A generalized split-window algorithm for retrieving land surface temperature from GF-5 thermal infrared data

GF-5 is the 5th satellite of China High-resolution Earth Observation System (CHEOS) and is designed with 4 thermal infrared (TIR) bands with a spatial resolution of 40 m at nadir, which is the highest one in the current civil satellites. According to the excellent spatial resolution of GF-5, it is definitely true that the LST estimated from GF-5 will obviously promote the development of the relevant studies, such as the city thermal environment detection, resources exploration, energy and water cycle monitoring and so on. In this paper, by considering the advantage of split-window algorithm in estimating LST from the space measurements, a Generalize Split-Window (GSW) algorithm is then developed in this paper to accurately estimate the LST from the GF-5 thermal infrared data. After comparing the performance of different band combination, the band 11 and band 12 are finally selected for building the GSW algorithm. To acquire the coefficients of the GSW algorithm, all the clear atmospheric profiles in TIGR2000 are selected and then used for driving the MODTRAN. More than that, a turbulence on LST is also included in the simulations for obtaining a more comprehensive dataset. The results show that, with the increase of the water vapor content, the uncertainty of the GSW algorithm for GF-5 is from 0.19 to 0.94 K for the simulated data. To evaluate the performance of the GSW algorithm, six standard profiles in MODTRAN are picked out together with 65 typical surface materials from the ASTER spectral library. According to the results, the overall Root Mean Square Error (RMSE) is approximately 0.5 K under the condition of no errors. When all kinds of errors are present, the performance will be then decreased to some extent. However, it is true that the overall accuracy will still less than 1 K when the WVC less than 3.5 g/cm
²
.