Land surface models (LSM) are used at a variety of resolutions in an effort to recreate the water and energy balances at the earth surface. Correct representations of these balances are critical for accurate weather and climate forecasts. Unfortunately, when used as a part of a general circulation model (GCM) to help predict weather and climate, LSMs are usually run at coarse resolutions. This results in a loss of accuracy of the description of the land surface and meteorological variables that are used by the LSM. Our goal is to assess how much error one can expect when moving from high to low resolution. Our investigation looks separately at the influence of the land surface parameters (based on vegetation and soil types) and the meteorological variables.

Our studies have used LSMs (such as Mosaic and CLM) run offline in the LDAS framework. Already we have seen that switching from 1/8 degree to 1 degree resolution induces error in various variables, such as latent heat and runoff. Also this error manifests differently depending up location in LDAS domain (roughly the United States). Between these two resolutions we have also seen that the degradation of the land surface description induces more error between simulations than the reduction of resolution of meteorological forcing variables. We have begun to determine if this error can be reduced by introducing a tiling or mosaic approach. We are also currently using a lake model in CLM to determine the error for the grid box associated with ignoring small bodies of water.