Abstract:
Agriculture is the biggest contributor to Pakistan’s GDP that makes around 22.67 percent to the total GDP. Cotton production alone accounts for 4.1 percent of Pakistan agriculture, 0.8% of GDP and for roughly 60% of Pakistan's international earnings. The fact that crop yield per hectare in Pakistan is less than its competitors call for methods like integration of GIS with Machine Learning to assist farmers and policymakers in making decisions for sustainable cotton growth. This study aims to integrate GIS with Remote Sensing data and Machine Learning algorithms to: (i) Map the cotton covered region. (ii) Predict cotton yield. (iii) Project the long-term impacts of climate factors (maximum & minimum temperature, and precipitation) on cotton yield. Sentinel 2A Imagery was imported in GEE to extract specific ranges of five vegetation indices in cotton field and apply these ranges to other time periods to delineate cotton covered region from all other. Then 11 vegetation indices and climate factors were used to model and estimate cotton yield. ALM, GLM, RF, GBT and SVM models were used to compare their results on the study area and provided data. Lastly, CMIP6 future projections of temperature and precipitations were used to correlate them with the vegetation indices and find out the pattern of crop yield from 2023 until 2099. The result not only showed the use of GIS, Remote Sensing and Machine Learning to map cotton fields, model cotton yield, but also emphasized the need for mitigation and adaptation for climate change to save cotton crops for better crop management practices.
