. Talukdar et al.
Crop yield benefits with a low slope. According to the slope study, in the
southern and southwestern areas, extremely softly sloped lands (1–3
degree) are present. Gently sloping lands (3–8 degrees) were found in
the middle areas, whereas moderate to high slope lands (>8 degrees)
were found throughout the study, mainly in the northern and eastern
parts (Supplementary Fig. 2b). Aspect is the compass direction that a
slope faces in relation to land owing convergence. The aspect map was
made up of eleven classes (Supplementary Fig. 2c). The degree of
elevation variation between close cells of a digital elevation map is
indicated as the topographic roughness index (TRI). It is based on the
research area’s local terrain. The northern margin and a small portion of
the eastern region had greater TRI, whereas the rest of the study area
had lower TRI (Supplementary Fig. 2d).
3.2.3. Soil related parameters
Soil organic carbon (SOC) is important for soil fertility, a complex
water-nutrient trade-off in the plant root zone, and land deterioration
(Bandyopadhyay et al., 2010). It changes with elevation, slope, rainfall
and soil variability and management on a geographical scale. The
greater SOC is found in the eastern part of the country, whereas the
lower SOC is found in the northwest (Supplementary Fig. 3a). Soil
erosion is also an important factor in determining agricultural suit-
ability. Soil erosion occurs mostly in higher elevation areas, notably in a
small fraction of the northern region, although it also occurs irregularly
across the study territory (Supplementary Fig. 3b). There were seven
types of soil quality employed in this study (Supplementary Fig. 3c). This
soil quality is critical for effective low-input farming and, to a lesser
extent, intermediate-input farming. There are several diagnostics for
nutritional availability. Texture/structure, organic carbon (OC), pH, and
total exchangeable bases (TEB) are all important topsoil (0–30 cm)
features. The most significant factors to consider for the subsoil (30–100
cm) are texture/structure, pH, and TEB. The soil parameters that affect
soil nutrient availability are connected to some extent. As a result, to
represent soil, the most limiting soil feature is merged with the average
of the remaining less limiting soil qualities in the evaluation. The study
region showed nutrient availability in most sections of the study region,
but toxicity and rooting conditions in the northern and eastern regions.
Soil types are one of the most important determinants of land suitability
(Islam et al., 2021). This study used a soil map from the United States
Geological Survey (USDA) and divided it into six soil types using USDA
soil taxonomy (Supplementary Fig. 3d). According to the soil taxonomy
model, entisols and inceptisols are often found in the study area (Soil
Survey Staff, 2010).
3.2.4. Land use related parameters
Aridity is one of the essential aspects of land suitability. A high de-
gree of aridity was discovered in the eastern and a small section of the
southwestern regions, while low aridity was discovered in the central
and northwestern areas (Supplementary Fig. 4a). The land pattern,
desertification, and evapotranspiration are all influenced by LULC
(Yalcin et al., 2011) and, for evaluating agricultural suitability zones,
the LULC map is crucial. The LULC map (Supplementary Fig. 4b)
included 19 classes, and the area calculated is shown in supplementary
table 1.
3.3. Fuzzification of the data layers
A fuzzy inference system comprises four essential components: a rule
basis, a fuzzifier, an inference engine, and a defuzzifier. A fuzzy set is a
rule-based model that functions as an active system. An inference engine
is used to prepare an inference from fuzzy rules (Ostovari et al. 2016).
Defuzzification is used to map a fuzzy dataset for the output of crisp
parameters. The fuzzy membership tool reclassifies or converts the input
data to a 0 to 1 scale, depending on the likelihood of belonging to a
specific set. Sigmoid, inverted sigmoid, linear, and parabolic types of
membership functions were used in our research. In addition, the
7
Agricultural Systems 196 (2022) 103343
Mamdani fuzzy inference system was used in this study (Mamdani,
1977). Climatic, soil, topography, and land use parameters were trans-
lated into fuzzy-set data with numerical values ranging from 0 to 1 using
various fuzzy membership functions (Supplementary Figs. 5–8).
3.4. Agriculture suitability modeling
Fig. 3 shows the results of the land suitability zone for agriculture,
and supplementary table 2 shows the area and percentage coverage of
the various agricultural suitability zones in India. The final agricultural
suitability map was created by superimposing 14 fuzzy data layers on
this land suitability map using Fuzzy AND, Fuzzy Gamma 0.9, fuzzy
Gamma 0.8, and Integrated AHP models (Fig. 3a–d). According to the
Fuzzy (AND operator) map, 16.76% (546,019.2 km2) of the study area is
very highly suitable, except for a small piece in the southwest, 8.04%
(261,936.3 km2) is moderately suitable, and 15.52% (505,722.6 km2) is
marginally suitable. Approximately 6.11% (199,067.2 km2) has been
assessed as currently not suitable, with the remaining 53.55%
(1,744,325 km2) being permanently not suitable, concentrated in the
northern, central, southern, and western areas. According to the Fuzzy
(Gamma 0.9 operator) model, 59.69% (1,944,325 km2) of the study area
is permanently unsuitable, 4.02% (131,067.2 km2) is currently not
suitable, and 14.45% (405,722.6 km2) is marginally suitable. In some
parts of the research area’s eastern region, 7.05% (229,936.5 km2) is
assessed to be moderately suitable, while 15.76% (546,019.2 km2) is
assessed to be highly suitable (Fig. 3b). According to the Fuzzy (Gamma
0.8 operator) map, 16.15% (526,019.2 km2) of the region is perma-
nently not suitable, 7.42% (241,936.5 km2) of the territory is currently
not suitable, and 14.91% (485,722.6 km2) of the region is marginally
suitable. Only the easternmost region of the nation is home to 4.97%
(162,067.2 km2) of moderately suitable, with the remaining 56.53%
(1,841,325 km2) of highly appropriate land dominating the rest of the
nation (Fig. 3c). According to the integrated AHP model, 19.78%
(644,324.7 km2) of the study area is permanently not suitable in the
northernmost region, 19.87% (647,265.6 km2) is currently not suitable
in the northernmost region, and 20.05% (676,012.6 km2) is marginally
suitable in the north-western and central parts of the area. Similarly,
20.10% (671,592.1 km2) is determined to be moderately suitable in the
north-eastern sections of the area, while the remaining 20.20%
(617,875.3 km2) is discovered to be in the highly suitable category in
India’s eastern region (Fig. 3d).
Some climatic parameters in the current study, such as rainfall,
temperature, humidity, and evapotranspiration, are time variable or
vary with time. As a result, these parameters might influence the model.
Consequently, we used historical rainfall and temperature data
(1990–2000) as input data for ALSA modeling to propose a dependable
and precise agricultural suitable model. Historical humidity and
evapotranspiration data, on the other hand, are not accessible. As a
result, we maintained the same humidity and evapotranspiration levels
as before. The FAHP model was then utilized to build a time-variant
agricultural suitability model (Fig. 4). The results reveal that the time-
variant ALSA model and the FAHP model-based ALSA are pretty iden-
tical. Then, using ArcGIS software’s ‘band collection statistic’ tool, we
used the correlation coefficient approach (Table 3). It indicates that the
fuzzy AND, fuzzy GAMMA0.8, and 0.9 models are over 50% compared
to the time-variant model (Table 3). It is, nevertheless, over 80% iden-
tical to the FAHP-based ALSA model. As a result, time-variant parame-
ters can affect the ALSA model to some level. As a result, the authors
propose that long-term climate data be used for agricultural suitability
modeling to achieve extremely robust and trustworthy results.
3.5. Ground validation
Ground validation is critical for any prediction model. However,
ground data is highly sparse and difficult to come by for some types of
complicated prediction models. As a result, proxy ground data may
S. Talukdar et al. Agricultural Systems 196 (2022) 103343

Fig. 3. Agricultural suitability modeling based on fourteen parameters for India using (a) fuzzy AND, (b) fuzzy gamma 0.9, (c) fuzzy gamma 0.8, (d) fuzzy AHP.

Fig. 4. Historical climatic variables such as rainfall (a) and (b) temperature used for FAHP based time variant (c) agricultural suitability model.

8
S. Talukdar et al. Agricultural Systems 196 (2022) 103343

Table 3
Correlation coefficient among three fuzzy operators based ALSA models, one fuzzy-AHP based ALSA model, and one fuzzy-AHP
based time variant ALSA model.

AND Gamma0.8 Gamma0.9 FAHP Time variant model

AND 1 0.99348 0.99504 0.82338 0.5264
Gamma0.8 0.99348 1 0.99848 0.81918 0.52898
Gamma0.9 0.99504 0.99848 1 0.82288 0.53266
FAHP 0.82338 0.81918 0.82288 1 0.80645
Time variant
model 0.5264 0.52898 0.53266 0.80645 1

evaluate and test the model’s’ reliability. Ground data on agricultural depending on the model’s present state, it may be claimed that FAHP
suitability zones is not accessible in the present research. As a result, we outperformed other models.
used several key crop production gridded historical data (1997–2003),
such as rice, wheat, potato, pulses, and tea, as proxies (latest data is
3.6. Sensitivity analysis
unavailable) (Fig. 5). Visually, the Indo-Gangetic plain is very fertile,
and most of crops have been farmed. Fig. 5 indicates that rice, wheat,
For four agricultural suitability models, sensitivity analysis used
and potato output were all relatively high in these areas. In contrast, tea
different machine learning algorithms and statistical approaches,
and pulses were abundant in the Western and Eastern Ghats. As a result,
including global sensitivity analysis, RF, and Pearson’s correlation ap-
these areas are also productive and suitable for farming. In India’s
proaches (Figs. 6–8). The agricultural suitability models (produced
complicated situation, determining an agriculturally suitable zone based
using fuzzy operators and fuzzy AHP) were employed for the sensitivity
on certain crops is highly challenging. As a result, four agriculture-
analysis. The target variables were land suitability models, while the
appropriate models have been compared to the production of these
independent factors were fourteen agriculture suitability conditioning
five key crops (Table 4). It indicates that the FAHP model, rather than
factors. We randomly gathered data from the target and independent
other models, has a strong association with these crops. Other models
variables based on 5000 points before doing sensitivity analysis. We
have a positive relationship with these crops as well. As a result,
used the retrieved data to do a sensitivity analysis to see how the

Fig. 5. Production of major crops of India in gridded format, such as (a) rice, (b) wheat, (c) potato, (d) tea, and (e) pulses for 1997–2003.

9
S. Talukdar et al. Agricultural Systems 196 (2022) 103343

Table 4
Correlation between agricultural suitability models and major crops, such as rice, wheat, potato, tea, and pulses.

Fuzzy AND Gamma 0.9 Gamma 0.8 FAHP Potato Pulses Rice Tea Wheat
Fuzzy AND 1 0.823 0.819 0.824 0.563 0.287 0.738 0.224 0.561
Gamma0.9 0.823 1 0.998 0.995 0.640 0.441 0.842 0.397 0.752
Gamma0.8 0.819 0.998 1 0.993 0.599 0.425 0.842 0.393 0.762
FAHP 0.824 0.995 0.993 1 0.641 0.448 0.837 0.405 0.756
Potato 0.563 0.640 0.599 0.641 1 0.507 0.557 0.303 0.357
Pulses 0.287 0.441 0.425 0.448 0.507 1 0.393 0.810 0.287
Rice 0.738 0.842 0.842 0.837 0.557 0.393 1 0.358 0.379
Tea 0.224 0.397 0.393 0.405 0.303 0.810 0.358 1 0.296
Wheat 0.561 0.752 0.762 0.756 0.357 0.287 0.379 0.296 1

Fig. 6. Global sensitivity analysis for evaluating the influence of the parameters to the agricultural suitability models using (a) fuzzy AND, (b) Gamma 0.9, (c)
Gamma 0.8, and (d) FAHP. (N.B. X1-Elevation, X2-Slope, X3-Aspect, X4-TRI, X5-Rainfall, X6-Temperature, X7-Wind Speed, X8-Evapotranspiration, X9-Soil organic
carbon, X10-Soil types, X11-Gross soil erosion, X12-LULC types, X13-Aridity).

independent factors affected the modeling of the target variable or land
suitability models.
3.6.1. Global sensitivity analysis
The Morris technique was used to conduct a global sensitivity anal-
ysis to test the dependability of the Fuzzy and AHP models in this study.
Agricultural suitability models are used as dependent factors, while 14
influencing variables are included as independent factors. We did a
global sensitivity analysis based on the extracted agriculture suitability
affecting factors to determine the most sensitive independent variable
for modeling agricultural suitability zones in the research region. Four
models were used to assess the effects of independent factors on agri-
cultural suitability zones. Because ground-based monitoring is not
feasible, sensitivity analysis is essential. Temperature, TRI, and soil
quality were identified using the Fuzzy AND model as critical indicators
that account for agricultural land suitability zones. On the other hand,
wind speed has the most negligible impact on agriculturally appropriate
models (Fig. 6a). Temperature, elevation, rainfall, and aspect were
anticipated to significantly affect agricultural suitability variables by the
Fuzzy Gamma 0.9 model, whereas soil quality was anticipated to be the
least influencing element for the research region (Fig. 6b). Soil quality,
soil organic carbon, and temperature were strongly responsible for
agricultural suitability using Fuzzy gamma 0.8. On the other hand, the
Fuzzy gamma model identified TRI as the least responsible variable
(Fig. 6c). The most influential markers for the agricultural adaptability
zone were temperature and rainfall, whereas evapotranspiration was the

10
 S. Talukdar et al.

11
Fig. 7. Random forest based sensitivity analysis for evaluating the influence of the parameters to the agricultural suitability models using (a) fuzzy AND, (b) Gamma 0.9, (c) Gamma 0.8, and (d) FAHP.
Agricultural Systems 196 (2022) 103343
S. Talukdar et al.
Fig. 8. Correlation coefficient-based sensitivity analysis for evaluating the influence of the parameters to the agricultural suitability models using (a) fuzzy AND, (b)
Gamma 0.9, (c) Gamma 0.8, and (d) FAHP.
least responsible variable (Fig. 6d).
3.6.2. RF based sensitivity analysis
In this work, we utilized two RF model error matrices, MDA and
MDG, to determine the relevance of factors versus agricultural suit-
ability models (Fig. 7). In this study, four agricultural suitability models
were used as target variables, and the relevance of predictor factors was
calculated for each of the stated models. The sensitivity analysis of the
fuzzy AND model is depicted in Fig. 7a. According to the MDA and MGD
values, the aridity index, rainfall, temperature, evapotranspiration, and
soil quality are the most sensitive factors, whereas aspect and soil
organic matter are the least susceptible. Fig. 7b, on the other hand,
depicts the sensitivity analysis of the GAMMA0.8 model. It shows that
the most sensitive factors are aridity index, rainfall, temperature,
evapotranspiration, and elevation, whereas the least sensitive ones are
aspect and gross soil erosion. Fig. 7c also displays the sensitivity analysis
12
Agricultural Systems 196 (2022) 103343
of GAMMA0.9, which demonstrates that the results are equal to
GAMMA0.8. Finally, Fig. 7d depicts the FAHP model’s sensitivity
analysis. The most sensitive factors are evapotranspiration, temperature,
soil organic matter, slope, and rainfall, whereas TRI and soil charac-
teristics are less susceptible. As a result, evapotranspiration and tem-
perature are the most ubiquitous sensitive factors, whereas rainfall, soil
organic matter, and slope are the most dynamic sensitive parameters. As
a result, to provide solid and sustainable management plans for eradi-
cating hunger, these indicators should be examined regularly.
3.6.3. Correlation coefficient-based sensitivity analysis
Using Pearson’s correlation coefficient, we computed the relation-
ship between agricultural suitability models as fourteen factors affecting
agricultural suitability (Fig. 8). At the significance level of 0.01, Fig. 8a
revealed that aridity index had a greater correlation (r: 0.59), followed
by rainfall (0.51), and evapotranspiration (0.34), while soil types had a
S. Talukdar et al.
very insignificant negative influence on agricultural suitability models.
On the other hand, similar to the fuzzy AND model, gamma 0.8 and
gamma 0.9 identified that aridity index, rainfall, and evapotranspiration
had the most influence on the agricultural suitability models (Fig. 8b &
c). In the case of the fuzzy AHP model, elevation, TRI, slope, rainfall,
temperature, wind speed, LULC, and soil erosion had the most influence
(Fig. 8d). Based on this analysis, it can be stated that rainfall, elevation,
slope, evapotranspiration, and aridity index strongly influenced the
agricultural suitability model.
4. Discussion
This paper proposes ALSA’s state-of-the-art and prospects. The SDG2
focuses on “ending hunger,” “achieving food security,” and “promoting
sustainable agriculture.” Food security can be attained through sus-
tainable agriculture, which will eventually lead to the abolition of
hunger. On the other hand, climate change is increasing the pressure on
agricultural areas by causing calamities such as drought and flooding.
To feed a growing population until 2050 and beyond, a significant
transformation in global food and farming systems is required. Agri-
cultural Land Suitability Analysis is a method of evaluating appropriate
agricultural land to promote sustainable agriculture and food security to
end hunger. On the other hand, ALSA must be implemented to under-
stand and consider current and future climate change consequences
fully. As a result, ALSA is an essential prerequisite for sustainable agri-
culture and food security in the face of climate change. As a result, in this
work, we propose hybrid models for constructing ALSA and reliability
testing. A closer look at the area and percentage coverage of the coun-
try’s agricultural land suitability zones shows a much difference be-
tween the agricultural suitability zones in India (Supplementary
Table 1). Just 16 to 20% of the research area in eastern and northeastern
areas and a limited portion of southwestern regions is designated as
“moderately suitable,” while more than half of the study area is desig-
nated as “highly unsuitable.” The region and percentage coverage of
“marginally suitable” lands range slightly, ranging from 14.45% to
20.05%.
To maximize crop yields across India, it is critical to identify suitable
agricultural land for various crops. As a result, many studies have esti-
mated the suitability of agricultural land in existing works (FAO, 1976;
Fontes et al., 2009; Bandyopadhyay et al., 2010; Grassano et al., 2011;
Akıncı et al., 2013; Zolekar and Bhagat, 2015; Kazemi et al., 2016;
Kazemi and Akinci, 2018; Ostovari et al., 2019; Pilevar et al., 2020;
Ramamurthy et al., 2020; Rukanee et al., 2020; Saha et al., 2021).
Topographic conditions, climate variables, soil characteristics, and local
expert knowledge all play a role in crop production on agricultural land
(Grassano et al., 2011). Temperature, soil quality, rainfall, soil organic
carbon, and elevation, as shown in supplementary Figs. 2–4, are
important factors in increasing crop yield.
Some previous studies have used fuzzy sets to determine land suit-
ability (Zhang et al., 2015; Keshavarzi et al., 2010; Braimoh et al.,
2004). The fuzzy set model has many advantages, including the ability
to (i) convert all data to a range of 0–1, which is an excellent mechanism
for solving different magnitudes at various data layers, and (ii) generate
a detailed land suitability assessment for each crop production on a
continuous scale in various land categories using fuzzy membership
functions. Many researchers have used the combined Fuzzy set with the
AHP approach to measuring contributing variables’ weights in recent
decades (Zhang et al., 2015; Kazemi et al., 2016; Pilevar et al., 2020).
The weighting of factors in our research can be compared to other
studies, including Kazemi et al. (2016), Ostovari et al. (2019), and
Pilevar et al. (2020), which found that soil factors are an important
factor in modeling land suitability for maize crop production. According
to the Morris method, temperature, soil quality, and rainfall play a
critical role in agricultural suitability for the entire country. Elevation,
which can be linked to land suitability, primarily influenced tempera-
ture and soil properties. Increasing elevation raises the likelihood of
13
Agricultural Systems 196 (2022) 103343
frost, cooler temperatures, and a decrease in plant growth months.
The most sensitive variables are temperature, soil quality, elevation,
and rainfall, resulting in a highly suitable to the marginally suitable
region for crops. High wind speed (northern parts of the area), evapo-
transpiration (western and southern portions), and TRI (northern and
eastern locations) limitations, on the other hand, curtailed the country’s
agricultural crop production capacity (Fig. 8). (Ahmed and Jeb, 2014)
reported similar results. Rainfall, temperature, and soil organic content,
on the other hand, were identified as limiting factors in the suitability of
a micro-water watershed for sorghum production (Mohan and Logana-
than, 2008). The northern, north-central, and western parts of the
country are called environmentally sensitive zones to be preserved and
protected. As a result, farming operations are prohibited in these fields.
It is recommended that land managers and decision-makers priorities
those extremely suitable areas in states. Nevertheless, micro-level
planning, like district and block-level, can be encouraged. It can be
fruitful for achieving SDGs.
The AHP-Fuzzy logic method has been effective in combining data
layers with a variety of subjective knowledge. This method is an
important model for scientifically integrating heterogeneous datasets
and calculating weight, thus assessing any inconsistency in this
appraisal. In the present study, we developed AHP based fuzzy logic
model for generating ALSA at the national level. Most previous studies
have generated the agriculture suitability models for particular micro-
regions using multi-criteria decision-making techniques (Kumar and
Shaikh, 2013; Akıncı et al., 2013; Zhang et al., 2015; Jamil et al., 2018).
Machine learning approaches to agriculture suitability models have
been utilized in very few researches, especially in developing nations
like India and Bangladesh (Mohan and Loganathan, 2008). As a result, in
this work, we combined multi-criteria decision-making techniques, such
as AHP, with machine learning techniques such as fuzzy logic to
generate ALSA and a highly reliable and accurate model. Finally,
farming managers and land-use developers will use the proposed model
to identify alternate land uses/land covers for particular crop produc-
tion. Compared to traditional agricultural suitability appraisal models,
AHP-Fuzzy logic analysis of agricultural suitability evaluation has the
advantage of being able to render appraisals at the national level by
appraising each data unit individually. As a result, it is possible to
determine the contribution of each soil, climatic, and topographic aspect
to the assessment and make the results applicable to each landowner.
The innovation of this study is the application of global sensitivity
analysis, such as the Morris method, and machine learning-based
sensitivity on agricultural suitability model based on the integrated
fuzzy logic-AHP model. Therefore, this study provides reliable findings,
which can propose sustainable agriculture management strategies to
increase agricultural production. To achieve optimum crop yield, how-
ever, significant attention should be given to crop selection that is well
suited to the agro-ecological conditions and proper land management.
5. Policy implication and conclusion
The present study used Fuzzy AND, Fuzzy Gamma 0.9, Fuzzy Gamma
0.8, and Integrated AHP models using fourteen parameters to create
India’s agricultural land suitability maps. Land suitability evaluation is
the first step to obtain optimum climatic, land use, topographic, hy-
drological, pedogenic, and social factors for the most sustainable agri-
cultural suitability in the studied area. According to the integrated AHP
model, which is regarded as the best model, 19.78% of the study area is
permanently unsuitable in the northernmost region, 19.87% is currently
unsuitable in the northernmost region, and 20.05% is marginally suit-
able in the area’s northwestern and central parts. Similarly, in the north-
eastern regions of the area, 20.10% is revealed to be fairly appropriate,
while the remaining 20.20% is identified to be very suited to India’s
eastern region. Aridity index, rainfall, temperature, evapotranspiration,
and elevation are the most sensitive elements, whereas aspect and gross
soil erosion are the least susceptible. However, in the present research,
S. Talukdar et al.
various environmental factors, such as rainfall, temperature, humidity,
and evapotranspiration, are time variable or fluctuate with time. As a
consequence, these parameters may influence the model over time. As a
result, historical rainfall and temperature data (1990–2000) and the
other parameters were used as input data for ALSA modeling to suggest a
reliable and accurate agricultural-suited model. The correlation coeffi-
cient results suggest that the time-variant model is over 80% compared
to the FAHP-based ALSA model. Therefore, temporal variant parameters
may influence the ALSA model to some extent. As a consequence, the
authors advise that long-term climate data be utilized for agricultural
suitability modeling to get exceptionally stable and trustworthy
conclusions.
Our study on agricultural land suitability maps relying on FAO
guidance would benefit India’s agriculture policymakers in expanding
the projects to new areas. Since cropland is strongly suitable for S1
categories (46%), there is a massive opportunity to harness agroforestry
in cropland areas. In India, over 80% of farmers are small landowners,
and most of them depend on rain for their agricultural practices. During
the monsoon season, they only use the groundwater for agricultural
purposes. Because of a lack of irrigation, the land is not used for agri-
culture for the rest of the period. The rainwater harvest scheme would
enable soil moisture preservation on the land before the next harvest
season when combined with different soil and water management ac-
tivities at the watershed level. The above research would significantly
aid in creating and establishing new agriculture centers for further
research on a regional scale.
There are a few flaws in this analysis. First, combined fuzzy-AHP
results are based on a subjective assessment of the relative signifi-
cance of the two variables, and associations between variables were
overlooked. Second, the spatial resolution of the images is low. There-
fore, micro-level like village or mouza level planning based on this work
can be critical.
The present work can be improved further by considering some vital
parameters for modeling, such as agricultural land availability and the
social context, such as irrigation data, fertilizers’ use, and machinery
use. The application of advanced machine learning algorithms can
improve the agricultural suitable model. One of the most complex as-
pects of this work is incorporating social perspectives into the spatial
demonstration. As a result, these viewpoints need to be explored further.
Declaration of Competing Interest
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influence
the work reported in this paper.
Acknowledgement
The authors extend their appreciation to the Deanship of Scientific
Research at King Khalid University for funding this work through
Research Groups (short) under grant number (R.G.P2 /89/42).
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.
org/10.1016/j.agsy.2021.103343.
References
Ahmed, M., Jeb, D.N., 2014. Land suitability for sorghum using multicriteria evaluation
(MCE) and analytical hierarchy process (AHP) in Bunkure Kano State, Nigeria.
J. Agric. Vet. Sci. (IOSRJAVS) e-ISSN 7 (9), 2319–2380.
Ahmed, G.B., Shariff, A.R.M., Balasundram, S.K., Abdullah, A.F.B., 2016. Agriculture
land suitability analysis evaluation based multi criteria and GIS approach. IOP Conf.
Series: Earth Environ. Sci. 37, 012044.
Akıncı, H., Özalp, A.Y., Turgut, B., 2013. Agricultural land use suitability analysis using
GIS and AHP technique. Comput. Electron. Agric. 97, 71–82.
14
Agricultural Systems 196 (2022) 103343
Akpoti, K., Kabo-bah, A.T., Zwart, S.J., 2019. Review - agricultural land suitability
analysis: state-of-the-art and outlooks for integration of climate change analysis.
Agric. Syst. 173, 172–208.
Anderson, R., Bayer, P.E., Edwards, D., 2020. Climate change and the need for
agricultural adaptation. Curr. Opin. Plant Biol. 56, 197–202.
Arora, N.K., 2019. Impact of climate change on agriculture production and its sustainable
solutions. Environ. Sustain. 2, 95–96.
Bandyopadhyay, P.K., Saha, S., Mani, P.K., Mandal, B., 2010. Effect of organic inputs on
aggregate associated organic carbon concentration under long-term rice–wheat
cropping system. Geoderma 154 (3–4), 379–386.
Bhatt, R., Hossain, A., 2019. Concept and Consequence of Evapotranspiration for
Sustainable Crop Production in the Era of Climate Change. Available online: https://
www.intechopen.com/books/advanced-evapotranspiration-methods-and-applica
tions/concept-and-consequence-of-evapotranspiration-for-sustainable-crop-pro
duction-in-the-era-of-climate.
Bocchiola, D., Brunetti, L., Soncini, A., Polinelli, F., Gianinetto, M., 2019. Impact of
climate change on agricultural productivity and food security in the Himalayas: a
case study in Nepal. Agric. Syst. 171, 113–125.
Braimoh, A.K., Vlek, P.L., Stein, A., 2004. Land evaluation for maize based on fuzzy set
and interpolation. Environ. Manag. 33 (2), 226–238.
Breiman, L., 2001. Random forests. Mach. Learn. 45, 5–32.
Chen, W., Li, Y., Xue, W., Shahabi, H., Li, S., Hong, H., Wang, X., Bian, H., Zhang, S.,
Pradhan, B., Ahmad, B.B., 2020. Modeling flood susceptibility using data-driven
approaches of naïve bayes tree, alternating decision tree, and random forest
methods. Sci. Total Environ. 701, 134979.
Chozom, K., Nimasow, G., 2021. GIS- and AHP-based land suitability analysis of Malus
domestica Borkh. (apple) in West Kameng district of Arunachal Pradesh, India. Appl.
Geomat. https://doi.org/10.1007/s12518-021-00354-7.
Dağdeviren, M., Yavuz, S., Kılınç, N., 2009. Weapon selection using the AHP and TOPSIS
methods under fuzzy environment. Expert Syst. Appl. 36 (4), 8143–8151.
Durán, O., Aguilo, J., 2008. Computer-aided machine-tool selection based on a fuzzy-
AHP approach. Expert Syst. Appl. 34, 1787–1794.
El Baroudy, A.A., 2016. Mapping and evaluating land suitability using a GIS-based
model. CATENA 140, 96–104.
Ertuğrul, I., Karakaşoğlu, N., 2008. Comparison of fuzzy AHP and fuzzy TOPSIS methods
for facility location selection. Int. J. Adv. Manuf. Technol. 39, 783–795.
FAO, 1976. A Framework for Land Evaluation. Food and Agriculture Organization of the
United Nations, Soils Bulletin No. 32. FAO, Rome.
Feizizadeh, B., Blaschke, T., 2013. GIS-multicriteria decision analysis for landslide
susceptibility mapping: comparing three methods for the Urmia lake basin, Iran. Nat.
Hazards 65, 2105–2128.
Fischer, G., Nachtergaele, F., Prieler, S., van Velthuizen, H.T., Verelst, L., Wiberg, D.,
2008. Global Agro-Ecological Zones Assessment for Agriculture (GAEZ 2008). IIASA,
Laxenburg, Austria and FAO, Rome, Italy.
Fontes, M.P., Fontes, R.M., Carneiro, P.A., 2009. Land suitability, water balance and
agricultural technology as a Geographic-Technological Index to support regional
planning and economic studies. Land use policy 26 (3), 589–598.
Fukase, E., Martin, W., 2020. Economic growth, convergence, and world food demand
and supply. World Dev. 132, 104954.
Gardner, A.S., Maclean, I.M.D., Gaston, K.J., Bütikofer, L., 2021. Forecasting future crop
suitability with microclimate data. Agric. Syst. 190, 103084.
Gilliams, S., Orshoven, J.V., Muys, B., Kros, H., Heil, G.W., Van Deursen, W., 2005.
AFFOREST sDSS: a metamodel based spatial decision support system for
afforestation of agricultural land. New Forest 30, 33–53.
Grassano, N., Tedone, L., Verdini, L., De Mastro, G., 2011. Evaluation of rapeseed
cultivation suitability in Apulia with GIS-multicriteria analysis. Ital. J. Agron. 6 (2),
e16.
Halder, J.C., 2013. Land suitability assessment for crop cultivation by using remote
sensing and GIS. J. Geogr. Geol. 5 (3).
Hathaway, M.D., 2016. Agroecology and permaculture: addressing key ecological
problems by rethinking and redesigning agricultural systems. J. Environ. Stud. Sci. 6,
239–250.
Herzberg, R., Pham, T.G., Kappas, M., Wyss, Tran, C.T.M., 2019. Multi-criteria decision
analysis for the land evaluation of potential agricultural land use types in a hilly area
of Central Vietnam. Land 8 (6), 90.
Huang, C.C., Chu, P.Y., Chiang, Y.H., 2008. A fuzzy AHP application in government-
sponsored R&D project selection. Omega 36 (6), 1038–1052.
Hunter, M.C., Smith, R.G., Schipanski, M.E., Atwood, L.W., Mortensen, D.A., 2017.
Agriculture in 2050: recalibrating targets for sustainable intensification. BioScience
67 (4), 386–391.
IPCC, 2014. AR5 Synthesis Report: Climate Change 2014. In: Contribution of Working
Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on
Climate Change [Core Writing Team, R.K. Pachauri and L.A. Meyer (Eds.)]. IPCC,
Geneva, Switzerland.
Islam, A.R.M.T., Talukdar, S., Mahato, S., Kundu, S., Eibek, K.U., Pham, Q.B., Kuriqi, A.,
Linh, N.T.T., 2021. Flood susceptibility modelling using advanced ensemble machine
learning models. Geosci. Front. 12 (3), 101075.
Jakhar, S.K., Barua, M.K., 2014. An integrated model of supply chain performance
evaluation and decision-making using structural equation modelling and fuzzy AHP.
Prod. Plan. Control 25 (11), 938–957.
Jamil, M., Ahmed, R., Sajjad, H., 2018. Land suitability assessment for sugarcane
cultivation in Bijnor district, India using geographic information system and fuzzy
analytical hierarchy process. GeoJournal 83 (3), 595–611.
Kazemi, H., Akinci, H., 2018. A land use suitability model for rainfed farming by Multi-
criteria Decision-making Analysis (MCDA) and Geographic Information System
(GIS). Ecol. Eng. 116, 1–6.
S. Talukdar et al.
Kazemi, H., Sadeghi, S., Akinci, H., 2016. Developing a land evaluation model for faba
bean cultivation using geographic information system and multi-criteria analysis (A
case study: Gonbad-Kavous region, Iran). Ecol. Indic. 63, 37–47.
Keshavarzi, A., Sarmadian, F., Heidari, A., Omid, M., 2010. Land suitability evaluation
using fuzzy continuous classification (a case study: Ziaran region). Mod. Appl. Sci. 4
(7), 72.
Kumar, M., Shaikh, V.R., 2013. Site suitability analysis for urban development using GIS
based multicriteria evaluation technique. J. Indian Soc. Remote. Sens. 41 (2),
417–424.
Lambin, E.F., Meyfroidt, P., 2011. Global land use change, economic globalization, and
the looming land scarcity. Proc. Natl. Acad. Sci. 108 (9), 3465–3472.
Leisner, P., 2020. Review: climate change impacts on food security- focus on perennial
cropping systems and nutritional value. Plant Sci. 293, 110412.
Liu, Y.S., Wang, J.Y., Guo, L.Y., 2006. GIS-based assessment of land suitability for
optimal allocation in the Qinling Mountains, China. Pedosphere 16 (5), 579–586.
Malczewski, J., 1999. GIS and Multicriteria Decision Analysis. John Wiley and Sons, Inc,
New York.
Malczewski, J., 2004. GIS-based land-use suitability analysis: a critical overview. Prog.
Plan. 62 (1), 3–65.
Mallick, J., Alqadhi, S., Talukdar, S., AlSubih, M., Ahmed, M., Khan, R.A., Kahla, N.B.,
Abutayeh, S.M., 2021. Risk assessment of resources exposed to rainfall induced
landslide with the development of GIS and RS based ensemble metaheuristic
machine learning algorithms. Sustain. 13 (2), 457.
Mamdani, E.H., 1977. Application of fuzzy logic to approximate reasoning using
linguistic synthesis. IEEE Trans. Comput. 26 (12), 1182–1191.
Minda, T.T., van der Molen, M.K., Struik, P.C., Combe, M., Jiménez, P.A., Khan, M.S., de
Arellano, J.V.G., 2018. The combined effect of elevation and meteorology on potato
crop dynamics: a 10-year study in the Gamo Highlands, Ethiopia. Agric. For.
Meteorol. 262, 166–177.
Mohan, S., Loganathan, G., 2008. Strategic decision-making in irrigated agriculture using
sustainability–related criteria. In: Twelfth international water technology conference.
IWTC12 Alexandria, Egypt.
Morris, M.D., 1991. Factorial sampling plans for preliminary computational experiments.
Techinometrics 33, 161–174.
Mosleh, Z., Salehi, M.H., Fasakhod, A.A., Jafari, A., Mehnatkesh, A., Borujeni, I.E., 2017.
Sustainable allocation of agricultural lands and water resources using suitability
analysis and mathematical multi-objective programming. Geoderma 303, 52–59.
Mozumdar, L., 2012. Agricultural productivity and food security in the developing
world. Bangladesh J. Agricult. Econ. 35 (1–2), 53–69.
Nabati, J., Nezami, A., Neamatollahi, E., Akbari, M., 2020. GIS-based agro-ecological
zoning for crop suitability using fuzzy inference system in semi-arid regions. Ecol.
Indic. 117, 106646.
OECD/FAO, 2019. "Overview", in OECD-FAO Agricultural Outlook 2019-2028. OECD
Publishing, Paris. https://doi.org/10.1787/dd8a74b8-en.
Olesen, J.E., Bindi, M., 2002. Consequences of climate change for European agricultural
productivity, land use and policy. Eur. J. Agron. 16, 239–262.
Ostovari, Y., Honarbakhsh, A., Sangoony, H., Zolfaghari, F., Maleki, K., Ingram, B., 2019.
GIS and multi-criteria decision-making analysis assessment of land suitability for
rapeseed farming in calcareous soils of semi-arid regions. Ecol. Indic. 103, 479–487.
Özkan, B., Dengiz, O., Turan, İ.D., 2020. Site suitability analysis for potential agricultural
land with spatial fuzzy multi-criteria decision analysis in regional scale under semi-
arid terrestrial ecosystem. Sci. Rep. 10, 22074.
Pal, S., Talukdar, S., Ghosh, R., 2020. Damming effect on habitat quality of riparian
corridor. Ecol. Indic. 114, 106300.
Pilevar, A.R., Matinfar, H.R., Sohrabi, A., Sarmadian, F., 2020. Integrated fuzzy, AHP and
GIS techniques for land suitability assessment in semi-arid regions for wheat and
maize farming. Ecol. Indic. 110, 105887.
Qiu, F., Chastain, B., Zhou, Y., Zhang, C., Sridharan, H., 2014. Modeling land suitability/
capability using fuzzy evaluation. GeoJournal 79 (2), 167–182.
Ramamurthy, V., Reddy, G.O., Kumar, N., 2020. Assessment of land suitability for maize
(Zea mays L) in semi-arid ecosystem of southern India using integrated AHP and GIS
approach. Comput. Electron. Agric. 179, 105806.
Rukanee, D., Sangchan, S., Choomjaihan, P., 2020. Assessment of the suitability of land
use for agriculture by analytical hierarchy process: AHP in lower Prachinburi
watershed, eastern Thailand. Agricultural Engineering International: CIGR Journal
22 (3), 19–26.
Saha, T.K., Pal, S., Sarkar, R., 2021. Prediction of wetland area and depth using linear
regression model and artificial neural network based cellular automata. Ecol. Inform.
62, 101272.
15
Agricultural Systems 196 (2022) 103343
Sahoo, S., Sil, I., Dhar, A., Debsarkar, A., Das, P., Kar, A., 2018. Future scenarios of land-
use suitability modeling for agricultural sustainability in a river basin. J. Clean. Prod.
205, 313–328.
Salam, R., Islam, A.R.M.T., 2020. Potential of RT, Bagging and RS ensemble learning
algorithms for reference evapotranspiration prediction using climatic data-limited
humid region in Bangladesh. J. Hydrol. 590, 125241.
Saltelli, A., Tarantola, S., Campolongo, F., 2000. Sensitivity analysis as an ingredient of
modeling. Stat. Sci. 15 (4), 377–395.
Salvacion, A.R., 2019. Mapping land limitations for agricultural land use planning using
fuzzy logic approach: a case study for Marinduque Island, Philippines. GeoJournal
1–11.
Sands, R.D., Jones, C.A., Marshall, E., 2014. Global Drivers of Agricultural Demand and
Supply ERR-174 (Washington, DC: US Department of Agriculture, Economic
Research Service).
Schneider, U.A., Havlík, P., Schmid, E., Valin, H., Mosnier, A., Obersteiner, M.,
Böttcher, H., Skalský, R., Balkovič, J., Sauer, T., Fritz, S., 2011. Impacts of
population growth, economic development, and technical change on global food
production and consumption. Agric. Syst. 104 (2), 204–215.
Seng, V., Bell, R.W., Hin, S., Schoknecht, N., Vance, W., White, P.F., 2009. Soil factors
affecting crop suitability for upland crops in Cambodia. Cambodian J. Agricult. 9
(1–2), 24–37.
Seyedmohammadi, J., Sarmadian, F., Jafarzadeh, A.A., McDowell, R.W., 2019.
Development of a model using matter element, AHP and GIS techniques to assess the
suitability of land for agriculture. Geoderma 352, 80–95.
Shahabi, H., Hashim, M., Ahmad, B.B., 2015. Remote sensing and GIS-based landslide
susceptibility mapping using frequency ratio, logistic regression, and fuzzy logic
methods at the central Zab basin. Iran Environ. Earth Sci. 73, 8647–8668.
Singh, S.K., Singh, K.M., Singh, R., Kumar, A., Kumar, U., 2014. Impact of rainfall on
agricultural production in Bihar: a zone-wise analysis. Environ. Ecol. 32 (4A),
1571–1576.
Soil Survey Staff, 2010. Keys to Soil Taxonomy, 11th edition. USDA — Natural Resources
Conservation Service, Washington DC, p. 338.
Song, G., Zhang, H., 2021. Cultivated land use layout adjustment based on crop planting
suitability: a case study of typical counties in Northeast China. Land 10 (2), 107.
Talukder, B., Blay-Palmer, A., vanLoon, G.W., Hipel, K.W., 2020. Towards complexity of
agricultural sustainability assessment: Main issues and concerns. Environ. Sustain.
Indicat. 100038.
Tashayo, B., Honarbakhsh, A., Azma, A., Akbari, M., 2020. Combined fuzzy AHP–GIS for
agricultural land suitability modeling for a watershed in southern Iran. Environ.
Manag. 66 (2020), 364–376.
Tercan, E., Dereli, M.A., 2020. Development of a land suitability model for citrus
cultivation using GIS and multi-criteria assessment techniques in Antalya province of
Turkey. Ecol. Indic. 117, 106549.
Vasu, D., Srivastava, R., Patil, N.G., Tiwary, P., Chandran, P., Singh, S.K., 2018.
A comparative assessment of land suitability evaluation methods for agricultural
land use planning at village level. Land Use Policy 79, 146–163.
Wang, D., Li, X.R., Li, Y., 2013. China’s “smart tourism destination” initiative: a taste of
the service-dominant logic. J. Destin. Mark. Manag. 2 (2), 59–61.
Yalcin, A., Reis, S., Aydinoglu, A.C., Yomralioglu, T., 2011. A GIS-based comparative
study of frequency ratio, analytical hierarchy process, bivariate statistics and
logistics regression methods for landslide susceptibility mapping in Trabzon, NE
Turkey. Catena 85 (3), 274–287.
Yi, X.B., Wang, L., 2013. Land suitability assessment on a watershed of loess plateau
using the analytic hierarchy process. PLoS One 8, e694987.
Yohannes, H., Soromessa, T., 2018. Land suitability assessment for major crops by using
GIS-based multi-criteria approach in Andit Tid watershed. Ethiopia. Cogent Food &
Agriculture 4 (1), 1470481.
Zabihi, H., Ahmad, A., Vogeler, I., Said, M.N., Golmohammadi, M., Golein, B.,
Nilashi, M., 2015. Land suitability procedure for sustainable citrus planning using
the application of the analytical network process approach and GIS. Comput.
Electron. Agric. 117, 114–126.
Zhang, S., Mishra, Y., Shahidehpour, M., 2015. Fuzzy-logic based frequency controller
for wind farms augmented with energy storage systems. IEEE Trans. Power Syst. 31
(2), 1595–1603.
Zolekar, R.B., Bhagat, V.S., 2015. Multi-criteria land suitability analysis for agriculture in
hilly zone: remote sensing and GIS approach. Comput. Electron. Agric. 118,
300–321.