GeoAI for Natural Disaster Assessment

Climate change has led to a sharp increase in the number and severity of extreme events, such as floods, tornados and wildfires. These events have resulted in adverse effects on human lives and the infrastructure. Swift disaster assessment is crucial for the effective planning of disaster response and relief efforts. AI and big data have provided unprecedented opportunities to enable swift disaster assessment, but two significant hurdles exist: (1) the scarcity of annotated geospatial data to train AI models, and (2) the lack of AI solutions that encode physics knowledge in a geospatial context. My research aims to address both challenges by developing an active-learning-based annotation platform that improves the annotation productivity of geospatial data for geospatial machine learning, and by developing physics-guided machine learning models for accurate natural disaster assessment.