In the first two decades of the 21st century (2001–2020), the global surface air temperature was 0.99°C higher than 1850–1900 (IPCC 2021). Global warming has become an accepted fact, particularly at northern high latitudes and high-altitude areas, such as the Rockies Mountains, the Alps, and the Tibetan Plateau(Hansen et al. 2010); (Yao et al. 2012). The Tibetan Plateau (TP), the inland plateau of Asia, is the largest and highest plateau in the world, with a mean altitude exceeding 4,000 m. It is described as the “Water Tower of Asia” because several major rivers originate from the TP, including the Yellow River, the Yangtze River, the Mekong River, and the Brahmaputra River (Yao et al. 2012).
The climate over the TP is complex and has intense sensitivity to climate change, primarily affected by the interaction of the Asian monsoon and the mid-latitude westerlies (Yao et al. 2019). The TP has experienced remarkable warming during the recent decades (Zou et al. 2014; Cai et al. 2017; Duan et al. 2020; Wang et al. 2021). Meteorological datasets indicate that the TP’s warming rate is twice of that observed worldwide in the last 50 years (Duan and Xiao 2015). The strongest warming occurred in the winter months (Du et al. 2004; Chen et al. 2006; Zou et al. 2014; Chen et al. 2016). Growing evidence reveals enhanced warming with elevation (Li et al. 2017) or elevation-dependent warming (Pepin et al. 2015) over the Tibetan Plateau. It is manifested that such inclinations may remain in the future, particularly in winter and spring (Liu et al. 2009; You et al. 2020). The TP has the largest glaciers other than the polar regions, as well as widespread alpine permafrost, snow cover, and lake ice (You et al. 2021). Significant glacier retreating, snowmelt, the degradation of permafrost, desertification, and vegetation change over the TP have been rapidly influenced by the amplified climate warming (Thakuri et al. 2019; Kuang and Jiao 2016; Wang et al. 2017; Gao et al. 2015). Hence, Future warming and its elevation-dependency are of significant impacts on the ecological environment and water resources over the Tibetan Plateau.
Global climate models (GCMs) are crucial tools for climate attribution and projection researches and have been widely applied to reproduce the historical climate and project the potential future climate change. Although climate models have experienced significant evolutions over the past decades, they still fail to capture the dynamics of a local scale process (e.g., cloud feedback, land-use changes, topography) because of their coarse spatial resolutions (Reichler and Kim 2008; Sharma et al. 2018; Gusain et al. 2020). Climate change projections have been reported in various study areas under different emission scenarios from the most frequent GCMs participating in the Coupled Model Intercomparison Project (CMIP) Phase 3/5 (Jia et al. 2019; Fahad et al. 2018; Gu et al. 2015; Chen and Frauenfeld 2014a, 2014b; Maloney et al. 2014; Kharin et al. 2013; Rogelj et al. 2012). Despite the improvements (including terrestrial and marine carbon cycles, dynamic vegetation and etc.) comprised in CMIP5 models (Jia et al. 2019; Fahad et al. 2018; Gu et al. 2015; Chen and Frauenfeld 2014a, 2014b; Maloney et al. 2014; Kharin et al. 2013; Rogelj et al. 2012), CMIP5 models still have pronounced cold biases in simulating surface air temperature over the Tibetan Plateau (Su et al. 2013; Chen and Frauenfeld 2014b).
Currently, the new sixth phase CMIP multi-model datasets developed by different institutions have become available. CMIP6 models display a substantial extension relative to CMIP5, owing to the remarkable changes in CMIP6 dynamical core structure, such as improved spatial and vertical resolutions, revised microphysics parameterizations, advanced deep convective schemes, modified ocean-ice models (Eyring et al. 2016; Gerber and Manzini 2016). The deficiencies identified in CMIP5 models are expected to be remedied or even overcome, notably in the TP. Shared Socioeconomic Pathway (SSP) scenarios are advocated for CMIP6 future projections, in place of the Representative Concentration Pathway (RCP) scenarios of CMIP5 (O'Neill et al. 2016; Taylor et al. 2012). SSP scenarios clearly describe and quantify both emission pathways and land-use changes (Riahi et al. 2017). Therefore, SSPs provide a more specific explanation of future socio-economic evolution and offer more realistic future climate projections.
Several recent studies compare the simulations and outputs of CMIP5 and CMIP6 (Tokarska et al. 2020; Su et al. 2021; Wang et al. 2021; Tebaldi et al. 2020; Chen et al. 2021). The performance of CMIP5 and CMIP6 models has been found to be site-specific and irregular (Song et al. 2021), and CMIP6 models exhibit better capabilities in historical climate simulations (Zhu et al. 2021; Zhu and Yang 2020; Lun et al. 2021; Zamani et al. 2020). The above studies primarily focused on the evaluation and comparison of the performance of CMIP5 and CMIP6 models in reproducing the historical climate. However, little achievements have been revealed for the future projections of surface air temperature over the TP under different emission scenarios from CMIP5 and CMIP6 models.
GCMs are widely used for climate projections, but are subjected to high uncertainty owing to standard errors mainly resulting from the model structures, scenarios and ensembles (Hawkins and Sutton 2011; Woldemeskel et al. 2014; Yip et al. 2011; Bennett et al. 2012). Several studies have demonstrated that models are the main source of total uncertainty, followed by emission scenarios (Jobst et al. 2018; You et al. 2021). Square Root of Error Variance (SREV) method proposed by Woldemeskel et al. (2012) can quantify such uncertainties in space and time. Woldemeskel et al. (2016) used the SREV to quantify and compare uncertainty in climate projections using 6 CMIP3 and 13 CMIP5 models. Kim et al. (2020) adopted the SREV to access the uncertainty in extreme daily precipitation from 45 projections of CMIP5. Eghdamirad et al. (2016) utilized the SREV to quantify the contribution of different sources of uncertainty in upper air climate variables. These studies manifested that uncertainty estimates using the SREV method have a high confidence level.
This study uses CMIP5 and CMIP6 experiments to perform an assessment of the capability of the models to simulate the surface air temperature over the TP and to quantify differences in their projections of future temperature changes for the near-term (2021–2040), mid-term (2061–2080) and long-term (2081–2100) future from spatio-temporal and elevational perspectives. The moderate and high scenarios (SSP2-4.5/SSP5-8.5 for CMIP6 and RCP4.5/RCP8.5 for CMIP5) are mainly focused on in this study. The key issues that we emphasize are as follows. (1) How do the CMIP6 models perform in simulating the historical surface air temperature over the Tibetan Plateau, whether CMIP6 models exhibit improvements over their CMIP5 predecessors? (2) What are the differences between CMIP5 and CMIP6 in the possible changes of the projected warming pattern over the Tibetan Plateau? (3) Does the Tibetan Plateau’s warming show dependency with elevation in the future projections? (4) For different elevation zones, are there any differences between CMIP5 and CMIP6 in uncertain estimates of surface air temperature over the Tibetan Plateau? This study can contribute to adjusting adaptive measures based upon RCP scenarios for the future natural ecosystem.