Precipitation From a Multiyear Database of Convection‐Allowing WRF Simulations

Convection‐allowing models (CAMs) have become frequently used for operational forecasting and, more recently, have been utilized for general circulation model downscaling. CAM forecasts have typically been analyzed for a few case studies or over short time periods, but this limits the ability to judge the overall skill of deterministic simulations. Analysis over long time periods can yield a better understanding of systematic model error. Four years of warm season (April–August, 2010–2013)‐simulated precipitation has been accumulated from two Weather Research and Forecasting (WRF) models with 4 km grid spacing. The simulations were provided by the National Center for Environmental Prediction (NCEP) and the National Severe Storms Laboratory (NSSL), each with different dynamic cores and parameterization schemes. These simulations are evaluated against the NCEP Stage‐IV precipitation data set with similar 4 km grid spacing. The spatial distribution and diurnal cycle of precipitation in the central United States are analyzed using Hovmöller diagrams, grid point correlations, and traditional verification skill scoring (i.e., ETS; Equitable Threat Score). Although NCEP‐WRF had a high positive error in total precipitation, spatial characteristics were similar to observations. For example, the spatial distribution of NCEP‐WRF precipitation correlated better than NSSL‐WRF for the Northern Plains. Hovmöller results exposed a delay in initiation and decay of diurnal precipitation by NCEP‐WRF while both models had difficulty in reproducing the timing and location of propagating precipitation. ETS was highest for NSSL‐WRF in all domains at all times. ETS was also higher in areas of propagating precipitation compared to areas of unorganized diurnal scattered precipitation. Monthly analysis identified unique differences between the two models in their abilities to correctly simulate the spatial distribution and zonal motion of precipitation through the warm season. Key Points This is a unique study of precipitation properties from two convection‐allowing models over a long period (4 years) Monthly analysis revealed unique changes in diurnal and propagating systems throughout the warm season The selection of verification method is greatly dependent on the purpose of verification (climate versus operational forecasting)