Land cover characterization for hydrological modeling using thermal infrared emissivities

Modeling the spatial distribution of evapotranspiration at landscape scales requires frequent monitoring of land cover conditions in order to accurately characterize the relative amounts of living vegetation, senescent vegetation, and bare soil surfaces. These three components constrain estimates of transpiration and partitioning of sensible heat flux. The usual remote sensing approach to perform this characterization is to create and synthesize visible - near infrared vegetation indices throughout a growing season. Such indices are usually excellent discriminators between living green vegetation and bare soils, but are less successful discriminators between senescent vegetation and soil. However, by utilizing multispectral thermal infrared observations from sensors such as ASTER and MODIS it is feasible to more accurately discriminate between all three land cover components. This capability is achieved because thermal infrared emissivities are sensitive to soil surface properties and to vegetation canopy density, but are entirely insensitive to plant greenness. Hence remotely sensed emissivities are complementary to conventional vegetation indices. Examples of the thermal emissivity approach will be illustrated using satellite data acquired over two areas in the U.S.: arid lands of southern New Mexico and grazing lands in central Oklahoma.