What AI Sees from Space: Skoltech Researchers Explore Capabilities and Limitations of Models for Monitoring Natural Threats

Researchers at the Skoltech Center for Artificial Intelligence (part of the VEB.RF group) conducted an analysis of AI models used for monitoring natural threats from satellite imagery. The study examined how these models can be applied in Earth observation to detect and track phenomena such as wildfires, floods, and other environmental hazards. The findings reveal that while AI shows promise in automating and improving the accuracy of threat detection, there are notable limitations, including data availability, model generalization across diverse geographic regions, and the need for high-quality training datasets. The researchers emphasized that addressing these challenges is crucial for the reliable deployment of AI-based monitoring systems. They also discussed potential future directions, including the integration of multi-source satellite data and the development of more robust algorithms to handle varying atmospheric and lighting conditions.