Recovering Land Surface Temperature Under Cloudy Skies Considering the Solar-Cloud-Satellite Geometry: Application to MODIS and Landsat-8 Data

Clouds play a significant role in the derivation of land surface temperature (LST) from optical remote sensing. The estimation of LST under cloudy sky conditions has been a great challenge for the community for a long time. In this study, a scheme for recovering the LST under cloudy skies is proposed by accounting for the solar-cloud-satellite geometry effect, through which the LSTs of shadowed and illuminated pixels covered by clouds in the image are estimated. The validation shows that the new scheme can work well and has reasonable LST accuracy with a root mean square error<4.9K and bias<3.5K. The application of the new method to the Moderate Resolution Imaging Spectroradiometer (MODIS) and Landsat-8 data reveals that the LSTs under cloud layers can be reasonably recovered and that the fraction of valid LSTs in an image can be correspondingly improved. The method is not data specific; instead, it can be used in any optical remote sensing images as long as the proper input variables are provided. As an alternative approach to derive cloudy sky LSTs based only on optical remote sensing data, it gives some new ideas to the remote sensing community, especially in the fields of surface energy balance.