USING SPATIAL CONTEXT IN SATELLITE DATA TO INFER REGIONAL SCALE EVAPOTRANSPIRATION

Several methods have been used to estimate regional scale evapotranspiration from satellite thermal infrared measurements. These procedures assume knowledge of surface properties such as surface roughness, albedo, vegetation characteristics, etc. In many areas of the earth the separameters are not accurately known due to the rapidity of change of vegetation, lack of a dequate geographical data bases, and the low spatial resolution of satellite data which results in multiple surface types corresponding to each satellite measurement. In this paper, an estimate of evapotranspiration is developed by relating variations of satellite-derived surface temperature to a vegetation index computed from satellite visible and near infrared data. The method requires independent estimates of evapotranspiration for a completely vegetated area and for a nonvegetated area, although such areas need not appear in the satellite data. A regional estimate of evapotranspiration is derived despite the lack of precise estimates for individual satellite measurements. The method requires spatial variability in the satellite data: it does not apply in uniform areas. In addition, a propertv is identified which permits discrimination of cirrus clouds from areas of varying soil moisture. A data bank of surface characteristics should be developed to support description of surface processes at large scales. © 1990 IEEE