Limits to hydrologic predictability: lessons for high-resolution forecast systems

Verification studies evaluating deterministic Quantitative Precipitation Forecast (QPF) have shown QPF to have considerable error when evaluated against observed precipitation. Hydrologic forecast error resulting from the use of QPF is significant, at levels that bring into question the overall utility of using deterministic QPF for flood forecasting. The operational implementation of ensemble hydrologic forecasting methodologies are now widespread, developed, in part, to deal with model forcing errors, such as observed with QPF. However, a fundamental problem exists related to the predictability of QPF and, consequently, with hydrologic forecasts. The research presented demonstrates that with uncertainties in the spatial placement of precipitation forecasts alone, expected hydrologic forecast errors can range over several orders of magnitude, both with current levels of QPF accuracy and with substantial improvements in QPF. The limitations of QPF accuracy portends limits to hydrologic predictability. This fact has repercussions on the use of hydrologic ensemble methodologies and with efforts to make hydrologic predictions at increasingly finer spatial scales. The implications are especially evident with continental scale hydrologic prediction systems. Model errors, independent of forcing errors, compound the problem of hydrologic forecast uncertainty.