Verification of experimental ensemble forecasts of precipitation and streamflow produced by the U.S. National Weather Service (NWS) [INVITED]

The River Forecast Centers (RFCs) of the U.S. National Weather Service (NWS) produce ensemble forecasts of temperature, precipitation and streamflow at a variety of lead times to support a wide range of hydrologic and water resources applications. As part of the Experimental Ensemble Forecast System developed by the NWS, ensemble traces of precipitation and temperature are generated from single-valued forecasts using an Ensemble Pre-Processor (EPP). These traces are input into the Ensemble Streamflow Prediction (ESP) subsystem of the Community Hydrologic Prediction System (CHPS), from which ensemble traces of streamflow are output. In order to account for the residual sources of bias and uncertainty in the streamflow forecasts (i.e. those not explicitly modeled via EPP and ESP), a statistical bias-correction is applied, which is implemented in the Ensemble Post-Processor (EnsPost). There is a need to verify the forecasts from EPP and ESP, both with and without EnsPost, in order to establish a baseline for improved forecasting techniques, and to identify the key factors responsible for model error and skill in different situations. Verification is required at multiple temporal and spatial scales, ranging from minutes and kilometers (e.g. for flash flood guidance) to years and entire regions (e.g. for water resource planning and national verification). This paper presents some verification results from the EPP and ESP subsystems (with and without EnsPost), identifies some of the factors responsible for model error and skill in different situations (in terms of various attributes of forecast quality, such as reliability and discrimination) and identifies several areas where future enhancement should lead to more reliable and skilful ensemble products.