LAND SUITABILITY ANALYSIS FOR OLIVE AND MAIZE CROP IN RAINFED AGROFORESTRY SYSTEM USING MACHINE LEARNING ALGORITHM

Abstract

The demand for edible oil is on the rise due to an increase in the population and affordability all over the world. But such an increase in the demand for a nation like Pakistan is a big problem by inducing pressure on the import bill and increasing the trade deficit. In our study, we focus on land suitability analysis of olive and maize crops in an Agroforestry system based on the Food and Agriculture Organization (FAO) “land suitability assessment framework” by using machine learning and traditional technique. The soil, climatic and topographic data were collected in district Dir Lower of Khyber Pakhtunkhwa to identify suitable areas for the agroforestry system of olive and maize crops in the rainfed agriculture areas. after determining the land suitability classes for the agroforestry system of olive and maize crops, they were classified and maps were produced through a machine learning algorithm of (RF, SVM) and by weighted overlay technique. The ML-based random forest (RF) gives overall accuracy and kappa index of (0.94, 0.90) for olive and (0.94, 0.91) for maize while the support vector machine (SVM) gives the overall accuracy and kappa index of (0.92, 0.88) for olive and (0.93, 0.90) for maize. On the other hand, the weighted overlay (WOL) technique gives the overall accuracy and kappa index of (0.89,0.85) for olive and (0.93, 0.87) for the maize land suitability classes. The traditional technique o WOL doesn’t predict the permanently non-suitable class for the maize while the ML-based technique did it. The maps produced by the two different techniques depict clear and prominent differences.