In addition to the design of a comprehensive land surface model for mesoscale modeling, we also need to develop many other realistic physical parameterization for the model atmosphere. For example, in addition to the ground wetness, soil and vegetation types, the land surface temperature can be affected by cloud shadow, long wave reflection from cloud, surface radiative cooling, evaporative cooling, downdraft outflow, topographical flow, low level advection, cool pool, surface heat, moisture and momentum fluxes, non-local PBL mixing, mesoscale circulation (e.g., dryline, sea breeze and land breeze) and synoptic system etc. To predict the surface temperature correctly, we need to parameterize the aforementioned physical processes successfully.

For example, to parameterize cloud-radiation, we modified a cloud radiation model used in general circulation model for efficiency. We also used an accurate method to compute solar zenith angle according to the formula used in the 2002 Astronomical Almanac. The difference can be as much as 80 watts/m2. But what is so good about a realistic cloud-radiation scheme if there is no cloud at initial time and the model spin-up time is long? So we have developed a cloud initialization scheme based on satellite, radar and surface observation information. Examples will be shown to demonstrate the impact of cloud initialization on the prediction of a sharp gradient of surface temperature.

We are using a modified Noilhan and Planton (1987) scheme for our land surface model. However, since some of the vegetation parameterizations are quite arbitrary, we are implementing the Common Land-Surface Model (CLM) into our mesoscale model. CLM has been developed recently by NASA, NCEP, Office of Hydrology and several Universities. We will report about how to simplify the CLM for a mesoscale numerical prediction model at the Conference.

The vegetation and soil type maps with resolution of 1km or less over US is available now. Similar resolution over the whole globe is under development for the Common Land-Surface Model. These maps need to be used very carefully as they are still full of obvious mistakes. Several soil types or vegetation types with appropriate portions are allowed at the same grid point. In addition, bi-weekly satellite derived vegetation coverage and leaf area index data has been used. The 12th hour surface temperature forecast looks better.

The Land Data Simulation System (LDAS) provides soil moisture and temperature suitable for “real time” forecasting. With this data as first guess field, we can make use of the Oklahoma Mesonet and OASIS project data for nudging method to produce initial fields. The Oklahoma Mesonet and OASIS data are suitable for initializing, tuning and validating land surface and surface layer models. Moreover, the OASIS network measures surface heat, moisture and radiative fluxes directly.