Using the model simulation to improve the Land Surface Temperature retrieval for JPSS and GOES-R Missions

Satellite products are being increasingly used in weather and climate prediction systems. Currently satellite measurements over ocean have been successfully utilized to improve numerical weather prediction. However, the utilization of satellite data over land is far less than over ocean. Land surface temperature (LST) retrievals are more challenging than sea surface temperature retrievals because of surface heterogeneity and the difficulty in simulating surface emissivity. In this study, we attempt to improve LST retrievals and test new approaches using radiative transfer model simulations. The MODTRAN radiative transfer model is applied to simulate the brightness temperature at the top of the atmosphere. In the LST retrieval, some biases are caused from uncertainty of surface emissivity. We have analyzed 10 years MODIS emissivity data and considered seasonal variations of the emissivity within same surface type, and then built up the emissivity pairs for representing 17 IGBP surface types. Based on the simulated brightness temperatures corresponding to these emissivity pairs, we evaluated the VIIRS Split-Window algorithm. We have also compared several different LST retrieval algorithms including surface-type dependent and emissivity-dependent algorithms. The comparison results will be used for improving the LST retrieval algorithms for JPSS and GOES-R missions.