Authors: Jose Castejon Villalobos, Anzy Lee, Noelle Patterson, Jared Stieve, Belize Lane, Colin Phillips – Utah State University Rebecca Diehl – University of Vermont
Title: A terrain-based metric for predicting HAND Flood Inundation Mapping Skill
Presentation Type: Poster Presentation
Abstract: Effective operational flood forecasting systems depend on accurate estimates of discharge and a means to convert these estimates into flood inundation extents. To serve this purpose, the terrain-based Height Above Nearest Drainage Flood Inundation Mapping (HAND-FIM) approach has gained prominence for its flexibility, ease and low computational requirements. However, despite these advantages, HAND-FIM has demonstrated limitations to generate accurate flood inundation extents in regions with highly complex terrain and for lower magnitude floods. What remains unclear is whether these limitations stem primarily from the underlying assumptions in HAND-FIM and terrain processing or from the influence of local channel characteristics. To address this gap, we evaluate the ability of HAND-FIM to produce flood inundation extents across a wide range of rivers and channel settings and introduce a terrain-based metric capable of predicting best possible HAND-FIM performance. We introduce a Signal to Noise Ratio (SNR) approach that identifies flow stages and channel settings where HAND-FIM generated maps might perform poorly relative to calibrated flow models. The method computes SNR values across the whole inundated domain and over multiple flow stages by extracting the signal (geometric mean of all depths) and the noise (geometric standard deviation of all depths) and combining these with river slope and the rate of change in inundated area with water depth. We then relate these SNR values to the maximum achievable Critical Success Index (CSI). Our findings show a consistent relationship where the maximum CSI increases with SNR. Low SNR and CSI values are observed where the scale of topographic variability rivals the flow depth and provides an explanation for HAND-FIM’s poorer performance for lower magnitude floods. This SNR framework offers a low-complexity and scalable diagnostic tool that can help identify channel settings, flow regimes and terrain conditions where alternative modeling approaches may be preferable for achieving more accurate FIM.