Geostatistical solutions for downscaling remotely sensed land surface temperature

Abstract

Remotely sensed land surface temperature (LST) downscaling is an important issue in remote sensing. Geostatistical methods have shown their applicability in downscaling multi/hyperspectral images. In this paper, four geostatistical solutions, including regression kriging (RK), downscaling cokriging (DSCK), kriging with external drift (KED) and area-to-point regression kriging (ATPRK), are applied for downscaling remotely sensed LST. Their differences are analyzed theoretically and the performances are compared experimentally using a Landsat 7 ETM+ dataset. They are also compared to the classical TsHARP method.