HS Kamide Lecture Satellite Microwave Remote Sensing for Hydrological Applications in the Era of AI Foundation Models
Hyunglok Kim
Gwangju Institute of Science and Technology
Share:
Biography

Hyunglok Kim is an Associate Professor in the Department of Environment and Energy Engineering at the Gwangju Institute of Science and Technology (GIST) in the Republic of Korea, where he leads the Hydro AI Laboratory. His research integrates microwave remote sensing, UAV observations, machine learning, and land surface data assimilation for hydrological applications. He leads the NASA-GIST core validation site in Hampyeong and is developing a microwave foundation model for hydrology. He earned a Ph.D. in Civil Engineering in 2022 and an M.S. in Data Science in 2021, both from the University of Virginia. He was a NASA Future Investigator (FINESST) and a postdoctoral fellow at the USDA Hydrology and Remote Sensing Laboratory. He is an IEEE Senior Member and an Associate Editor of IEEE Transactions on Geoscience and Remote Sensing. He received the 2020 AGU Horton Research Grant.

Abstract

Satellites provide continuous global observations of the terrestrial water cycle, including soil moisture, vegetation water, surface water, snow, precipitation, evapotranspiration, and water storage. Artificial intelligence, particularly foundation models, is being developed to draw hydrological insights from these expanding records.Foundation models are being built using optical, radar, and atmospheric sounder observations alongside weather and climate reanalyses. Microwave satellites are especially important for hydrology because they sense soil moisture, vegetation water, snow, and surface water through clouds, day and night.Our research seeks to use satellite information more fully for hydrological understanding and prediction. We are developing a foundation model using microwave and other hydrological satellite observations and applying it to hydrological research. We also use existing satellite, weather, and Earth system foundation models to estimate and predict soil moisture, evapotranspiration, and precipitation, and to support drought prediction. Data assimilation using foundation models could overcome the limitations of assimilating satellite observations directly and improve hydrometeorological prediction skill.An equally important challenge is in-situ validation. Sites for evaluating such models and products are scarce, especially in Asia. We have therefore been working with NASA since 2023 to develop Korea's first core validation site, the NASA-GIST site in Hampyeong. It combines a dense in-situ soil moisture network, tower-based L-band radiometers, and periodic unmanned aerial vehicle (UAV) campaigns. UAV payloads include L-band radiometers; multispectral cameras; and light detection and ranging (LiDAR) sensors. These observations bridge the gap between point measurements and satellite footprints and are intended to support next-generation radiometer and synthetic aperture radar (SAR) satellites.Connecting in-situ, UAV, and satellite observations with foundation models could contribute to hydrological understanding and early warning in Asia-Oceania, where observations are sparse and water-related hazards are pressing concerns.

Session Details
Section Type
Lecture Type
HS HS Kamide Lecture