CMIP6 Projections of the “Warming–Wetting” Trend in Northwest China and Related Extreme Events Based on Observational Constraints

2022-05-07 07:05QianWANGandPanmaoZHAI
Journal of Meteorological Research 2022年2期

Qian WANG and Panmao ZHAI

1 State Key Laboratory of Severe Weather, Chinese Academy of Meteorological Sciences, China Meteorological Administration, Beijing 100081

2 University of Chinese Academy of Sciences, Beijing 100049

ABSTRACT This study presents the improved future projections of the climate “warming–wetting” trend and climate extremes with different return periods in Northwest China at different global warming levels. The projections are based on the Coupled Model Intercomparison Project phase 6 (CMIP6) simulations constrained by the high-resolution observation dataset using the equidistant cumulative distribution functions (EDCDF) method. The results indicate that the climate will experience continuous warming and wetting as reflected by average temperature and total precipitation over Northwest China, especially under the scenario of the shared socioeconomic pathway 5–representative concentration pathway 8.5 (SSP5-8.5). Most parts of Northwest China will continue to warm in the future more than global average. Spatially, areas with prominent “warming–wetting” trends will be mainly distributed in western Northwest China. It is worth noting that extreme heat and precipitation events will also increase with the climate warming and wetting over Northwest China. Moreover, frequencies of rarer extreme events will increase more apparently than weaker extreme events and frequency increase of extreme heat events responds to global warming faster than that of extreme precipitation events. Limiting global warming within 2°C relative to 1850–1900 would slowdown the increase in extreme heat events and considerably suppress the increase in frequencies of extreme precipitation events,especially the rare (i.e., 50-yr) extreme events.

Key words: Northwest China, projection, precipitation, temperature, extreme events

1. Introduction

Northwest China is located in the interior of Eurasia,with lack of water resources, complex topography, and fragile ecosystems. It is one of the sensitive areas of global climate change (Wu et al., 2010; Zhao et al., 2010;Lu et al., 2021). Many studies have been carried out on climate change in Northwest China. As early as the end of the 20th century, Zhai et al. (1999) pointed out that the precipitation over Northwest China showed an increasing trend in 1951–1995. Shi et al. (2003) indicated that the climatic conditions have changed from “warm–dry”to “warm–wet” over Northwest China since the mid-1980s. This is the first time that the statement “warm–wet” has been proposed. Since then, the warming and wetting in Northwest China has received widespread attention. Recent studies suggest that with global climate warming, the climate has shown a significant “warming–wetting” trend over Northwest China in the past 60 years(Wu et al., 2019; Wang et al., 2020; Zhang et al., 2021).To a certain extent, the current climate warming and wetting is conducive to the development of ecosystems (Zhang et al., 2021). However, extreme precipitation events have notably increased with the climate warming and wetting,which may bring greater challenges to water resource utilization and disaster prevention and mitigation (Wang et al., 2020). Therefore, future projections of climate“warming–wetting” trend and changes in extreme events in Northwest China are of great significance for climate change response and disaster risk management.

In recent years, the issue of climate change projections in Northwest China has received increasing attention. As shown in Table 1, most studies indicate that the temperature and precipitation over Northwest China will continue to increase in the future, especially under representative concentration pathway 8.5 (RCP8.5) (Xu et al., 2003; Shi et al., 2007; Pan et al., 2020; Wang et al.,2021). Qualitatively, the projection results of temperature and precipitation in Northwest China based on the multimodel ensemble results of global climate models(GCMs) in Table 1 are both suggesting increasing trends.Studies based on GCMs simulations and using regional climate models (RegCMs) for dynamic downscaling can provide high-resolution local details. Due to the large uncertainty of the model for precipitation simulation, some studies based on the output of a single model may have a large uncertainty in the projection of precipitation. Specifically, Du et al. (2021) found that the future precipitation increase in Xinjiang under the RCP4.5 scenario is greater than that under the RCP8.5 scenario based on simulations of regional climate model driven by the Coupled Model Intercomparison Project phase 5(CMIP5) models. Some studies based on the output of a single model or considering only natural variability found that precipitation over Northwest China will see a generally decreasing trend in the 21st century (Yu et al.,2015; La et al., 2019). In addition, a few studies have further projected the changes in the extreme climate indices over Northwest China in the future based on CMIP5 simulations. Pan et al. (2020) pointed out that the extreme climate index for summer days will increase, while the consecutive dry days will decrease in the future based on regional climate model (RegCM4.6) driven by CMIP5 models. Wang et al. (2021) indicated that the minimum temperature, maximum temperature, and maximum precipitation in Xinjiang will increase in the future. Recently, a few studies have carried out related research based on the latest CMIP6 data. CMIP6 models exhibit a general improvement in the simulation of climate indices and climate extremes compared to CMIP5 simulations (Chen et al., 2020; Zhu et al., 2020). Under the scenario of the shared socioeconomic pathway 5–representative concentration pathway 8.5 (SSP5-8.5), the drought frequency will decrease, the drought duration will increase, and the drought severity will increase over Northwest China in the future (Li et al., 2020). Meanwhile, the simulation of CMIP6 models also shows that in the long term (2081–2100), annual mean temperature and annual precipitation in Northwest China will increase by 6°C and 27.3% respectively under SSP5-8.5 scenario relative to 1995–2014 (Qin et al., 2021).

Table 1. A summary of climate change projections in Northwest China from different references

With the deepening study on climate projections, the issue of constraints on the simulations of climate models is getting more and more attention (Watanabe et al.,2016). Within many constraint methods, the equidistant cumulative distribution functions (EDCDF) method is more popular. This method is developed from the traditional quantile-based matching method. The traditional method assumes that variables have the same cumulative distribution functions (CDFs) in the historical period and in the future projection period (Li et al., 2010), but IPCC(2007, 2013) pointed out that this assumption may not be true because of the nonstationary of the climate change process. Therefore, the traditional CDF method may artificially influence the climate change signal and destroy the distribution of future climate element sequences (Li et al., 2010; Yang et al., 2018). The EDCDF method applies a quantile-based mapping of the CDF between both the historic and projection periods and matches the climatic fields in the future projection period (Zhu et al.,2019; Qin et al., 2021). Yang et al. (2018) compared the constraint projections of different methods on temperature and precipitation simulations for China from CMIP5 models, and confirmed that the EDCDF method can more effectively reduce the model biases.

Although some studies on future climate change in Northwest China have been carried out, most studies still rely on the simulation results of the CMIP5 models, and there are few studies related to constraint projections and climate changes under different warming levels in the future. In addition, due to the different data and analysis methods used in different studies, there are also certain differences in the existing projections about precipitation(Table 1). In general, the current understanding of future climate change in Northwest China is still lacking, especially for the projections of extreme events. Although a few studies have involved future changes in extreme indices of temperature and precipitation, our study pays more attention to extreme events with return periods of 10, 20, and 50 yr. These rarer extreme events often have more serious impacts on natural ecosystems and socioeconomics (IPCC, 2012, 2018). Therefore, this study uses the latest CMIP6 model projections and constrains them based on the observation data to address the following key scientific questions. (1) How is the performance of CMIP6 models in simulating the climate states and their trends in Northwest China? Will the simulations be improved based on the observation data constraints? (2)What are the temporal and spatial characteristics of climate “warming–wetting” trends in Northwest China under different emission scenarios and at different global warming levels? (3) How will the frequencies of extreme events with different return periods in Northwest China change under different emission scenarios and at different global warming levels?

2. Data and methods

2.1 Data

The outputs of 16 CMIP6 models on daily mean and maximum temperature and daily precipitation are selected, mainly including historical simulations and outputs of three future scenario experiments (SSP1-2.6, SSP2-4.5,and SSP5-8.5; https://esgf-node.llnl.gov/projects/cmip6/).SSP1-2.6 represents the combined scenario of a sustainable socioeconomic development path (i.e., SSP1) with forcing at a low level of greenhouse gas emissions (i.e.,RCP2.6); SSP2-4.5 represents the combined scenario of a moderate socioeconomic development path (i.e., SSP2)with forcing at an intermediate level of greenhouse gas emissions (i.e., RCP4.5); and SSP5-8.5 represents the combined scenario of a high socioeconomic development path (i.e., SSP5) with forcing at a high level of greenhouse gas emissions (i.e., RCP8.5) (O’Neill et al.,2016; Riahi et al., 2017). The first member (r1i1p1f1) for each model is used in this study. The detailed information of 16 models is shown in Table 2. The observation data are from the high-resolution grid observation dataset CN05.1 (Wu and Gao, 2013), which is developed based on the observation data of more than 2400 meteorological stations in China. Observation dataset is mainly used to evaluate the performance of CMIP6 models and to constrain the model projection data of precipitation. In order to facilitate calculation and comparison, the observation and model data are interpolated uniformly by bilinear interpolation method. Considering that the observation dataset CN05.1 is of resolution of 0.25° × 0.25°and only contains data for territory of China, its interpolation to a coarser resolution will result in missing data near the national boundary. Therefore, in this study, we interpolated the model data to a resolution consistentwith CN05.1 dataset. We also compared the historical temperature and precipitation series before and after interpolation, and found that the results are very consistent,which shows that the interpolated results are reliable.

Table 2. Information on 16 global coupled climate models (the full model names can be found at https://esgf-node.llnl.gov/search/cmip6/)

2.2 Methods

The study area is Northwest China (30°–50°N,73°–112°E), mainly including Xinjiang, Qinghai, Gansu,Ningxia, and Shaanxi. The anomalies of the projected period in this study are relative to the preindustrial period(1850–1900). Since the observation data cover the period from 1961 to 2018, when comparing the observed and simulated temperature and precipitation anomalies in the historical period, the anomalies are calculated relative to 1961–1990. Linear trends are estimated by using Theil–Sen trend estimation (Sen, 1968). The significance test of the trend is based on the Mann–Kendall test(Kendall, 1975).

The EDCDF method (Li et al., 2010) is used in this paper to constrain model projection data of precipitation.For precipitation, the formula of EDCDF is:

whereXm-pis the simulated variable during future projection period,is the CDF ofis the inverse function of CDF obtained from the observation during the training period,is the inverse function of CDF obtained from the simulation during the training period, andXm-p,adjis the bias-corrected value.The training period selected in this study is 1961–2014.The gamma distribution is used for the CDF calculation of precipitation (Li et al., 2010; Yang et al., 2018).

The method for defining the global warming levels in this article refers to the definition in IPCC special report on the global warming of 1.5°C (SR15; IPCC, 2018),taking the global average temperature series of each model as a 31-yr running average, and the period of global warming 1°C is the first 31-yr period with global average temperature exceeding 1°C relative to the 1850–1900 average. The periods of 1.5 and 2°C of global warming relative to 1850–1900 for each model under three different emission scenarios are shown in Table 3. This study defines extreme events with different return periods based on the corresponding quantile values in the generalized extreme value (GEV) distribution, which has been wildly used to describe the distribution of climate indices such as extreme temperature and extreme precipitation (Kharin et al., 2007, 2013, 2018;Xu et al., 2018; Li et al., 2021). For example, the selection method of the thresholds of extreme heat and extreme precipitation with a return period of 50 yr (i.e., 50-yr event) is as follows. First, fit the GEV distributions for the annual maximum daily maximum temperature (TXx)and annual maximum 1-day precipitation (Rx1day) series from 1850 to 1900 on each grid of each model, and se-lect a CDF of 0.98 (1 - 1/return period) as the threshold of the grid. Then further calculate the change of the frequency or return period of the threshold under different global warming levels in the future. Estimating rarer extremes (e.g., 50-yr event) based on a short time window may produce greater uncertainty and potential deviations(Ben Alaya et al., 2020; Li et al., 2021). In order to minimize or avoid these deviations, changes of extreme events with different return periods in this study are analyzed based on a 31-yr time window.

Table 3. The periods reaching 1.5 and 2°C (31-yr running average)of global warming relative to 1850–1900 for each model projections under three different emission scenarios

3. Results

3.1 Historical climate changes and model performance

From the historical time series of temperature and precipitation in Northwest China, both temperature and precipitation have shown a significant increasing trend during 1961–2014. The increasing rates of temperature and precipitation are 0.27°C (10 yr)-1and 1.36% (10 yr)-1,respectively. Comparing the results of observations and simulations, it can be found that the CMIP6 models can reproduce the historical temperature changes in Northwest China very well, and the correlation coefficient between the observed and simulated temperature series is 0.85 which is statistically significant at the 95% confidence level (Fig. 1a). However, the simulation of precipitation is relatively poor. The simulated precipitation trend and the interannual variation have large differences from observations. The correlation coefficient between the observed and simulated precipitation series in Northwest China is only 0.27 (Fig. 1b).

Fig. 1. Historical time series (1961–2014) of (a) temperature and (b) precipitation anomalies in Northwest China (30°–50°N, 73°–112°E) from observations and model simulations.

Fig. 2. Taylor diagrams of the spatial distribution of historical (1961–2014) average (a) temperature and (b) precipitation in Northwest China(30°–50°N, 73°–112°E) simulated by CMIP6 models compared to observations.

Furthermore, the simulation performance of CMIP6 models on the spatial pattern of temperature and precipitation in Northwest China is evaluated. As shown in Fig. 2, the spatial correlation coefficients between observed temperature and simulations of 16 climate models are mainly distributed in the range of 0.85–0.98, and the correlation coefficient of precipitation is between 0.7 and 0.95. This shows that the simulation of the spatial distribution of temperature and precipitation in Northwest China is highly consistent with the observation, and the simulation of temperature is better than that of precipitation. This conclusion is consistent with the finding of Qin et al. (2021). The root-mean-square deviations(RMSDs) of temperature between observation and simulation of each model are small, while the RMSDs of precipitation mostly exceed 1.75, which suggests that there are large deviations between observed precipitation and simulations. In addition, for precipitation, the distribution of scattered points in the Taylor diagram is relatively discrete, which indicates that there are some differences in the ability of different models to simulate precipitation. This difference may be related to model internal variability (Zhou and Chen, 2015). Based on the ensemble mean of 16 CMIP6 model simulations, the spatial distribution pattern of the simulated precipitation is generally consistent with the observation (Figs. 3a1, b1).However, quantitatively, except for the Tianshan Mountains, the results of the multimodel ensemble mean overestimate historical average precipitation in Northwest China as a whole, and the models slightly underestimate the precipitation in the Tianshan Mountains (Fig. 3d1).Furthermore, models have poor simulating skills for precipitation trends over Northwest China. Specifically, the trends of precipitation in the Tianshan Mountains, and southern and eastern Qinghai are obviously underestimated, while the trends of precipitation in the Tarim Basin are overestimated (Figs. 3a2, b2, d2).

Considering that the CMIP6 models have relatively poor ability to simulate precipitation in Northwest China,and the differences in precipitation simulations between different models are large, this study uses the EDCDF method to correct the deviation of the precipitation simulation data to improve the reliability and credibility of future projection. As shown in Fig. 3, the spatial distribution of historical average precipitation described by biascorrected simulations is highly consistent with the observation (Figs. 3c1, e1). Meanwhile, the spatial distribution of the bias-corrected precipitation trend has also been significantly improved. The areas with prominent precipitation increase trends in the Tianshan Mountain area,southern Qinghai, and eastern Qinghai are all well described (Fig. 3c2). However, quantitatively, there is still a certain deviation between the simulation after the biascorrection of precipitation trend and the observation. For example, the precipitation trend in the Tianshan Mountains and Qinghai is still relatively small compared with the observation (Fig. 3e2). The precipitation data biascorrected by the EDCDF method are used in the following analysis.

3.2 Future climate changes

The results of future projections show that temperature and precipitation trends in Northwest China will continue to increase under three emission scenarios (Fig. 4).The increasing rates of future temperature and precipitation under SSP2-4.5 and SSP5-8.5 scenarios are more prominent than in 1961–2014 (Figs. 1, 4). Specifically,as shown in Table 4, the projections of different future periods indicate that under the SSP1-2.6 scenario, mean temperature in Northwest China will increase by 2.1°C(1.6–2.7°C) during the near-term (2021–2040) period relative to 1850–1900, and annual precipitation will increase by 11% (3.8%–18.8%). Temperature and precipitation will increase by 2.6°C (1.9–3.2°C) and 15.4%(6.0%–24.1%) respectively during the long-term (2081–2100) period. Under the SSP5-8.5 scenario, the future temperature and precipitation increase will be the most remarkable, which is also consistent with previous research conclusions (Li et al., 2020; Qin et al., 2021).Temperature will rise by 2.3°C (1.9–2.6°C) and precipitation will increase by 11.5% (3.5%–19.8%) during the near-term period, while temperature rise will reach 6.8°C(5.2–9.4°C) and precipitation will increase by 39.3%(23.9%–56.4%) during the long-term period under the SSP5-8.5 scenario (Table 4).

Figure 5 shows the spatial distribution of “warming–wetting” climate states in Northwest China at global warming levels of 1.5 and 2°C relative to 1850–1900.The results imply that most parts of the Northwest China will continue to warm more than global warming. The warming in the western part of Northwest China, especially in Xinjiang, is more pronounced than in the eastern part (Fig. 5). At 1.5°C of global warming, the warming in Northwest China is 1.7°C (1.4–2.3°C) under the SSP5-8.5 scenario (Table 5), and the warming in Xinjiang is the most prominent, with warming exceeding 1.75°C (Figs. 5a1, c1, e1). At 2°C of global warming, the temperature increase in Northwest China is 2.4°C(1.9–3.1°C) under the SSP5-8.5 scenario (Table 5), and especially the temperature exceeds 2.5°C in northern Xinjiang (Figs. 5b1, d1, f1). Regarding precipitation, excluding a few areas in the east of Northwest China, precipitation in most areas of Northwest China will increase in the future (Fig. 5). Under the SSP5-8.5 scenario, annual precipitation will increase by 7.6% (1.6%–13.4%) and 11.3% (4.0%–18.3%) at 1.5 and 2°C of global warming,respectively (Table 5). In terms of spatial distribution,the increase in precipitation will be more remarkable in the Junggar Basin, the Tarim Basin, and the Qaidam Basin. In particular, the increase in precipitation in the Tarim Basin will exceed 30% relative to 1850–1900,while the precipitation in southeast of Northwest China will decrease slightly (Figs. 5a2–f2). On the whole, the“warming–wetting” trend in Northwest China will continue in the future and the magnitude of trend will depend on levels of warming or scenarios. Regionally, it will mainly be in the western Northwest China.

Table 5. Temperature and precipitation anomalies (relative to 1850–1900) in Northwest China (30°–50°N, 73°–112°E) at 1.5 and 2°C of global warming under three different emission scenarios

Fig. 3. Spatial distributions of (a1, a2) observation, (b1, b2) simulation before and (c1, c2) after the bias-correction, and (d1, d2, e1, e2) the differences between observation and simulation of the average annual precipitation (mm) and trend [mm (10 yr)-1] over Northwest China (30°–50°N,73°–112°E) during 1961–2014. The simulation results are computed based on the multimodel ensemble mean (MME).

Fig. 4. Time series of (a) annual mean temperature and (b) annual precipitation anomalies in Northwest China (30°–50°N, 73°–112°E) during 2015–2100. Thick lines stand for multimodel mean and shading is the 5%–95% range across CMIP6 models.

Table 4. Projections of temperature and precipitation anomalies in Northwest China (30°–50°N, 73°–112°E) during different projection periods under three different emission scenarios

Fig. 5. Spatial distributions of (a1–f1) temperature (°C) and (a2–f2) precipitation (%) anomalies (relative to 1850–1900) in Northwest China at 1.5 and 2°C of global warming. The dotted area indicates a significant change in temperature and precipitation at the 95% confidence level.

3.3 Future climate extremes change

As pointed out above, the climate “warming–wetting”trend in Northwest China will continue in the future.Then, how will the frequencies of extreme events change in the future? In the following, we will analyze the changes of TXx and Rx1day in the future, and further analyze the changes of the frequency of 50-, 20-, and 10-yr extreme heat and extreme precipitation events at different global warming levels.

Judging from the probability density function (PDF)distributions of TXx and Rx1day (Fig. 6), the probability of extreme heat and extreme precipitation events will increase remarkably in the future, especially under the SSP5-8.5 scenario. For the extreme heat, the PDF distributions under the three different emission scenarios are basically the same in the near-term period, the differences gradually appear from the mid-term period, and the differences are very obvious in the long-term period(Figs. 6a–c). Under the SSP5-8.5 scenario, the probability of 50-yr extreme heat events at the end of the 21st century is close to 100% (Fig. 6c). For extreme precipitation, especially in the long-term period, the PDF distributions under the three emission scenarios are apparently different (Figs. 6d–f). This further shows that the temperature response to different future SSP scenarios is faster than precipitation. The main reason is that temperature responds more directly to different scenarios, and the impact of precipitation is more complex. Under the background of global warming, the midlatitude circulation system is possibly adjusted, which in turn affects precipitation changes (IPCC, 2021).

Fig. 6. Probability density function distributions of TXx and Rx1day in Northwest China (30°–50°N, 73°–112°E) during different projection periods under three emission scenarios obtained by GEV fitting. The dotted lines indicate the thresholds corresponding to extreme events with different return periods.

With global warming, the current and future frequencies of extreme heat and extreme precipitation events in Northwest China both exceed those of the 1850–1900 period. It shows that extreme heat and extreme precipitation events with different return periods are becoming more frequent as the global warming, especially under the SSP5-8.5 scenario (Fig. 6). Further, Fig. 7 shows the frequency changes of 50-, 20-, and 10-yr extreme heat and precipitation events at different global warming levels under the SSP5-8.5 scenario. The changes in the frequencies of extreme heat events respond to global warming faster than extreme precipitation events. When the global warming is 1.5°C, the frequency ratios of 10-,20-, and 50-yr extreme heat events will become 3.4, 5.1,and 9.3, respectively, and the frequency ratios of extreme precipitation events are 1.1, 1.5, and 2.9, respectively. When the global warming reaches 2.0°C, the frequency ratios of 10-, 20-, and 50-yr extreme heat events will become 4.8, 7.7, and 14.9, respectively, and the frequency ratios of extreme precipitation events are 1.2, 1.6,and 3.1, respectively. It can be seen that the frequencies of extreme heat events at these above global warming levels are markedly greater than those of extreme precipitation events, and the frequencies of 50-yr extreme events are clearly greater than those of 10-yr extreme events. This indicates that projected changes in frequencies will be higher for rarer extreme events. For 50-yr extreme events, the frequencies of extreme heat and extreme precipitation events will be respectively 37.5 and 4.6 times that of 1850–1900 at the 4°C of global warming, while these frequencies will be 14.9 and 3.1 times at the 2°C of global warming (Fig. 7). Obviously, limiting global warming within 2°C would substantially suppress the increase in frequencies of extreme heat and extreme precipitation events, especially the 50-yr extreme events.

Fig. 7. Projected frequency ratio changes of 50-, 20-, and 10-yr (a) extreme heat and (b) extreme precipitation events in Northwest China(30°–50°N, 73°–112°E) under the SSP5-8.5 scenario relative to 1850–1900. The boxes show the 25th and 75th percentiles, the lines in the boxes mark the median, the asterisks mark the mean, and the line below (above) the 25th (75th) percentiles shows the minimum (maximum) value.

4. Conclusions

This study evaluates the simulation performance of 16 CMIP6 models for temperature and precipitation in Northwest China, and uses the EDCDF method to constrain the projection data of precipitation based on the high-resolution observation data. Furthermore, the changes in “warming–wetting” trend in Northwest China under different global warming levels and emission scenarios are analyzed. Particular attention is paid to the projections of the frequencies of extreme heat and extreme precipitation events with different return periods in Northwest China. The main conclusions are summarized as follows:

The simulation performances of CMIP6 models for temperature are obviously better than those for precipitation. The simulation of spatial distribution pattern of precipitation is generally consistent with the observation,while the quantitative simulations of the total precipitation and precipitation trend have a certain deviation from the observation. In addition, the simulation results of different models for precipitation are quite different. After the precipitation simulation is corrected by the EDCDF method based on the observation, the differences of the precipitation simulated by each model are reduced, and the bias-corrected spatial distributions of multimodel ensemble precipitation and precipitation trends are more consistent with the observations.

In the context of global warming, the climate “warming–wetting” trend in Northwest China is expected to continue in the future, especially under the SSP5-8.5 scenario. The temperature and precipitation in Northwest China will increase by 6.8°C and 39.3% respectively in the late 21st century relative to 1850–1900. Compared with the global average warming, the warming in Northwest China is even more dramatic, especially in Xinjiang. Specifically, under the SSP5-8.5 scenario, temperature and precipitation in Northwest China will increase by 1.7°C and 7.6% respectively at the 1.5°C of global warming. Temperature and precipitation in Northwest China will increase by 2.4°C and 11.3% respectively at the 2°C of global warming. In terms of spatial distribution, areas with a more prominent climate “warming–wetting” in the future will be mainly distributed in the west of Northwest China.

It is worth noting that in the background of continuous “warming–wetting,” there will be an apparent increase in extreme heat and extreme precipitation events over Northwest China. The frequency changes of extreme heat events respond to global warming faster than that of extreme precipitation events. For extreme events with different return periods, the increase in frequencies of rarer extreme events will be more pronounced than weaker extreme events. The return periods of 50-yr extreme heat and extreme precipitation events will become about 3.3 and 16.7 yr during the 2°C period of global warming under the SSP5-8.5 scenario. When the global warming reaches 4.0°C, the return periods of 50-yr extreme heat and extreme precipitation events will be 1.3 and 11 yr, respectively. Limiting global warming within 2°C relative to preindustrial levels can slow down the increase in extreme heat events and substantially suppress the increase in frequencies of extreme precipitation events, especially the 50-yr extreme events.

Acknowledgments. The authors acknowledge the World Climate Research Program’s Working Group on Coupled Modeling and thank the climate modeling groups for producing and sharing their model outputs.