Modeling the Rice Land Suitability Using GIS and Multi-Criteria Decision Analysis Approach in Sindh , Pakistan

The objective of this research was to evaluate rice land suitability in Sindh, Pakistan, by designing GIS-based Multi-Criteria Decision Analysis (MCDA) spatial model to aggregate interdisciplinary aspects including factors of soil physical and chemical properties, ground water quality, soil pH, agro-ecological zones, canal command area and temperature. A constraint map of water bodies was also utilized in this model. On the basis of these parameters, standardized raster maps were created, and then Pair-Wise Comparison Matrix (PWCM) of Analytical Hierarchy Process (AHP) was developed to calculate significant weights by means of Principal Eigen vector by Saaty’s method, with accepted Consistency Ratio (CR) of 0.10. Furthermore, Multi-Criteria Evaluation (MCE) employing Weighted Linear Combination (WLC) aggregated all the suitability maps to yield rice land suitability map. Final output map of this work demonstrated 30.2% increase in area suitable for rice cultivation with an increased production of 14,716,592.17 tonnes as compared to existing rice practices in Sindh. This increase in the area and production of the potential land shows the capability of our novel model and offers an opportunity to improve cultivation by providing the much required information at local level that could benefit farmers, vision scientists and decision makers to select appropriate cropping site and agricultural planning making the best use of available data.


INTRODUCTION
Rice is widely consumed staple food for a large part of population, especially in Asia with second highest worldwide production.Top rice exporting countries are: India, Thailand, Pakistan, United States, Vietnam, Uruguay, Brazil, and China [1].Pakistan has produced 5.8 million tonnes of rice in 2013-14, from which Sindh contributes to 2.6 million tonnes (45%).The main research problem is that rice is grown in Sindh on 1.8 million acres (27%) from total rice arable land of 6.9 million acres of Pakistan, with 1420.8Kg/acre yield in 2013-14, that does not reflect actual potential to fulfill high export demand [2].
The GIS-based land-use suitability analysis has been applied in ecological, agricultural activities, landscape evaluation, and planning the environmental impacts [3].For rice land suitability in Sheikhupura and Nankana Sahib, Pakistan, AHP computes significant weights for soil, environment and ground water quality attributes by PWCM [4].For the delineation of suitable soils for zero-till wheat cultivation in Gujranwala, Pakistan, GIS-based remote sensing and field data for soil texture, bulk density and ground water quality have been suggested.Overlay integrates the parameters for generating final map to identify best, normal, moderate and unsuitable soils [5].For identification of maize land suitability in Okara, Pakistan, GIS-based AHP model has been employed using parameters of soil pH, electrical conductivity, soil texture, organic matter and ground water quality to recognize areas of highly, moderately, marginally and not suitable [6].For the locations and distributions of rice cultivation in Nile Delta, PWCM and weighted overlay have produced suitability map [7].For rice land suitability in Prachuap Khiri Khan, Thailand, GIS-based AHP ranking technique has been utilized to weight coefficients.Further WLC has produced diverse suitable lands for cultivation [8].For rice land suitability map, based on physical and climatic factors in Great Mwea Region, Kenya, GIS-based MCE technique has been employed [9].For rice suitability sites in Morobe, Papua New Guinea, parameters of topography, physical and chemical soil properties, and climate have been used to construct index model [10].
Due to current depleting resources and poor agricultural land management, in Sindh, Pakistan, this work proposed a GIS-based MCDA model for rice land suitability, to boost the production.Data analysis techniques, experts knowledge and individual requirements for rice were essential in this decision making process.This study aimed to evaluate practice of rice farming and to help decision makers to adopt a GIS based more flexible, comprehensive and reliable suitability map to elevate production.

Study Area
Sindh is the third largest province of Pakistan with an area of 140,914 km 2

Environmental Variables
As a Kharif crop, rice is planted in summer while harvested in early winter.For 75 days rice fields should have 6 inches of slow moving water, so the deficiency of rainfall in study area has been met by canals and groundwater.Heavy clayey subsoil having water retaining capacity with level land is suitable.Rice requires high temperature during growing season of 4-6 months [12].In this work, the data source for all the criteria for rice in Sindh was: statistical data for rice production and area sown [2], shape file of Sindh administrative boundary [13], map of ground water quality with its parameters as adopted by SMO-WAPDA [14,15], shape file of Sindh mean maximum annual temperature [16], map of canal command area [15], digital scanned map of water bodies, soil physical and chemical properties, soil pH, and agro-ecological zones [17].The detailed description of rice suitability in Sindh for all the criteria used in this work is given in Table 1, as suggested by [3].These values are in agreement with those considered in literature.

Framework for GIS-Based MCDA Model
To determine rice land suitability, GIS-based MCDA model, as shown in Figure 1 depicts all steps used in this research procedure.In Erdas Imagine 9.2, the downloaded digital scanned maps for all criteria, which were originally in geographic latitude/longitude projection, were geometric transformed into real world projected coordinate maps.In rectification module of Erdas, by triangulation geo-coding method, linear rubber sheet map transformation was used to rectify scanned maps.In order to geo-reference all the criteria maps to WGS84 projection, 10 Ground Control Points were used to assign coordinates.First order polynomial equation and nearest neighbor re-sampling method quantified new values for output image.
All geometrically rectified criteria maps were reprojected to WGS84 Universal Transverse Mercator (UTM) zone 42N in ArcGIS 10.2.2, as the projection of Sindh administrative boundary shape file was already in WGS84 UTM zone 42N.Manual digitization of all the    criteria maps was performed in ArcGIS with projected administrative boundary shape file and using geometrically rectified digital scanned criteria maps as base maps.In this work, Figure 2 represents vector models denoting entire classes in a particular layer.For production of standardized criteria, rasterization of all digitized criteria maps was performed in Idrisi Selva with resolution of 100x100m.For standardization by reclass module of Idrisi, categories of interest that meet particular criterion were isolated by value of 1 while unconcerned categories were set to 0. Figure 3 illustrates suitable and not suitable classes of standardized criteria maps for this work.
Derivation of weights was central step for this work.In Idrisi, by employing AHP of decision support system, weights were derived for the factors by utilizing PWCM by Saaty's method.In developing weights, every possible pairing was compared by rating relative importance of factors on a 9 point rating scale and entered into a PWCM, where 1/9 indicates extremely less important and 9 indicates extremely more important [18].For this work, PWCM of all factors is given in Table 2; furthermore weights for all factors were calculated by Principal Eigen vector sum to 1 with acceptable CR of 0.10, as shown in Table 3.

Multi-Criteria Decision Analysis
MCE was implemented by WLC using Idrisi, with additive weighting concept based on weighted average of the criteria.Total score is obtained by multiplying weights to the scaled value of each criteria and then summing the products over all attributes.
In final step constraint modifies the procedure by multiplying the suitability calculated from the factors by the product of the constraints.This overlay capability of GIS allows evaluation of criterion maps into a final composite map, by following mathematical formula: Where, S is composite suitability score, w i is weight of each factor i, x i is criterion score of factor i, c j is criterion score of constraint j, ∑ is sum of weighted factors, and Π is product of constraints [18].The result of aggregation of weighted factors and constraint using WLC approach by our GIS-based MCDA model to create a map of situational weight-based sites is shown in Figure 4, which demonstrates areas with favorable conditions ranked as suitable while unfeasible areas ranked as not suitable.

RESULTS AND DISCUSSION
According to our final rice land suitability map for Sindh, as shown in Figure 4, obtained by integrating geographical data and decision maker's preferences, area of 4,940,961.6230345hectares (35.6%) was permanently suitable for rice cultivation with production capability of 17,333,892.17tonnes and area not suitable was 8,938,120.3180755hectares (64.4%).
According to Agriculture statistical report of Pakistan [2], rice planting area in Sindh was 746,091 hectares in 2013-14 with 2,617,300 tonnes production.According to this report, existing rice cropland area covers 5.4% while area not under rice cultivation was 94.6% of Sindh.Our final rice land suitability map presented a potential increase in suitable area of 4,194,870.623hectares (30.2%) with an increased potential in production of 14,716,592.17tonnes.This difference in the area and production of actual and potential land showed capability of our novel model.
The final rice land suitability map was influenced mainly by agro-ecological zones, soil physical and chemical properties and soil pH.Results of this work were also in conformity with the findings of: Agriculture statistics of Pakistan [2], Minister for agriculture [19], Rice growing areas [20], Major rice growing areas [21], Rice production in Sindh districts [22], and Rice production regions [23] who concluded sites suitable for rice cultivation in Sindh, Pakistan.

CONCLUSION
Land suitability is a complex process that could significantly impact the profit and loss of principal investments of the country.Proper land evaluation is most important concern in agriculture to enhance crop production by utilizing feasible potentials.The application of our proposed GIS-based MCDA model was successful in providing a powerful tool to evaluate rice land suitability in Sindh, by transforming complex decision making problems into series of transparent steps.In this study it was observed that for suitability analysis, consideration of river constraint and factors: temperature, canal command area, soil physical and chemical properties, groundwater quality, soil pH and agro-ecological zones, with their suitability conditions, relative importance and weighting, were important in obtaining useful results having geographic precision.This model offered enhanced sustainable production by mapping accurate cropland with much required information for farmers and agricultural planners.

Figure 1 :
Figure 1: GIS-based MCDA model for evaluating rice land suitability in Sindh, Pakistan.

Figure 4 :
Figure 4: District-wise final rice land suitability map of Sindh, Pakistan, using GIS-based MCDA model.