The last decade has seen an increase in the application and complexity of mesoscale models. The Regional Atmospheric Modeling System (RAMS) has been applied at scales simulating flow around buildings at 1.0m grid sizes to regional climate simulations encompassing continental scales. Coinciding with the expanded range of applications the sophistication of the model physics has increased. These include the explicit treatment of microphysics, long and shortwave radiation parameterizations, advection schemes, new techniques for the assimilation of meteorological variables, and new parameterizations for treating the land-atmosphere continuum. One area that has not kept pace is the initialization of the land surface models (LSM). A promising technology is emerging, called the North American Land Data Assimilation Systems (NDLAS).

Two separate NLDAS research efforts are currently ongoing at the NASA Goddard Space Flight Center; a continental scale 1/8th degree project, and a 12km project on the Arakawa E grid (NLDAS-E). A joint effort between NASA GSFC, NOAA NCEP, NOAA OHD, Princeton University, Rutgers University, the University of Washington and the University of Maryland, the 1/8th degree NLDAS program has examined issues ranging from model intercomparison and validation, to hydrologic prediction and snowfall assimilation. Participating land surface models include the Mosaic, Noah, Sacramento, CLM, and VIC LSM. Advances have been made in the improvement of model parameterizations, in the understanding of model spinup processes, and in the production of high quality real-time and retrospective land surface fields. Applications have ranged from the initialization of climate models, to assistance in the prediction of West Nile outbreaks. In this study the NLDAS model derived fields are used to initialize the LSM contained in the RAMS model for long and short-term integrations.

In both sensitivity experiments the output of NLDAS derived soil moisture and temperature, as well as leaf area index (LAI) will be used to initialize the RAMS LSM. The first experiment examines the impact of soil memory on regional climate simulations. The RAMS model was configured to match the grid structure of the NLDAS 1/8th degree grid. The model was then integrated for 3 months with and without the initialization of the LSM using the NLDAS derived fields. A third integration again uses NLDAS to initialize the LSM and also uses the bottom soil temperature and moisture fields to nudge the bottom boundary of the RAMS soil model. The LAI was also updated with the NLDAS values of LAI. Preliminary results indicate a significant improvement in simulated meteorological conditions when using the NLDAS initialization procedure. The decrease in domain biases was persistent throughout the entire 3-month integration. In addition, improvement in spatial correlations with observed meteorology is noted. The use of the bottom soil boundary conditions and LAI was also found to improve the model performance.

In the other simulations, the RAMS model was configured to investigate the impacts of the initialization on the genesis and evolution of a severe line of thunderstorms in the summer of 1998. The model was integrated for a period of 3 days. The model grid in this case employed 2km horizontal grid spacing on a fine nest, which was embedded in a mesh with 6km horizontal grid spacing. Finally, these grids were contained within a grid with an 18km mesh increment. Again the initialization used the default procedure in RAMS as well the new procedure employing the NLDAS derived fields. In this case the fields were only used as initial conditions. Here it was found that the integration using the default initialization in RAMS produced no organized convection with sporadic cells throughout the domain. When NLDAS data was used, the formation of the line, and its subsequent evolution, was found to simulate the actual events of this summertime case study. Coinciding with an improvement in the simulation of convection was the improvement in surface simulated winds and temperatures throughout all three grids.

This study was conducted to investigate the value added by initializing an LSM using data supplied by NLDAS. We chose to perform the experiments over a wide range of temporal and spatial scales. It was found that improvements were noted over all of the simulated scales. It was also found that the use of NLDAS to force LSM boundary conditions also was beneficial to model performance. We conclude that NLDAS shows promise and convenience to the users of mesoscale atmospheric modelers in providing the data necessary to initialize LSM. There are also ongoing developments that will make this data available globally at 1km mesh sizes in real-time (see http://lis.gsfc.nasa.gov), as well as plans to incorporate additional LSM.