Spatial-Temporal Adaptive Land Surface Temperature Data Fusion By Combining TsHARP Model And STITFM Algorithm

Land surface temperature (LST) is an important parameter that describes energy balance of substance and energy exchange between the surface and the atmosphere, and LST has widely used in the fields of urban heat island effect, soil moisture and surface radiative flux. Traditionally, acquisition of land surface temperature data is mainly through the way of fixed point measurement, which has high precision but the observation area is small. The development of satellite remote sensing technology makes it possible to obtain the global land surface temperature data. One of the effective ways to get large-scale LST is quantitative retrieval by using thermal infrared remote sensing data. Landsat 7 satellite provides a thermal infrared band with spatial resolution of 60m, but the long revisit cycle and clouds effects lead to less availability of data, which greatly limits the application in temporal analysis. Due to its one-day revisit frequency, MODIS thermal infrared image has great advantages for continuous observations of land surface temperature. However, its low spatial resolution of lkm provides less detail information, facing more serious problem of spatial scale. Currently, no satellite sensor can deliver thermal infrared data at both high temporal resolution and spatial resolution, which strongly limits the wide application of thermal infrared data. Spatial-temporal fusion of remote sensing data regarded as one of the effective methods to solve this problem.The combination of thermal sharpening (TsHARP) model and the spatio-temporal integrated temperature fusion model (STITFM), namely CTSSTITFM is proposed in this study. Firstly, the TsHARP method is used to downscaling the lkm MODIS land surface temperature image to LST data at spatial resolution of 250m. Then the accuracy is verified by the retrieval LST from Landsat ETM+ image at the same time. Finally, fine resolution of land surface temperature image is predicted by fusing Landsat ETM+ and MODIS data using STITFM. The estimated LST from Landsat ETM+ data for the same predicted date is used to validate the accuracy of the predicted LST. The results are shown as follows:(1) Comparing to the traditional fusion methods of STITFM, the CTSSTITFM method has a better precision. In the default parameter settings, the predicted LST values using CTSSTITFM fusion method have a root mean square error (RMSE) less than 1.33K.(2) By adjusting the window size of CTsSTITFM fusion method, we found that the fusion results in the selected areas show some regularity with the increasing of the window. In general, a reasonable window size set might slightly improve the effects of LST fusion.(3) The CTsSTITFM method can solve the mixed pixel problem caused by the coarse MODIS LST image to some degree.