Abstract:
This study addresses the spreading disparity between global food demand and production, particularly focusing on optimizing agricultural processes in Dera Ismail Khan, Pakistan, with a specific emphasis on sugarcane. Leveraging ground survey data and Sentinel 2 satellite imagery, the research employs a two-step methodology, utilizing NDVI variations for sample selection and conducting a comprehensive field survey with the Kobo Toolbox. A total of 450 samples were collected, with 85% used for model training and 15% for validation. The methodology incorporates RF, Support Vector Machine (SVM), Classification and Regression Tree (CART), K-Nearest Neighbor (KNN) and K-Means clustering classification models, with SVM demonstrating the highest overall accuracy of 90% followed by Random Forest (RF) at 81.43%. The study identifies a total sugarcane area of 54,055 acres. A novel approach for sugarcane harvest monitoring is introduced, establishing a threshold NDVI value of 0.25 for identifying harvested areas, and providing real-time insights into cropping patterns. The study also addresses challenges posed by atmospheric conditions and advocates for integrating SAR data. The objectives include enhancing precision in sugarcane identification and mapping, improving growth and harvest tracking, and comparing classification models. The results showcase the potential of this SVM for sustainable crop monitoring, contributing to informed decision-making in agriculture and addressing the global food demand-production gap.
