Modeling the observed angular anisotropy of land surface temperature in a Savanna

Several recent studies have found that retrievals of land surface temperature (LST) from remote sensing measurements depend upon the angle of observation. To understand, predict, and ultimately correct this sensitivity, simple but physically based models of LST angular anisotropy are needed. In this study, we describe and evaluate the modified geometric projection (MGP) model, a highly parameterized model of scene thermal infrared (TIR) radiance applicable to both homogeneous and discontinuous canopy environments. Based on geometric optics modeling, MPG assumes that the angular anisotropy of TIR radiance over discontinuous canopies is due strictly to the different proportions of scene endmembers (e.g., sunlit tree crowns, background shadows) visible to a sensor at different sun-view geometries. We tested MGP against DART, a rigorous three-dimensional radiative transfer model, and against field-measured data from a southern Africa savanna. For a prescribed set of canopy conditions, MGPs estimates of observable endmember fractions and scene temperatures in the solar principal plane compared well with estimates from DART. We also parameterized MGP with field-measured endmember data for an acacia/combretum savanna near Skukuza, South Africa. We angularly integrated the MGP-predicted radiances and compared the results with measurements of scene hemispherical exitance from a tower-based pyrgeometer. The modeled exitances exhibited the normal diurnal behavior. Model predictions generally agreed with the pyrgeometer measurements; however, model accuracy decreased as the difference in endmember temperatures increased. These tests suggest that the assumptions inherent in the MGP model do not seriously impact the accuracy of the simulated radiances. We conclude that the MGP model accurately captures the predominate thermal emission directionality resulting from discontinuous canopy structure, and could therefore be applied at continental and global scales.