Temporally Extended Satellite-Derived Surface Air Temperatures Reveal a Complete Warming Picture on the Tibetan Plateau
Jun Qin,Min He,Wei Yang,Ning Lu,Ling Yao,Hou Jiang,Jin Wu,Kun Yang,Chenghu Zhou
DOI: https://doi.org/10.1016/j.rse.2022.113410
IF: 13.5
2023-01-01
Remote Sensing of Environment
Abstract:The Tibetan Plateau (TP) is the highest plateau in the world, which imposes the intense thermal and dynamical forcings on the atmosphere and then impacts the climate in its surroundings. The TP has been undergoing a rapid warming, which accelerates glacial melting, and causes more natural hazards. Although the warming on the TP has been widely investigated, there is no complete picture of its thermal status during the past decades for lack of high-quality, long-term, spatiotemporal-continuous observations. The number of weather stations are rather limited and are mainly located in the east of the TP. The analysis based on these stations is confronted with the spatial representativeness problem. On the other hand, Satellites can monitor the earth seamlessly in space and time, but reliable land surface temperatures (LSTs) have only been available in recent 20 years, and moreover their physical meaning also differs from that of the most commonly used surface air temperatures (SATs). For climate change research, the period length of these satellite LSTs is too short to obtain a definitive conclusion. In this study, the entire algorithm consists of two primary steps. One is to develop an stacking-based ensemble learning algorithm to convert LSTs to SATs with the random forest model as both base learner and meta learner. The other is to construct a Bayesian-based temporal extension algorithm to merge satellite SATs and station SATs to obtain long-term, spatiotemporal-continuous SATs. After validating the reliability of these SATs and the warming trends based on them, 60 years (1961-2020) of SATs on the TP are implemented to examine the warming status of the TP. The spatial pattern of temperature trends illustrates that the warming occurs almost everywhere on the TP, and lots of areas with intensive warming, cannot be detected only based on station ob-servations, especially in the western part of the TP. Similarly, ERA5-Land and CRU datasets underestimate the warming in these areas. The newly-derived warming rate arrives at 0.03 degrees C/year and is 50% greater than those computed based on ERA5-Land and CRU dataset, implying an unexpected severe threat to the cryosphere.