<p>Over the past several decades, global climate change, particularly the rising temperature has caused public concerns. In the context of climate warming, many environmental and water problems such as decreasing runoff, shrinking glaciers and permafrost, vegetation degradation and desertification can be attributed to rapid climate change. Surface air temperature (SAT) plays a key role in land-atmospheric interactions and is an important parameter for climate change studies. Traditional SAT data are collected by ground meteorological observation. Nevertheless, such traditional measurements at ground stations cannot capture the spatial variations of SAT, especially over complicated areas such as the Tibetan Plateau, where meteorological stations are with large elevation variability and unreasonable spatial distribution. In contrast, satellite remote sensing provides an direct observation of land surface temperature (LST) and, thus, also provides an possible way to obtain SAT since LST and SAT are generally closely related to each other. The scientific communities have developed various methods to estimate SAT from LST through statistical or physical models. The widely used satellite LST, however, is derived from satellite thermal infrared remote sensing and thus, significantly affected by the clouds.</p><p>In this study, we report an examination of the estimation of daily 1-km SAT from the all-weather satellite LST over the Tibetan Plateau. The estimation of SAT is based on a noval method that dynamicall integrates the newly published 1-km all-weather LST data by merging satellite thermal infrared and microwave remote sensing observations based on the random forest. The matchups of the ground measured SAT at stations and the corresponding all-weather LST were separated into the training set and valiation set. In addition, independent SAT measured at experimental ground sites were used to evaluate the SAT method. Results indicate that reasonably integrating multiple LST terms provides daily average all-weather SAT with satisfactory accuracies over the Tibetan Plateau. The estimated SAT based on the proposed method has ignorable systematic error and low root-mean squared error when validated with ground measured SAT under all-weather conditions. Further comparison demonstrates that the SAT estimate agree well with other SAT estimated from satellite thermal infrared LST under cloud-free condition. In addition, the SAT method has the potential to be generalized and extended to various complicated areas. With this method, the daily 1-km SAT for the entire Tibetan Plateau from 2003 to 2018 were produced. This dataset is of great value to examine recent climate warming trend and the land-atmospheirc interactions in the entire Tibetan Plateau.</p>
ESTIMATION OF MONTHLY NEAR SURFACE AIR TEMPERATURE USING GEOGRAPHICALLY WEIGHTED REGRESSION IN CHINA
