Authors: Yu-Fen Huang (UHM), Mirce Morales Velazquez (UVM), Yinphan Tsang (UHM), Beverley Wemple (UVM), Hannah Lohman (RTI), Schuyler DeBree (RTI), Katie van Werkhoven (RTI)
Title: What drives flash flooding in mountainous watersheds? A case-study approach from Hawaiʻi to Vermont
Presentation Type: Poster Presentation
Abstract: Flash flood forecasting in small, steep mountainous watersheds remains one of the hardest problems in operational hydrology, hindered by rapid watershed response, fine-scale sensitivity to rainfall errors, and high uncertainty in quantitative precipitation forecast (QPF). Within the CIROH project – Pathways to Increase the Effectiveness of Flash Flood Forecasts in Mountain Regions, using recent high-impact events as case studies, we are investigating which storm and antecedent characteristics most strongly control flood severity, and therefore, which must be adequately represented by alternative forcing and forecasting approaches.
We present a preliminary analysis of the 23 March 2026 Mānoa Valley flood (Oʻahu, Hawaiʻi), a ~11 km² tropical watershed that received 765 mm of rainfall over 13 days from back-to-back storms. At USGS gauge 16241600, located at the downstream watershed, peak discharge during the second storm reached 92 cms — 2.6x the first storm’s peak despite comparable event totals — with only 20 minutes separating the maximum rainfall from peak streamflow. Co-located soil moisture observations confirm that antecedent saturation prevented infiltration, causing rainfall during the second storm to translate almost entirely into runoff. We are examining MRMS gridded precipitation, USGS discharge, and local soil moisture observations to characterize the processes that drove the flood and evaluate the operational forcing’s ability to resolve that sub-hourly intensity. We will replicate the analysis at Ranch Brook, a ~10 km2 mountainous watershed in Vermont with similar remote and in-situ observations, to test transferability across physiographic settings.
Planned next steps include coupling the observed forcings with pyMARRMoT or RavenPy model structures to explore how sensitivity to rainfall intensity, spatial distribution, and antecedent conditions varies across model formulations. Findings will inform the development of low-latency, high-resolution alternative flash flood forecasting approaches within the broader CIROH effort. Feedback from the community on case selection, candidate model structures, and evaluation of metrics is welcomed.