Uncertainty in River Forecasts and Implications for Emergency Management

Sunday, February 17, 2013
Auditorium/Exhibit Hall C (Hynes Convention Center)
Frauke Hoss , Engineering & Public Policy, Carnegie Mellon University, Pittsburgh, PA
Emergency Management Coordinators (EMCs) are responsible for their community’s preparation and response to floods. The daily flood forecasts from the National Weather Service (NWS) available to the EMCs are deterministic expected water levels for the next four days. Based on an historical analysis of flood events, the uncertainty of those forecasts can be several feet depending on the lead time and the water level. The forecasts also have an underestimating bias increasing with lead time. The uncertainty has not been decreased in decades because of factors in the hydrologic model that are difficult to predict. EMCs often have very limited or no experience with floods and the uncertainty in daily flood forecasts. Uncertainty thus needs more explicit attention from NWS. This research quantifies the uncertainty of flood forecasts for the Neosho River in Commerce, OK and the Chikaskia River in Blackwell, OK, explores its sources using regression analysis, and proposes a visual presentation of uncertainty and other decision aids. Additionally, the research shows that statistical post-processing of the forecast can improve its accuracy by removing its underestimating bias. This gives the EMCs more lead time to prepare. Finally, based on 17 interviews with EMCs, the research describes the use of flood forecasts in emergency management and the implications of uncertainty in flood forecasts.