Reconstructing complete MODIS LST based on temperature gradients in northeastern Qinghai-Tibet Plateau

MODIS LST products provided by NASA may suffer from missing values and noises from various sources, which can degrade the LST quality and hamper its efficient applications. The paper presents an algorithm to reconstruct complete LST image based on regression analysis of LST with elevation in each sliding window, after filtering low-quality and unreliable pixels. Comparison of reconstructed LST with meteorological temperature measurements (T) indicates that LST is significantly correlated with T with an average correlation coefficient of 0.96 and a mean absolute difference (MAE) of 2.02 K. LSTs and Ts show no significant differences at monthly and yearly scale. The differences between LSTs and Ts have certain correlations with their different spatial and temporal definitions; however some residual noises existing in the reconstructed LSTs indicate more meticulous algorithm needed to work out more accurate RS-LST data.