| Description |
In August 2018 and June 2019, NCEP upgraded the operational versions of the High-Resolution Rapid Refresh (HRRR) and Global Forecast System (GFS), respectively. To inform forecasters and model developers about the capabilities, limitations, and biases of these modeling systems over the contiguous western U.S., we validate and compare quantitative precipitation forecasts (QPFs) produced by the experimental HRRRv3 and GFSv15.0 with the then operational HRRRv2 and GFSv14 during the 2017/18 cool season. We also compare the GFSv14 and GFSv15.0 with the operational, high-resolution configuration of the ECMWF Integrated Forecast System (HRES). We validate at lower elevations using observations from Automated Surface Observing System (ASOS) stations and at upper elevations using observations from SNOw TELemetry (SNOTEL) stations. We identify little or no change in overall HRRR precipitation bias or forecast skill as evaluated by traditional metrics. The GFSv14, GFSv15.0, and HRES all produce forecasts with a wet bias at lower elevations and neutral or dry bias at upper elevations, reflecting insufficient terrain representation. Forecast skill for large events increases from the GFSv14 to the GFSv15.0 to the HRES. Furthermore, as forecast lead-time increases, the improvement of GFSv15.0 over GFSv14 increases and the separation between GFSv15.0 and HRES decreases. In summary, HRRRv3 produces similar, and GFSv15.0 more skillful, QPFs over the iv western continental U.S. compared to their predecessors, but GFSv15.0 skill continues to lag HRES. |
