Real-Time Probabilistic Delineation of Flood Warning Polygons to Reduce Flood Alert Fatigue

Overview

86% of flood deaths since 1959 are from people driving or walking into floodwaters. As little as 6 inches of flood current can knock over an adult, while 12 inches can sweep a car away. As global warming increases the frequency and intensity of floods, governments are investing in flood-alert systems to mitigate preventable deaths. However, in regions like Central Texas, where floods claimed at least 134 lives in July 2025, residents have criticized flood warning systems for inaccuracy. False alarms erode public trust, causing a “boy-cried-wolf” effect making residents more likely to dismiss a warning to evacuate. Traditional alerting systems use meteorologists’ analyses of rigid weather forecasting models to hand-draw flood warning polygons. Often, entire regions are overwarned as a precaution. This research project utilizes a novel machine learning-based approach to generate more accurate flood warning polygons to minimize false alarms. This approach counters the “boy-cried-wolf” effect, inducing prompter responses in warned areas by increasing the public’s trust in flood warnings. Using topographical, atmospheric, soil-water saturation, historical flood coverage, and 12 additional data features, a multimodal dual-head Residual U-Net was trained on over 50,000 flood samples to predict, for a given 250m by 250m tile, the probability of a flood occurrence and its intensity. Compared to the average NWS-flood warning polygon, this model reduces the false warning area by 38.2%, while still encompassing over 99% of true flood coverage. Ultimately, this model’s precision may save lives and reduce property damage by restoring public trust in flood warnings.

Competition history

• CSEF 2026 Environmental Engineering (Senior Division) · Entry S-11-22

Related projects

Closest projects by meaning, across every fair and year in the corpus.

Browse more like this

• All Environmental Engineering projects
• CSEF Environmental Engineering projects
• Every CSEF 2026 project