Estimation of Land Surface Temperatures From Hyperspectral Infrared IASI Sensor

Infrared atmospheric sounding interferometer (IASI) optical sensor operates in the infrared spectrum and offers fine spectral samplings of the atmosphere in the infrared band between wavelengths of 645 and 2760 cm(-1). A physical model was proposed to estimate land surface temperatures (LSTs) from IASI observations in the 645-1600 cm(-1 )spectral range. In this method, the radiative transfer equation (RTE) was linearized into a form related to LST, land surface emissivity (LSE), atmospheric equivalent temperature and water vapor content considering the spectral characteristics of the IASI sensor. To reduce the number of unknown variables in RTE, LSE was expressed as a piecewise linear function. Regularization algorithms and iterative algorithms will further update the initial estimate LSTs and improve the retrieval accuracy of LSTs. This algorithm was tested on both simulated and real data obtained from the IASI sensor. The root-mean-square error (RMSE) of the simulated LST was improved (1 K) compared to the initial estimation of artificial neural network (ANN) method (2 K). This algorithm was also applied to real IASI daytime and nighttime observations and showed that it is capable of retrieving LST with retrieval accuracy similar to that achieved by the Advanced Very High Resolution Radiometer onboard MetOp (AVHRR/MetOp) LST product.