Atmospheric sensitivity on land surface temperature retrieval using single channel thermal infrared remote sensing data: Comparison among models

Atmospheric effect is one of the most important error sources of land surface temperature (LST) retrieval using single channel thermal infrared (TIR) remote sensing data. This paper focuses on the theoretical analysis of the atmospheric correction error sensitivity on LST retrieval, which applied on a new TIR remote sensing data of HJ-1B satellite. For this purpose, atmospheric effects on three different methods for LST retrieval are analyzed and compared: the radiant transfer method (RTM), the mono-window algorithm (Qin method) and the generalized single-channel method (JM&S method). The MODTRAN 4 codes are compiled in order to simulate the atmospheric radiance in different situations. Differential equations are built from each LST model by converting the derivatives of atmospheric parameters, then the change rate of the model with respect to the atmospheric variables are exhibited. The results show that for all methods, the increasing of LST error is ensued from the increasing of water vapor content and viewing angle, and aerosols effect on LST is negligible for all methods; JM&S method is most sensitive to water vapor content uncertainty followed by JM and Qin methods, with an error on LST of 0.8K, 0.6K and 0.4K, respectively, on typical condition that water vapor uncertainty is 0.15 g cm
⁻²
with mid-latitude summer atmospheric; for a view angle of 33°, error on LST is lower than 2.3K, 1.9K and 1.1K for JM&S, RTM and Qin method, respectively; Qin method is also sensitive to mean atmospheric temperature, which leads to an error on LST of 0.5K, assuming an atmospheric temperature uncertainty of 0.1K; JM method is still sensitive to effective wavelength, which leads to an error on LST of 0.3K, assuming an wavelength uncertainty of 0.3μm; Hence, JM&S method is more suitable for the condition that precise water vapor profiles are applied, while Qin method can be more precisely if atmospheric temperature is fully estimated. For non-mathematical simplification, RTM could be the most accurate and complex choice if in situ measurement and angular correction of atmospheric are undertook.