Herein, we present the results of a series of numerical experiments using the Ex-BATS biosphere model, which is an adaptation of Dickinson's biosphere-atmosphere transfer scheme (BATS). These simulations are used to assess how the model performs when remotely sensed data are used to estimate three key canopy variables. These canopy variables, which effectively represent the slowly changing boundary conditions of a vegetated surface, consist of the total surface albedo, leaf area index, and the nondiurnally varying component of stomatal resistance, referred to as stressed stomatal resistance. The surface albedo is retrieved from NOAA-AVHRR (advanced very high resolution radiometer) channel 1 spectral reflectance information in conjunction with a directional reflectance model which accounts for the strong diurnal variations in surface reflectance. A 4-channel vegetation index also retrieved from AVHRR measurements is used to estimate the leaf area index. A similar index derived from high-resolution SPOT visible and near-infrared information has been used to describe the spatial variations in such indices which impact the retrieval of the leaf area index. Satellite retrieval of stomatal resistance is based on split-window skin temperatures from AVHRR channels 4 and 5 from the afternoon overpass (~1630 LT).

Variables derived from the hourly skin temperature observations of GOES-VISSR have also been examined with respect to retrieval of stomatal resistance. It was found that although stomatal resistance has little correlation with the diurnal amplitude of skin temperature, it is closely related to the daily maximum of skin temperature. Numerical experiments have been conducted to examine model sensitivity to each of these canopy variables. Results indicate that Ex-BATS is not sensitive to small variations of surface albedo or leaf area index within the range of estimation uncertainty. The rms measurement-model flux differences in every numerical trial were within 6 W m-2 of the rms differences obtained for the simulations performed using measured albedo and leaf area index. Measured stomatal resistance values were obtained using an inversion form of the model. The resulting stomatal resistances were used to perform a control experiment simulating an ideal satellite retrieval scenario involving one observation per day. The control experiment resulted in improvements of approximately 20 W m-2 in the rms flux differences over the model using a purely hypothetical formulation for stomatal resistance. Simulations using the remotely retrieved stomatal resistances produced significantly reduced rms differences for latent and sensible heat fluxes over the model using the hypothetical formulation. Based on a 55-day composite involving all days from the four FIFE intensive field campaigns, the sensible and latent heat flux improvements are approximately 25 and 20%, respectively (11 and 8 W m-2). The satellite retrievals are only 20 and 30% less accurate (7 and 10 W m-2) than the idealized results of the control experiment. ¿ American Geophysical Union 1993