The University of Alabama - Integrating Geographical Hidden Markov Tree with the FIST Model towards Operational Observation-based Flood Inundation Mapping - Model Integration and ArcGIS Tool Development

(NOAA Collaborators:Sam Contorno)

This project aims to develop a computational tool for observation-based flood inundation mapping in an operational setting. While techniques for flood mapping from earth imagery advance over the years, significant challenges remain due to view obstructions (cloud, tree canopies) for optical and SAR sensors and limited high-quality imagery coverage in certain flood areas. This is particularly true for higher spatial resolutions of aerial and commercial imagery products. In addition, an effective and efficient tool that can produce or refine high quality observation-based flood inundation maps can help validate and calibrate the parameters of the NOAA National Water Model, which is currently only validated and calibrated by observations from over 7000 river gauges. Our new tool uses geospatial machine learning techniques called geographical hidden Markov tree to holistically incorporate earth imagery spectral information and the Earth’s surface topography at the same time. We will implement the algorithms into an ArcGIS tool so that it can be used in an interactive manner. The PI Dr. Jiang will collaborate with the NOAA National Water Center in this research.