Modeling the Surface Energy Balance and Soil Moisture at the Atmospheric Radiation Measurement Cloud and Radiation Testbed in the U.S. Southern Great Plains

We determine the vegetation type for the CART from a database generated by the U. S. Geological Survey (USGS 1990). Soil types were provided by the Atmospheric Radiation Measurement (ARM) Program from the State Soil Geographic (STATSGO) database and mapped into the USDA classifications described by Clapp and Hornberger (1978). To provide the input parameters of temperature, pressure, humidity, wind speed, insolation, and precipitation needed to drive SiB, we use a multiquadric interpolation scheme (Nuss and Titley 1994) to grid measurements from the CART extended observing network, the Oklahoma mesonet, National Weather Service surface stations, and the Arkansas Basin River Forecast Center's NEXRAD Stage III precipitation product. SiB also requires a time-varying Leaf Area Index (LAI) to estimate the biophysical rates that control local evapo-transpiration. We derived estimates of the LAI using a method described by Sellers et al. (1994) which uses the Normalized Difference Vegetation Index (NDVI). We determined NDVI from AVHRR (advanced very high resolution radiometer) imagery provided by ARM. Although the results of our calculations appeared reasonable, i.e., regions with recent precipitation and large values of NDVI showed greater latent heat fluxes than dry regions with low NDVI values, comparisons with flux measurements from the CART Energy Balance Bowen Ratio (EBBR) stations exhibited considerable deviations from the calculations. The largest differences between modeled and measured fluxes arose from differences in land use: all of the EBBR stations are located in pastures, whereas wheat fields are the dominant land use type in some of the 6.25x 6.25-km cells containing the EBBR stations. For these locations, the spatially averaged NDVI was quite low since the wheat had been harvested 4-6 weeks earlier but the pastures were generally quite verdant. Consequently, the NDVI input to SiB was not representative of the measurement area, and the modeled fluxes accordingly differed from the measurements.