An Effective Interpolation Method for MODIS Land Surface Temperature on the Qinghai–Tibet Plateau

Land surface temperature (LST) is a key parameter in the processes of energy and water exchange between land and atmosphere. However, the MODIS LST products are often obscured by clouds and other atmospheric disturbances, resulting in severe data loss. Traditional interpolation methods cannot be effectively applied when there is large area of missing data. Thus, in this study, an effective LST interpolation method is developed to address this issue and is used to interpolate MODIS/Terra LST data on the Qinghai-Tibet Plateau in 2005. This method assumes that some pixels with spatially valid LSTs may follow a change trend over time similar to the null pixels, and the focus thus becomes to locate those similar pixels to interpolate each null pixel. First, LST images with a small amount of missing data, chosen as the reference images, were interpolated using a traditional interpolation method. Then, for each null pixel, other pixels with similar temporal changes of LST were identified by a similarity function. Finally, a transfer function for each null pixel was established based on those pixels most similar to it in the interpolated image and the corresponding reference image. The results were found to be much superior to those interpolated by traditional methods, such as regression Kriging, ordinary Kriging, and IDW. A specially designed experiment on an area that had ample valid LSTs confirmed that the proposed method can produce more favorable results than the other methods, and performed especially well when there was a significant lack of data.