An Evaluation of QPF from the WRF, NAM, and GFS Models Using Multiple Verification Methods over a Small Domain
The ARW model was run over a small domain centered on Iowa for 9 months with 4-km grid spacing to better understand the limits of predictability of short-term (12 h) quantitative precipitation forecasts (QPFs) that might be used in hydrology models. Radar data assimilation was performed to reduce spinup problems. Three grid-to-grid verification methods, as well as two spatial techniques, neighborhood and object based, were used to compare the QPFs from the high-resolution runs with coarser operational GFS and NAM QPFs to verify QPFs for various precipitation accumulation intervals and on two grid configurations with different resolutions. In general, NAM had the worst performance not only for model skill but also for spatial feature attributes as a result of the existence of large dry bias and location errors. The finer resolution of NAM did not offer any advantage in predicting small-scale storms compared to the coarser GFS. WRF had a large advantage for high precipitation thresholds. A greater improvement in skill was noted when the accumulation time interval was increased, compared to an increase in the spatial neighborhood size. At the same neighborhood scale, the high-resolution WRF Model was less influenced by the grid on which the verification was done than the other two models. All models had the highest skill from midnight to early morning, because the least wet bias, location, and coverage errors were present then. The lowest skill was shown from late morning through afternoon. The main cause of poor skill during this period was large displacement errors.