By inverting BRDF models against satellite-sensor measurements of spectral reflectance recorded at different illumination and view angles, it is theoretically possible to obtain estimates of several important environmental variables (e.g., LAI, albedo). In recent years, a growing number of studies has exploited this basic approach, however, the relationships between the resulting model-derived quantities and the corresponding surface properties have yet to be rigorously explored. This thesis examines these relationships, by comparing upscaled observations of LAI with the outputs of BRDF model inversions. The research makes use of ground-base measurements of LAI, albedo, spectral reflectance and canopy biometric characteristics for a variety of arable crops over two growing seasons (1999/2000). These data sets are, however, recorded at very different spatial and temporal scales compared to the satellite-sensor data employed as input to the BRDF models (c.1km). The former data set was, therefore, upscaled to match the satellite data. Upscaling is achieved using a combination of plant growth models and intermediate spatial resolution satellite-sensor images. The plant growth models, in particular SUCROS, are used to provide estimates of LAI over the full growing season. These temporal profiles of LAI are up-scaled to the 1km spatial resolution of the SPOT-VGT image data employed in the BRDF model inversions using image data acquired by the Lansat-TM sensor. These data are used to generate a land cover map of the study area (84% accuracy). The simulated temporal profiles of LAI are applied to this land cover map on a cover-type by cover-type basis to generate images of LAI, initially at 30m resolution and, subsequently, at 1km resolution. In theory, the kernel weights are related to surface biophysical properties, however, it was subsequently determined that, in the case of LAI, this was not so. Therefore, evidence that the kernel weights may be related to LAI is sought by comparison of the temporal profiles of up-scaled LAI with temporal profiles of up-scaled LAI with temporal profiles of the kernel weights. Although positive correlation with the other kernel weights. It was, however, found that the up-scaled LAI was strongly correlated with the NDVI and the corrected NDVI than with the kernel weights.