Excellence in Research: Mutlimodal Geospatial and Remote Sensing Data Fusion for Flood Mapping and Damage Assessment

Flooding is one of the most catastrophic and frequently occurring natural disasters, causing extensive damage to life, infrastructure, and the environment. The severity and frequency of floods have increased in recent years due to extreme weather events such as hurricanes and the expansion of urbanization. Accurate monitoring and mapping of flood extent and damage assessment in both spatial and temporal measurements are critical to assessing flood risk and developing comprehensive relief efforts immediately after flooding occurs.

Remote sensing data, including both optical and radar data, have increasingly been used to develop flood mapping and modeling in a cost-effective and efficient manner, as establishing and maintaining rain and stream gauging stations can be costly. Remote sensing data are effective for determining the spatial extent of coastal and river flooding, providing essential information for delineating flood-affected areas, assessing damage to infrastructure such as roads and bridges, and feeding models that predict vulnerability to flooding in both inland and coastal areas.

The recent proliferation of remote sensing platforms, such as satellites, aircraft, and UAVs, equipped with advanced sensor technologies like optical, SAR, and LiDAR, has enabled the systematic production of massive amounts of high spatial, spectral, and temporal data. This research develops a novel framework for automatically extracting spatio-temporal features using integrated data-driven analysis and generative models to create a comprehensive and detailed knowledge base of environmental dynamics for rapidly changing events like floods.

Additionally, this project develops a language-guided self-supervision fusion method for heterogeneous remote sensing data for efficient damage assessment. The self-supervised multisource visual question answering framework allows real-time communication between human and robot agents for rapid response and recovery after natural disasters. This enables effective monitoring and notification of flood states, identification of affected or at-risk areas and people, which are essential for rescue and disaster operations.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.