Improvement of daytime land surface skin temperature over arid regions in the NCEP GFS model and its impact on satellite data assimilation

Comparison of the land surface skin temperature (LST) from the National Centers for Environmental Prediction (NCEP) operational Global Forecast System (GFS) against satellite and in situ data in summer 2007 indicates that the GFS has a large and cold bias in LST over the arid western continental United States (CONUS) during daytime. This LST bias contributes to large errors in simulated satellite brightness temperatures over land by the Community Radiative Transfer Model (CRTM) and hence the rejection of satellite data in the NCEP Gridpoint Statistical Interpolation (GSI) system, especially for surface-sensitive satellite channels. The new vegetation-dependent formulations of momentum and thermal roughness lengths are tested in the GFS. They substantially reduce the large cold bias of daytime LST over the arid western CONUS in the warm season. This, in turn, significantly reduces the large biases of calculated satellite brightness temperatures found for infrared and microwave sensors in window or near-window channels, so that many more satellite data can be assimilated in the GSI system. In the arid western CONUS, the calculation of surface emissivity for microwave sensors in the CRTM can be further improved, and the new microwave land emissivity model together with increased LST via changes in surface roughness length formulations reduces biases and root-mean-square errors in the calculated brightness temperature.