Forecasting Potential Wheat Cultivation Zones In Zakho District, Northwest Iraq Using Remote Sensing Data And Machine Learning Techniques

Abstract

This research highlights the significant challenges of accurate wheat area mapping in the Zakho district of the Kurdistan region of Iraq, through the development of an accurate wheat classification and mapping system by using the integration of optical and radar imagery from Sentinel 1 and Sentinel 2 datasets, geographic information system (GIS), and machine learning algorithms. The Support Vector Machine (SVM) were used in this research to classify wheat and non-wheat areas utilizing multi-source satellite imagery, with combination of normalized difference vegetation index (NDVI) and Ratio Vegetation Index (RVI) of optical imagery, along with VV and VH polarization bands in radar data. 54 wheat farm data collected as a sample to be used as a training data for SVM method and validation to create wheat map for the study area. The results indicated that Zakho district are covered by 126,486.07 donums of wheat which is 9.2% of total area of the district in 2023. The accuracy of the classification was validated using field-collected ground truth samples, achieving an overall accuracy of 95%, and further cross-validated by comparison with official agricultural directory records. The research addressed the significance of using multiple combinations of optical and radar datasets with GIS, remote sensing and machine learning techniques to precisely map crop and agricultural farms in complex geopolitical regions. The developed maps provide accurate evidence to support the rights of farmers, promoting informed decision-making and equitable resource distribution. The outcome of this research can be used in future agricultural assessments, resulting in economic stability and sustainable development