A water vapor scaling model for improved land surface temperature and emissivity separation of MODIS thermal infrared data

Abstract

Abstract We present an improved water vapor scaling (WVS) model for atmospherically correcting MODIS thermal infrared (TIR) bands in the temperature emissivity separation (TES) algorithm. TES is used to retrieve the land surface temperature and emissivity (LST&E) from MODIS TIR bands 29, 31, and 32. The WVS model improves the accuracy of the atmospheric correction parameters in TES on a band-by-band and pixel-by-pixel basis. We used global atmospheric radiosondes profiles to generate view angle and day–night-dependent WVS coefficients that are valid for all MODIS scan angles up to 65°. We demonstrate the effects of applying the improved WVS model on the retrieval accuracy of MODIS-TES (MODTES) LST&E using a case study for a granule over the southwest USA during very warm and moist monsoonal atmospheric conditions. Furthermore, a comprehensive validation of the MODTES LST&E retrieval was performed over two sites at the quartz-rich Algodones Dunes in California and a grassland site in Texas, USA using three full years of MODIS Aqua data. Results from the case study showed that absolute errors in the emissivity retrieval for the three MODIS TIR bands were reduced on average from 1.4% to 0.4% when applying the WVS method. A Radiance-based method was used to validate the MODTES LST retrievals for and the results showed that application of the WVS method with the MODTES algorithm led to significant reduction in both bias and root mean square error (RMSE) of the LST retrievals at both sites. When the WVS model was applied, LST RMSE's were reduced on average from 1.3 K to 1.0 K at the Algodones Dunes site, and from 1.2 K to 0.7 K at the Texas Grassland site. This study demonstrated that the WVS atmospheric correction model is critical for retrieving MODTES LST with