WorldClim global high-resolution climate gridded data with a spatial resolution of 2.5 min (∼0.04°) (http://worldclim.org; Hijmans et al., 2005) was used in this study. It includes the data in the mid-Holocene (MH, refers to ∼6 cal ka BP in the study), the present-day (PD, 1960–1990), and the projected future. The projected future refers to the average of 2041–2060 (represented by 2050s). The PD data were interpolated from in-situ data. The data for the MH was downscaled from the Coupled Model Intercomparison Project 5 (CMIP5), which includes nine models of monthly precipitation and maximum and minimum monthly temperature. For the 2050s, the monthly precipitation, monthly maximum, and minimum temperature data from 17 climate models under the Representative Concentration Pathway (RCP) 4.5 and RCP8.5 scenarios were downscaled to 2.5 min resolution. The data of the MH and 2050s were downscaled and calibrated using the PD data as baseline “current” climate. Thus, the downscaled data by the same eight models for both the MH and 2050s are selected. The ensemble mean of the eight models’ data is used for analysis, which is usually more reliable than a single model (Lambert and Boer, 2001).

The unique geographical location and surface characteristics of TP lead to substantial uncertainties in climate simulations (Figure 1). The Asian Precipitation Highly-Resolved Observational Data Integration Towards Evaluation (APHRODITE) dataset, including precipitation (V1101, 1960–1990) and temperature (V1808, 1961–1990), is applied to evaluate the reliability of WorldClim data in this study, which has a spatial resolution of 0.25° (Yasutomi et al., 2011; Yatagai et al., 2012). It is recommended to analyze the TP’s climate (Salunke et al., 2019; Tan et al., 2020). The temperature and precipitation results from WorldClim PD’s data and the APHRODITE dataset show similar spatial distributions (Figures 2A–D). The major differences in mean annual temperature between the two datasets are distributed in the western Himalayas, where WorldClim underestimates the temperature. The annual precipitation in Karakoram is underestimated by WorldClim (Figures 2A–D). Overall, WorldClim datasets could reproduce the TP’s mean annual temperature and precipitation well. Notably, the study aims to compare the climate types in TP among different warm periods, which means the accuracy of the spatiotemporal distribution is more important than absolute values. The vegetation data of the manual digitalization of the 1:1,000,000 vegetation map of China (The China Vegetation Atlas Editorial Committee and Chinese Academy of Sciences, 2001) is utilized to contrast the climate types in TP because the 1:1,000,000 vegetation map of China has more detailed vegetation resources than other global remote sensing data, which covers most of TP (more than 83%).

The range of average annual temperature of TP in the three warm periods is from −8°C to 20°C (Figure 3). There is a gradient of increasing values from northwest to southeast, with the highest temperature in the southern margin, which is closely linked to the altitude (Figures 1, 3). The spatial temperature distribution in TP is similar between the MH and PD. TP is predicted to be warm substantially in the 2050s, with the 0°C isotherms moving toward the northwest and the average annual temperature of the southern TP changing from below 0°C to above 0°C. The 0°C isotherms shift northwestward in the RCP8.5 scenario than in the RCP4.5 scenario (Figures 3C,D). The average annual precipitation in TP ranges from 100 mm to 1,000 mm, decreasing from the southeast to the northwest (Figure 4).

To further understand the spatiotemporal characteristics of the average annual temperature and precipitation of TP in the MH, the PD, and the projected future, the difference between temperature (Figure 5) and precipitation (Figure 6) for each period were analyzed. It should be pointed out that the significance test is inapplicable to our calculation because only monthly climatology data exists in WorldClim dataset for the three periods. Compared with the PD, the average annual temperature in the MH shows overall negative anomalies with a meridional gradient and a maximum negative center in the southern TP (Figure 5A). Compared with the PD, the average annual temperature will increase by more than 2°C in the entire TP in the 2050s under the RCP4.5 scenario. The most pronounced warming appears in the western TP (around Gangdise Mountains and Qiangtang Plateau) with high altitudes (Figures 1, 5B). The warming is more pronounced under the RCP8.5 scenario, which increases about 0.5°C–1°C higher than the RCP4.5, coupled with the same phenomenon at high-altitude areas (Figures 1, 5C,D). In summary, the MH under natural conditions shows a meridional gradient difference in average annual temperature, while increased greenhouse gas emissions caused by anthropogenic changes could lead to more substantial warming of TP in the 2050s, with higher temperature increase at the high-altitude areas.

Precipitation also changes evidently among the three periods. Compared with the PD, the average annual precipitation in the MH increases by more than 300 mm in the southwestern TP, while precipitation decreases in the northwestern TP and northeastern TP and enhances the disparity between southern and northern TP (Figures 4A, 6A). Average annual precipitation in TP is predicted to increase in most of TP in the 2050s under the RCP4.5 scenario by about 20–60 mm compared with the PD. The increasing precipitation is most pronounced in the southeastern TP and southwestern TP (Figure 6B), which may attribute to the enhanced monsoon. The average annual precipitation differences between the PD and RCP8.5 scenario are similar to that of the RCP4.5 scenario (Figures 6B,C). The average annual precipitation in TP will increase by about 0–20 mm under a higher emission scenario in most of TP in the 2050s, except the boundary region of TP, where the average annual precipitation is projected to decrease by about 0–20 mm (Figure 6D). Besides, the distribution of precipitation increase is consistent between the RCP4.5 and RCP8.5 scenarios except for the Pamirs and southeastern TP margins (Figure 6D), indicating the homogeneity effect of the greenhouse gas emissions on increasing precipitation in TP.

Based on the information about the Köppen-Trewartha climate classification listed in Table 1, the climate types in TP during the MH, the PD, and the 2050s were calculated using monthly average temperature and monthly precipitation. The areal percentages of the major climate types for each period were calculated and listed in Table 2. Most regions of TP are occupied by the dry climate and polar climate, accounting for more than 70% of the area of TP (Figure 7; Table 2). In terms of the spatial distribution of climates during the MH, the dry climate occupies most of the northern part of TP, and the polar climate is mainly distributed in the hinterland and southwestern region of TP. The subtropical climate, temperate climate, and subpolar climate are primarily distributed in the southeast and western margins (Figure 7A). The distribution of major climate types in the PD is analogous to the MH, except for the polar climate replaced by the dry climate in the hinterland of TP (Figure 7B). In the context of intensifying climatic warming in the 2050s, the areas of polar climate will shrink. The subtropical climate, temperate climate, and subpolar climate move northward and northwestward towards the hinterland (Figures 7C,D). The distribution of climate sub-types shows that the dry climate is mainly composed of BSk (semi-arid climate, fewer than 8 months >10°C) and BWk (arid climate, fewer than 8 months >10°C) (Figure 8). The dry climate area in the northern TP and hinterland TP are classified as BWk and BSk, respectively. The sub-types of polar climate are mostly an Ft (tundra climate), and subtropical climate is a Cwa (subtropical winter dry climate, warmest month >22°C). The sub-subpolar climate is mostly an EO (subpolar oceanic climate) in the low-elevation areas of the southeastern TP and mainly an EC (subpolar continental climate) in the area extending to the hinterland of TP (Figures 1, 8). Except for the BWk and BSk, major warming in the 2050s will result in large changes to these sub-types, such as DCwb (temperate continental climate, dry winter, warmest month <22°C), DOwb (temperate oceanic climate, dry winter, warmest month <22°C), EO and EC. Compared with the PD, the temperate climate and subpolar climate will extend towards the hinterland of TP, and the proportion of these sub-types is predicted to increase in the 2050s.

Figure 9 evaluate further the differences between climate types in TP during the selected warm periods. Compared with the PD, a large proportion of dry climate in the hinterland TP mainly changes to a polar climate, while the areas of temperate climate, subpolar climate, and polar climate increased slightly during the MH (Figure 9A; Table 2). In the 2050s, it is worth noting that tropical climate will appear on TP (Table 2). In addition, a large area in the southern TP will undergo a change from a cold climate type to a warmer climate type in the 2050s, with an apparent shrinkage of the polar climate (Figures 9B,C; Table 2). Compared with the RCP4.5 scenario, climate types show unnoticeable change. A slight transition occurs mainly from a cold to a warmer climate type, especially for the polar climate (Figure 9D). Different scenarios have little impact on the dry climate area (Table 2). It is concluded that natural conditions resulted in a decrease in the dry climate regions in the MH, while anthropogenic change will promote a cold climate type to a warmer one in the 2050s.

The change in temperature and precipitation can both lead to the transformation of climate types. The precipitation increased by more than 300 mm in the southwestern TP between the MH and the PD, and 2°C warming over TP in the 2050s. We found that the area experiences a dramatic change in climate types in the MH when using the PD’s precipitation and MH’s temperature (Table 3). It indicates that precipitation changes dominated the shift of climate types in the MH. By contrast, a great change in climate types will occur in the 2050s when using the precipitation of PD and the temperature of the 2050s. It demonstrates that temperature changes dominate the 2050s evolution of climate types across TP. It is also helpful to understand that the dry climate will expand slightly in the 2050s, which is inconsistent with the trend of warming and wetting in TP (Table 2). Strong warming plays a more crucial role than a slight increase in precipitation for the climate types of changes in the 2050s. Notably, the increasing temperature may lead to nearly 20% of the area of climate types across TP changing by the 2050s under the RCP8.5 (Table 3). In conclusion, precipitation changes are primarily responsible for the difference in climate types between the PD and MH, and temperature will lead to changes in the 2050s.

To further investigate the relationship between Köppen-Trewartha climate types and vegetation in TP, we extracted the mega vegetation groups to compare with major climate types in the PD (Figures 2E,F, 10). The results show that the dry climate regions mainly correspond to steppes and deserts areas in the northern TP; polar climate regions correspond to the meadows areas in the eastern TP and southwestern TP; subpolar climate regions correspond to the shrubs areas in the southeastern TP; the regions of temperate climate and subtropical climate correspond to the needleleaf forests and the broadleaf forests areas respectively in the southern margin. However, climate types and vegetation are inconsistent in the southwestern TP and the ecotone boundary, primarily due to the overestimation of precipitation compared with observation data in the southwestern TP (Figures 2C,D). Since the Köppen-Trewartha classification has more than 30 sub-types, it is challenging to compare climate sub-types to vegetation groups, for the detail is restricted by the data resolution and the regional area of TP. In general, the distribution of the major climate types in TP corresponds well with vegetation mage types in the PD.

Given the relationship between climate types in TP in the PD, it could be inferred that there is an expansion of meadows, shrubs, and forests, and a shrinking of steppes and deserts during the MH, due to stronger monsoonal precipitation (Figures 7A,B, 10). Under the RCP4.5/RCP8.5 scenarios, a prominent shrinking area of the polar climate in the hinterland TP suggests the degradation risk of meadows, and the meadow will mainly be replaced by shrubs in the future (Figures 7B–D, 10). The distribution of shrubs and forests in the southeastern TP may shift northwestward. More specifically, anthropogenic warming may decrease cryophilic plants while thermophilic plants may increase.