IMPROVEMENTS IN THE SPLIT-WINDOW TECHNIQUE FOR LAND-SURFACE TEMPERATURE DETERMINATION

Land surface temperature (LST) retrievals obtained from NOAA Advanced Very High Resolution Radiometer (AVHRR) are of considerable importance for climatic research. However, the accurate evaluation of LST from s pace has been severely limited because of the difficulty in separating atmospheric from surface effects as the surface cannot be modeled as a black-body radiator. With this goal in mind, a novel extension of th e split-window technique is presented in which the atmospheric contrib ution to the radiance measured by the satellite is investigated by the ratioing of covariance and variance of the brightness temperatures me asured in Channels 4 and 5 of AVHRR/2. Furthermore, the contribution o f emissivity is evaluated from coefficients that depend on the spectra l emissivities in both thermal channels. Using a wide range of simulat ions from an atmospheric radiative transfer model it is shown that the proposed algorithm provides an estimate of LST, to within 0.4 K if th e spectral surface emissivity is known, which is better than that give n by the currently used split-window algorithms for LST determination. Also the limitations on algorithm accuracy are discussed considering different values of noise equivalent temperature. Finally we present t he preliminary results obtained using the proposed method from AVHRR d ata over a semi-arid region of Northwestern Victoria in Australia prov ided by CSIRO, and a mountainous region of Northeast of France acquire d in the frame of Regio Klimat Projekt.