Introducing a new approach to temperature validation of remote sensing thermal images

Abstract

Assessing the accuracy of the Land surface temperature (LST) has been and remains a challenging task. Rapid LST fluctuations in time and space, and a spatial scale mismatch between satellite and terrestrial sensors, have made validation using terrestrial data challenging. In addition to using ground data, there are three techniques for evaluating this parameter: radiance-based validation, indirect assessment, and cross-validation between two sensors. One of the most prevalent approaches for measuring LST accuracy is cross-validation. The main requirements of the cross-validation approach are temporal, spectral, spatial, and sensor angle of view adaptation. A technique for cross-validating LST from Landsat 8 using MODIS sensors is provided in this research. MODIS ' temperature product was chosen as the reference since it was collected twice per day by each of their sensors. The suggested method's results revealed that the accuracy evaluation in regions with high homogeneity, using the parameters of mean differences and root mean square error, has an accuracy of 0.6 and 1.63 degrees Kelvin in the first study's image, respectively. These values, 0.94 and 1.27 kelvin, were likewise achieved in the image of the second research. The suggested approach is applicable to any thermal sensor at any time and place.