Reduction of NCEP Global Forecast System 2-m Temperature Forecast Errors

Abstract:

In this study the systematic deficiencies and cause of errors in 2-m temperature forecasts in the NCEP Global Forecast System (GFS) are identified by investigating the physics of the Noah land surface model and land-atmosphere interactions, and a practical solution is found to reduce this kind of forecast errors. This presentation focuses on further evaluation of the proposed modifications with two one-month experiments for summer and winter seasons through the verification of GFS forecasts against surface and sounding observations. It was found that the modifications can substantially avoid late afternoon rapidly dropping 2-m temperature and decoupling when a cessation of turbulent transport between the surface and the atmosphere due to high near surface atmospheric stability happens, and reduce the cold bias of 2-m temperature during nighttime. Furthermore, the surface dew point temperature, surface wind speed and scores for light and medium precipitation are also improved. In the future, new land data sets such as vegetation and soil types, near real-time green vegetation fraction and snow albedo will be updated and we expect to further reduction of 2-m temperature bias in the GFS model.