Soil-Vegetation-Atmosphere-Transfer (SVAT) models inherently work at the patch scale. The tool of wavelet multiresolution analysis is applied to remotely-sensed data collected at multiple resolutions over the Southern Great Plains 1997 (SGP97) data set. This analysis reveals vegetation and surface radiometric temperature length scales on the order of 800-1000 m. Wavelet cospectra between fractional vegetation, temperature and the SVAT modeled surface fluxes are shown to correlate well to this length scale for high resolution data (12 meter). When applied to coarser resolution data (1000 m), no dominant length scale is observed. The implications for aggregation of remotely sensed data is discussed.