Remote Sensing and Machine Learning for War Impact Assessment on Agriculture in Ukraine

The ongoing war in Ukraine has profoundly disrupted the economy and posed significant threats to both national and global food security by impeding agricultural activities. Our research proposes a pioneering approach to systematically monitor and assess war-induced damages using satellite imagery and machine learning models. Specifically, we implement Random Forest classification and regression, in conjunction with statistical analysis of spectral bands and vegetation indices. This investigation introduces several key contributions, including the development of a comprehensive cloud-based geospatial system for quantifying the war's impact on agricultural lands in Ukraine, the generation of novel datasets for continuous monitoring, and innovative methods for automated detection of damaged areas based on Earth observation data.