Using Ensemble Precipitation Forcing to Communicate the Uncertainty of Hydrologic Event Occurrence at River Infrastructure

Current hydrologic risk management on United States military installations heavily depends on in-situ observations to support decisions to close river crossings. Although decision makers have high confidence that observations are accurate, observations describe conditions in real-time thus constraining the time decision makers have available to prepare. A tragic example of how reactionary decision support can lead to disaster is the training fatalities of nine soldiers at a low water crossing on 2 June 2016 at Fort Hood, Texas. The Investigating Officer for the incident states in the executive summary that, “although not directly causing this incident, procedural shortcoming at the installation, company, and platoon levels–specifically, training on local hazards, driver’s training, and low water crossing hazard reporting–should be improved to more effectively mitigate risks in the future.” Deterministic forecasting methods can provide decision makers the lead times needed (12-24 hours) to implement a proactive risk management approach, but with greater uncertainty compared to observations. It is proposed that a probabilistic forecasting approach can provide decision makers hydrologic predictions that are at both sufficient lead time and accuracy. The probabilistic approach uses ensemble precipitation forecasts to drive hydrologic (HEC-HMS) and hydraulic (HEC-RAS) model simulations for each forcing ensemble member. The forcing uncertainty across ensemble members, described by the variability of precipitation spatially and temporally, “trickles-down” through the hydrologic and hydraulic models to produce ensembles of streamflow, flow velocity, and flow stage at strategic river crossings. Two hydrologic events that resulted in loss of life at Fort Hood low water crossings are hindcasted with the probabilistic methodology. The NCAR 3-km Ensemble is used as forcing in anticipation of using the operational version of the 3-km HRRR-Ensemble. The ensemble of predicted streamflow is compared to calculated return periods for the catchment to serve as an initial warning that a severe hydrologic event is in the forecast. For a more detailed description of the hydrologic risk, the ensembles of predicted flow velocity and depth are compared to safety standards for specific vehicles through rating curves generated per low water crossing. The results are intended to easily communicate to decision makers both hydrologic risk through the magnitude of predicted values and uncertainty of those predictions through ensemble member agreement. With this methodology, decision makers can quickly be informed of the probability of a hydrologic event occurring and the hydraulic implications of that event on river infrastructure trafficability, potentially saving soldier lives.