Leveraging Probabilistic Forecasting to Prevent Flood-Related Fatalities


Graduate student, Sean Matus, using a Trimble R10 unit to survey a bridge in Colorado Springs, CO. Surveying allows for more accurate representative of these structures in our model simulations.

In June of 2016, nine United States Army soldiers were on a routine training exercise at Fort Hood, Texas. They made the decision to traverse a low water crossing on Owl Creek. Unbeknownst to them, that crossing was closed and under 7 feet of water. Tragically, their military vehicle was swept downstream and those nine soldiers perished. The Army’s Investigating Officer stated that procedural improvements could be made to better mitigate future risks. To address this problem, ATMS associate professor Dr. Francina Dominguez and PhD student Sean Matus are collaborating with the United States Army Corps of Engineers’ (USACE) Engineering Research Development Center (ERDC) to develop a flood prediction tool, the Hydrologic Risk Forecaster (HydroRF), to forecast hydrologic and hydraulic conditions at mission-critical Department of Defense (DoD) riverine infrastructure. Such locations are typically in remote locations with minimal observations.

HydroRF uses deterministic and probabilistic weather forecasts run by the National Centers for Environmental Prediction (NCEP) and United States Air Force (USAF), and dozens of hydrologic simulations are run in parallel to produce conglomerated ensembles of streamflow. Rating curves developed by the team transform streamflow into flow depth and velocity at individual river crossings. This workflow enables HydroRF to communicate flood hazard predictions, and the uncertainty of those predictions.

The accuracy of a flood forecast is irrelevant if it is ignored. As such, the HydroRF team is implementing a two-way communication strategy with its DoD stakeholders. HydroRF communicates model output graphically with a focus on condensing probabilistic information into simple, yet detailed figures. These figures are designed to be ever evolving. By working directly with the eventual model users, their feedback can be heard, and these figures will be tailored to their needs. This way, HydroRF can be most effective in providing water managers with the information required to implement preventative action and ultimately save lives.


The proof-of-concept for HydroRF was published in the Journal of Hydrometeorology: https://doi.org/10.1175/JHM-D-19-0238.1