Dissertation Defense - Tianshu Yang - Department of Geography and Geoinformation Science

PhD Candidate: Tianshu Yang

PhD of Science, Earth Systems and Geoinformation Sciences
Department of Geography and Geoinformation Science 

Dissertation Title: Identifying Non-hazardous Floodwaters from the Operational VIIRS and GOES-R Flood Products

Dissertation Chair: Dr. Donglian Sun (GMU, Geography and Geoinformation Science) 

Committee Members: 

Dr. John Qu (GMU, Geography and Geoinformation Science) 

Dr. Ruixin Yang (GMU, Geography and Geoinformation Science) 

Dr. Viviana Maggioni (GMU, Department of Environmental and Water Resources Engineering) 

Abstract: 
With the increasing impact of climate change and the growing frequency of extreme weather events, large-scale flood occurrences are becoming more frequent and widespread. The flood detection algorithm and software, developed by George Mason University under the support of the NOAA JPSS Program Office, enables automatic flood detection from the Visible Infrared Imaging Radiometer Suite (VIIRS) sensor on board the polar-orbiting Suomi National Polar-orbiting Partnership (Suomi NPP), NOAA-20, and NOAA-21 weather satellites, and the ABI (Advanced Baseline Imager) from the GOES-R series satellites. The operational VIIRS flood products have been widely used by the National Weather Service (NWS), the Federal Emergency Management Agency (FEMA), and the International Charter Program for flood monitoring and relief and rescue efforts.

In order to create satellite-based flood maps, water bodies are first identified from satellite imagery. Floodwater can be detected by comparing pre-event and post-event water classification maps. However, in an operational environment, such comparisons are often impractical. Instead, operational flood products are generated by comparing water classification maps with permanent or normal water bodies. This approach may misclassify certain water bodies, such as paddy rice fields, wetlands, etc., as floodwaters because they are not considered as permanent or normal water. These kinds of floods don’t cause significant damage to properties or harm to life, therefore, there are some kinds of non-hazardous floodwaters, but may interfere with hazardous floods. To allocate limited resources to the real needed regions with sufferings, it’s important to distinguish these non-hazardous floodwaters from hazardous floods, especially for decision-makers to investigate disaster status. 

In this study, these types of non-hazardous floodwaters are further classified into four categories: paddy rice fields, seasonal water in wetlands, tidal flooding near riverbanks and coastlines, and other non-hazardous floodwaters, including possible fishponds. Methodologies combining long-term flood frequency, paddy rice phenological algorithms, and change detection analysis are developed. Through confusion matrix analysis with the USDA rice data, the paddy rice extraction algorithm achieved an accuracy of 93% and an F1-score exceeding 80%. In addition, the algorithm for extracting non-hazardous floodwaters in wetlands also demonstrated promising results, indicating an accuracy of 97.87%, with a precision of 79.99%, a recall of 81.47%, and an F1-score of 80.73%.