Forecasting Inundation Extents using REOF analysis (FIER) over Greater Houston in Texas

Abstract:

As a victim of Hurricane Harvey in 2017, we understands firsthand that a forecasted flood extents is incredibly vital information for communities and emergency responders to reduce property damage and save lives. Although Harris County Flood Control District (HCFCD) operationally provides real-time inundation maps based on in-situ gauge data, there is no operational inundation forecasting system for the Greater Houston area. It is also recalled that the in-situ gauge in Barker Reservoir, located upstream of the Buffalo Bayou flowing through the heart of Houston, has been blacked-out for 5 days during Harvey, giving no indication how fast the reservoir was being filled up to Houstonians.

Recently, a novel approach for inundation extent forecast, which implements Rotated Empirical Orthogonal Function (REOF) analysis on historical multi-temporal Synthetic Aperture Radar (SAR) images to extract spatiotemporal patterns, has been developed and demonstrated over the Tonle Sap Lake Floodplain in Cambodia (Chang et al., RSE, 2020). The approach, called Forecasting Inundation Extents using REOF analysis (FIER), utilizes regression analysis between temporal patterns of historical SAR image stack and hydrological discharge or water level data. Through the regression models, forecasted discharges can then be used to estimate forecasted temporal patterns which are to be integrated with spatial patterns to synthesize SAR intensity images. Finally, the forecasted inundation maps can be generated using a water classification technique.

In this study, using the Tropical Storm Imelda that lashed Houston with more than 50 inches of heavy rain in September 2019 as a test case, we apply FIER to the Greater Houston area using Sentinel-1 images and historical and forecasted (1) discharges from the National Water Model (NWM) and (2) river levels from the West Gulf River Forecast Center (WGRFC) of NOAA, and generate forecasted inundation extents. The forecasted inundation maps are compared with independently-derived inundation maps from spaceborne SAR and optical imagery, and historical inundation maps available from HCFCD. It is expected that the application of FIER to urban environment can be scalable to other flood-prone cities in US and around the world, and significantly contribute to disaster response and risk reduction.