In the wake of the Kakhovka Dam destruction in Ukraine, a team of researchers led by Bohdan Yailymov from the Space Research Institute of the National Academy of Sciences of Ukraine has published groundbreaking research that leverages machine learning and satellite data to assess the impact on cropland irrigation. The study, published in the IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, offers a compelling narrative of the disaster’s far-reaching consequences, particularly for the energy and agricultural sectors.
The Kakhovka Dam, a critical infrastructure element, collapsed on June 6, 2023, unleashing catastrophic flooding. Yailymov and his team set out to quantify the damage, focusing on flooded areas and the subsequent impact on irrigation. Using a combination of Sentinel-1, Sentinel-2, and Landsat-9 images, they generated detailed maps of water bodies before and after the flood. The team employed advanced machine learning algorithms, including the random forest classifier for flooded area mapping and multilayer perceptron and U-shaped network classifiers for identifying irrigated lands.
The findings are stark. As of June 9, 2023, the total area of flooding under the Kakhovka Dam was 47.33 thousand hectares, affecting a diverse range of landscapes, including 1.67 thousand hectares of cropland, 0.97 thousand hectares of forests, 12.3 thousand hectares of grasslands, 1.85 thousand hectares of settlements, and 29.4 thousand hectares of wetlands. The analysis of irrigated areas revealed a dramatic decrease in irrigated cropland—from 351 thousand hectares in 2019 to just 38 thousand hectares in 2024.
“Our study highlights the systematic destruction of agricultural infrastructure, with far-reaching consequences for regional food security,” Yailymov noted. The disappearance of water in irrigation canals, a critical component of the region’s agricultural infrastructure, underscores the severity of the situation. The flood also affected areas along the Ingulets River, further exacerbating the impact on agricultural land and water quality.
The research not only provides a detailed assessment of the immediate impacts but also offers a framework for future disaster monitoring. The high classification accuracy achieved—90.4% overall accuracy with F1-scores of 90.4% for both irrigated and nonirrigated classes—demonstrates the reliability of satellite remote sensing and machine learning approaches in quantifying flood-related natural disaster impacts.
For the energy sector, the implications are significant. The destruction of the Kakhovka Dam and the subsequent flooding have disrupted not only agricultural activities but also the broader energy landscape. The study’s findings could inform future disaster response strategies, helping to mitigate the commercial impacts on energy infrastructure and ensuring more resilient systems.
As Yailymov and his team continue to refine their methods, the potential for these technologies to shape future developments in disaster monitoring and agricultural management is immense. The study, published in the IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, serves as a testament to the power of satellite data and machine learning in addressing some of the most pressing challenges of our time.