High-Resolution Flood Hazard Simulation of the Red River of the North Using a Single-Pass InSAR DEM

In this study, we used a unique opportunity of the GLISTIN-A (NASA airborne mission designed to characterizing the cryosphere) track to Greenland to acquire a high-resolution InSAR DEM of a large area in the Red River of the North Basin (north of Grand Forks, ND, USA), which is a very flood-vulnerable valley, particularly in spring time due to increased soil moisture content near state of saturation and/or, typical for this region, snowmelt. Having an InSAR DEM that meets flood inundation modeling and mapping requirements comparable to LiDAR, would demonstrate great application potential of new radar technology for national agencies with an operational flood forecasting mandate and also local state governments active in flood event prediction, disaster response and mitigation. Specifically, we derived a bare-earth DEM in SAR geometry by first removing the inherent far range bias related to airborne operation, which at the more typical large-scale DEM resolution of 30 m has a sensor accuracy of plus or minus 2.5 cm. Subsequently, an intelligent classifier based on informed relationships between InSAR height, intensity and correlation was used to distinguish between bare-earth, roads or embankments, buildings and tall vegetation in order to facilitate the creation of a bare-earth DEM that would meet the requirements for accurate floodplain inundation mapping. Using state-of-the-art LiDAR terrain data, we demonstrate that capability by achieving a root mean squared error of approximately 25 cm and we then illustrate its applicability to simulating past flood events in the basin.