Scale Mismatch Implications for the Validation of Remote Sensing LST Products

The validation of remote-sensing land surface temperature (LST) can provide accurate feedback of the LST retrieving, which is known as the basis of various applications of LST products. The scale mismatch between LST observations of the remote sensors and in situ challenges feasibility and rationality of the conventional validation methods that directly use ground-measured “true” LST observing in small scope to evaluate remote sensed-based LST products for relative large space, especially validation for low or medium spatial resolution. To explore scale mismatch underlying effect on the conventional validation, the multi-scale ground-based in situ observations were compared with Moderate Resolution Imaging Spectroradmeter (MODIS) daily LST products (MOD11A1) with a spatial resolution of 1 km. The comparison results at daytime indicates that stronger heterogeneity leads to a greater scale mismatch uncertainty, which demonstrates that the scale mismatch can lead large errors to the validation process. Based on semivariance, the LST heterogeneity and structure of the MODIS 1 km mixed pixels were analyzed by the thermal TASI and ASTER images which are with a spatial resolution of 3 m and 90 m, respectively. The variogram-based results indicate that average scales from the structural and heterogeneous also depend on the spatial resolution of the analyzed images.