Recent frontiers of climate changes in East Asia at global warming of 1.5°C and 2°C | npj Climate and Atmospheric Science – Nature.com
East Asian monsoon circulations
The East Asian monsoon consists of the East Asian summer monsoon (EASM) and East Asian winter monsoon (EAWM) (Fig. 2 and Table 1). It is widely accepted that changes in both EASM and EAWM circulations can be quantified using monsoon indices29. However, there are still many controversies about future changes due to different choices of monsoon indices and uncertainties among the models. Here, we focus on summarizing the changes in the East Asian monsoon and its components at GW1.5 and GW2, including related uncertainties. In addition, the long-term changes of the East Asian monsoon under global warming are also included, which is useful to explain the changes at GW1.5 and GW2.
East Asian summer monsoon (EASM)
The EASM contributes nearly half of the annual precipitation in East Asia, and any changes will affect the intensity and position of the main monsoon rain belts, making prediction difficult30,31,32. There are many types of EASM indices, including land–sea thermal difference indices defined by sea-level pressure, circulation indices defined by geopotential height, and indices of water vapor transport flux33,34,35. The intensity of EASM increases (medium confidence) in a warmer future in Coupled Modeling Intercomparison Project phase 5/6 (CMIP5/6) models36,37 (Fig. 2a and Table 1), with dry-get-wetter climate38, especially in the long-term future under the higher emission scenarios, the enhancement of EASM also reflects in increased EASM precipitation (high confidence) and increased monsoon duration (medium confidence)1. Recent studies using CMIP6 models found that the changing rate of precipitation is distinct under the high emission scenarios in the long-term future39. It is reported that the summer monsoon precipitation increased by 2.1%/°C in CMIP6 multi-model ensemble mean under the shared socioeconomic pathway (SSP) 2–4.5 scenario37. Meanwhile, the duration of the summer monsoon precipitation over East Asian might increase by more than 1.6 pentads under the SSP5–8.5 scenario, which is related to the earlier onset and later retreat of the EASM39,40.
Although the projected strength of the EASM will intensify (medium confidence) at GW1.5 and GW241, results obtained using different EASM indices and datasets are varied. Ensemble projections using CMIP5 models show future EASM intensity increasing at GW1.5, and the interannual variabilities of the EASM and its associated precipitation are also expected to be enhanced, with large inter-model spreads remained28. Similar conclusions are obtained from the CESM-low warming experiment, that all ensemble means to show an increase in EASM intensity and associated precipitation over most parts of the East Asian region in 1.5 °C “never exceed”, 1.5 °C “overshoot” and 2.0 °C “never exceed” experiments and any significant difference in the future changes in EASM intensity and precipitation among the three scenarios42. Another study using large-ensemble simulations also showed the enhanced EASM circulation and precipitation with global warming, however, the increase was not significant at GW1.5. Further study indicates that the positive contribution of the dynamic component related to strengthening EASM is almost twice as large as that of the thermodynamic component when global warming increases by more than 2 °C, while both contributions are quite similar and small at GW1.543. Therefore, the changes of EASM at GW1.5 and GW2 still remain large projection uncertainties, and the confidence level is still low at GW1.5. Compared to the projections of surface temperature changes related to the EASM, there is lower confidence in projections of precipitation changes at GW1.5 and GW244. Nonetheless, the benefits of limiting global warming to 1.5 °C instead of 2 °C are clear45, and it will robustly reduce areal and population exposure to dangerous extreme precipitation events in the populous land monsoon regions46.
The EASM is closely associated with water vapor transport (WVT). There are four typical WVT corridors in the summer monsoonal circulation: the southwestern, South China Sea, and southeastern corridors at low latitudes, and the Eurasian westerly corridor at high latitudes47. The results from CMIP5 models showed that summer WVT is enhanced in East Asia at GW1.5 and GW2 under both RCP4.5 and RCP8.5 (medium confidence). For southern East Asia, enhanced southerly or southwesterly flow due to an enhanced South China Sea summer monsoon trough and western North Pacific subtropical high favors the enhanced WVT, while for northern East Asia, changes in lower-level circulation have less contribution to WVT, and the increased water vapor content mainly due to the increased tropospheric air temperature48. A dynamic component plays an important role in WVT changes at GW2, while the thermodynamic component caused by increased water vapor content is more important at GW1.549. The Tibetan Plateau also plays an essential role in enhancing WVT related to EASM circulations under global warming through enhancing precipitation-induced anomalous latent heating over the high plateau50.
The Western North Pacific Subtropical High (WNPSH) is another dominant component affecting the EASM and is a key high-pressure system controlling the summer monsoon rainfall and typhoon activities over the western Pacific47,51. Thus, the intensity, shape, and location of WNPSH have great influences on EASM, but whether the WNPSH will be strengthened or weakened remains inconclusive due to limitations in models, varied index definitions and constraining methods51,52,53,54,55. For example, the WNPSH may be strengthened and extended westward in a warming future (low confidence) (Fig. 2a and Table 1), with the most pronounced changes expected in RCP8.5; while the ridge line of WNPSH shows no obvious long-term trends56. However, based on the projected changes in 925 hPa wind, 925hPa relative vorticity, 925 hPa divergence, and 700 hPa vertical velocity, the CMIP5 model simulation under RCP8.5 projects reduced intensity for the subtropical anticyclones over the North Pacific57. Meanwhile, the multi-model ensemble mean from 33 CMIP5 models shows the projected changes in the WNPSH intensity (defined by zonal wind at 850 hPa) are approximately zero in the long-term future, and about half of the models indicate an enhanced WNPSH and about half of the models project a weakened WNPSH under both RCP4.5 and RCP8.5 scenarios58. Furthermore, using a sea level pressure index as a definition of the WNPSH constrained by observed sea surface temperatures, CMIP5 models project a strengthened and westward-shifted WNPSH due to suppressed warming in the western Pacific and enhanced land–sea thermal contrasts59. However, in contrast, based on eddy geopotential height at 500 hPa in CMIP5 models, the WNPSH will be weakened and retreat eastward in the middle troposphere in response to global warming, accompanied by an eastward expansion of the East Asian rain belt along the northwestern flank of the WNPSH51. Therefore, even in the long-term future, the projection of WNPSH is still uncertain, and the reasons for the intermodal spreads remain unknown.
For 1.5 and 2 °C targets, future projections of the WNPSH remain hugely uncertain at different geopotential levels. Generally, the study shows that the WNPSH approximately linearly weakens (eastward retreat) at 500 hPa and intensifies (westward extension) at 850 hPa with a rise in the warming target, while such changes are insignificant at GW1.5 to GW2, and becomes apparent at higher warming targets (low confidence). Meanwhile, the projected WNPSH might experience an interdecadal variation in the 21st century which is influenced by natural variability, thus the robustness of the WNPSH projection is low under the low warming targets like GW1.5 or GW252.
The Summer East Asian subtropical westerly jet (EASWJ), a strong westerly jet in the upper troposphere with maximum intensity at around 200 hPa, is also an important component of the EASM system. The EASWJ is located to the north of the EASM rain belt, and its intensity and location control EASM rainfall53,60,61,62. In summer, a southward EASWJ displacement often corresponds to strong convective activity and accompanies increased precipitation over south-central China, while a northward shift is associated with heavier rainfall in North China63. Many studies have projected changes in the EASWJ in response to global warming and have shown the increase of its interannual variability and a southward displacement in the EASWJ during the 21st century (medium confidence). This would also strengthen the relationship between EASM rainfall and the EASWJ60,64. The CMIP3, CMIP5, and CMIP6 model simulations all project a southward shift and intensification of the EASWJ53,60,61, which is associated with an enhanced poleward temperature gradient in the mid-high troposphere in East Asia60,64.
Noticeably, however, the increase of the westerly on the south side and decrease on the north side will intensify with increasing global warming targets, suggesting a lower signal-to-noise ratio for EASWJ changes at GW1.5 compared to GW2. The results show that the EASWJ weakens slightly with weak air temperature meridional gradient change at GW1.5 (low confidence), while the westerly wind is strengthened/weakened on the south/north of the EASWJ axis at GW2 (medium confidence), leading to a southward movement64.
East Asian winter monsoon (EAWM)
The EAWM is characterized by a prevailing northerly flow between the Siberian high and the Aleutian low and is associated with low winter temperatures, outbreaks of snowfall in northern China, and freezing rain in southeastern China65. The East Asian trough is a major feature at 500 hPa, usually around the longitude of Japan. Four main indices are often used to describe the circulation related to the EAWM: low-level meridional wind indices, east–west pressure gradient indices, upper-level zonal wind shear indices, and east-Asian trough indices66. Using low-level meridional wind indices, a strengthened northerly wind in a future warmer climate is projected and implies a stronger EAWM, which is associated with changes in Aleutian low and near-surface temperature in East Asia67. However, based on the rest three indices closely related to the near-surface temperature of East Asia, the EAWM is projected to become weaker in the future (medium confidence)68 (Fig. 2b and Table 1). In addition, interannual variability of these indices will be maintained with an intensity similar to that of the present68. Thus, the projections of future changes in EAWM are highly dependent on the models and indices employed.
It is also shown large uncertainties could be found in the projection of EAWM system members at GW1.5 and GW2, which are described in the following sections. Multi-model results from the CMIP5 models indicated that compared to the EAWM system members at GW1.5, the EAWM system members show more robust changes at GW2 under the RCP4.5 and RCP8.5 scenarios65.
The Siberian high (SH) and Aleutian low (AL) are located on the Asian continent and North Pacific at the lower troposphere, respectively, and are important components of the EAWM. A stronger (weaker) SH and AL lead to increased (reduced) zonal sea level pressure gradient and stronger (weaker) EAWM. At GW1.5 and GW2, the projected SH is weakened but the AL is intensified based on CMIP5 models (medium confidence)65, consistent with earlier results based on CMIP3 models69. The projected significant intensification and a northward shift of the AL are associated with strong warming over the high-latitude North Pacific due to melting sea ice in the Bering Sea and Okhotsk Sea65. At the end of the 21st century, the SH is expected to decrease with a much smaller magnitude than the intensification of the AL. Overall the projected poleward shifting of AL increases the horizontal pressure gradient between the SH and the AL regions and strengthens the EAWM67.
The East Asian trough (EAT) is an important part of the EAWM system in the middle troposphere (500 hPa), and a deeper (shallower) EAT is conducive to a stronger (weaker) EAWM65. The tilt of the EAT line is closely correlated with the EAWM pathway. When the EAT line tilts more (less) eastward, the EAWM tends to take the eastern (southern) pathway, which will bring more (less) cold air to the central North Pacific in comparison to East Asia70. According to the results of 28 CMIP5 models, the EAT is projected to weaken slightly and tilt more eastward with latitude at GW1.5 and GW2 under RCP4.5 and RCP8.5 (medium confidence)53, consistent with the previous studies67. A stronger eastward tilt with the latitude of the projected EAT suggests that the future EAWM will bring more cold surges out into the North Pacific and a less disturbed winter climate in East Asia67, although the current climate models still have large uncertainties in the simulation of the EAT behavior and its position71.
The winter East Asian jet (WEAJ) in the upper troposphere is also an important component of the EAWM. The core region and zonal wind speed as well as the jet axis at 200/300 hPa can be used to characterize the WEAJ. A strong jet is beneficial to a deeper East Asian trough and Aleutian low as well as EAWM enhancement65,72. At GW1.5 and GW2, the core region of the WEAJ is expected to shift westward, zonal winds at 200 and 300 hPa increase slightly and the jet axis shifts would shift northward under RCP4.5 and RCP8.567, suggesting a northward shift of the WEAJ (medium confidence)65,68.
The El Niño-Southern Oscillation (ENSO) is a prominent feature of interannual climate variability, and associated weather and climate events can cause pronounced socioeconomic losses in East Asia. The future projection of ENSO and its climatic impacts on East Asia is of great importance7. Based on CMIP3 and CMIP5 models in a warming world, ENSO intensity and the frequency of extreme El Niño events are expected to increase73,74, despite uncertainties in climate model prjections75. Moreover, the influence of El Niño on the interannual variability of climate in East Asia is ENSO-phase dependent in a warmer future76 (Fig. 3). It is projected that the El Niño-related sea surface temperature anomaly pattern becomes narrower and stronger under global warming76, leading to a strengthened positive (negative) precipitation anomaly in the equatorial central Pacific (northwestern Pacific). Furthermore, the strengthened negative precipitation anomaly in the northwestern Pacific further enhances an anomalous anticyclone over the tropical western North Pacific, which increases southerly flow to East Asia, leading to enhanced precipitation there76.
Meanwhile, the future behavior of ENSO under anthropogenic forcing remains uncertain. IPCC AR6 emphasized that there is no consensus from models for a systematic change in the amplitude of ENSO variability over the 21st century in any of the SSP scenarios assessed (medium confidence)1. A recent study identifies the sources of uncertainty in the ENSO amplitude projections in CMIP5 and CMIP6 models, which found that internal variability is the main contributor to the uncertainty in the near term future, and model uncertainty dominates thereafter, while scenario uncertainty is relatively small throughout the 21st century77. Thus, it is hard to determine the ENSO changes at GW1.5 and GW2 due to large internal ENSO variability on decadal time scales in the near-term future.
Climate changes and climate extremes
Mean temperature and extreme temperature
The surface means the temperature in East Asia will continue to rise under different scenarios in the future based on CMIP6 models (high confidence) (Figs. 1a and 4a–c). At both GW1.5 and GW2, warming in East Asia will be more significant than the global mean22,25,78. The results of the CESM low-warming experiment show the East Asian temperature would increase by approximately 1.4 and 2 °C at GW1.5 and GW2, and larger warming magnitudes would occur in the southern, northwestern, and northeastern regions of China, parts of Mongolia, the Korean Peninsula, and Japan than in other regions78. Similar results can be obtained from the CMIP5 and CMIP6 models. Multi-model ensemble mean from the CMIP5 models indicated the mean surface temperature in China will be increased by 1.82 °C/1.83 °C at GW1.5 and 2.60 °C/2.52 °C at GW2 under RCP4.5/8.5, respectively. Regionally, there are similar spatial patterns at both GW1.5 and GW2 under different RCP scenarios, and the projected changes in northern China, northwestern China, southern China, and the Tibetan Plateau being particularly sensitive to the additional GW0.579. The spatial pattern of seasonal warming in China at GW1.5 is similar to the annual mean80. Recent studies show the annual surface mean temperature in China is increased by 1.49 and 2.21 °C (relative to 1986–2005) at GW1.5 and GW2 in CMIP6 models under the SSP5-8.5 scenario, while the counterpart in CMIP5 models under RCP8.5 is 1.20 and 1.93 °C, respectively, and the changes of annual surface mean temperature in China are generally larger in the new generation of models and scenarios3. However, there are projection uncertainties that remain at GW1.5 and GW2, which can also be reflected in signal-to-noise. It is shown that the Southern China/Tibetan Plateau has the smallest/largest signal-to-noise ratio at GW1.5 and GW2, and both the model and internal variability uncertainty are the main sources of uncertainties over China at GW1.5 under RCP4.5 and RCP8.5 scenario, while both the model and scenario uncertainty contribute more at GW279.
In terms of extreme temperature events, global warming will significantly increase extreme high temperatures, including eastern China, the Korean Peninsula, and Japan will see more intense, more frequent, and longer-lasting extreme high-temperature events78, but decrease the extremely low temperature at both GW1.5 and GW2 (high confidence) (Fig. 4d–i), and Table 2 summarizes projected extreme temperature at GW1.5/2 in China. At GW1.5/GW2 relative to 1986–2005, the temperatures of the hottest day and coldest night are projected to increase about 1.0 °C/1.6 °C and 1.1 °C/1.8 °C, whereas warm days and warm spell duration will increase about 7.5%/13.8% and by 15/30 days, respectively (Fig. 4j–l)25. The spatial distribution of changes in extreme temperature is generally consistent between CMIP5 and CMIP6 models but with significantly stronger increases in CMIP6 models in northeastern and northwestern China for the hottest days, and southern China for the coldest nights3. An additional GW0.5 is projected to have substantial impacts on the extreme temperatures in China. As shown in the results of the CESM low-warming experiment, Compared to the higher scenarios GW2, GW1.5 will help avoid 35–46% of the increases in extreme high temperature in terms of intensity, frequency, and duration in East Asia with maximal avoidance values (37–49%) occurring in Mongolia78. CMIP6 models indicate about 36–87% of occurrences of extremely high temperature will be avoided in China81, and similar results can also be found in CMIP5 models25. Therefore, it is beneficial for East Asia to limit the warming at GW1.5.
In addition to intensity, global warming will also affect the probability of extreme temperature events. It is found that the frequency, duration, and intensity of extreme warm events in China under RCP4.5 and RCP8.5 at GW2 will increase82. Extreme heatwave events with a recurrence interval of 5 years under the current climate will recur every 2 years at GW1.5, while the severity of annual heatwaves at GW2 will be higher than that recorded in the summer of 201383. Moreover, the risks of the current 1-in-100-year hottest day and hottest night occurring are projected to increase by 14.4/23.3 and 31.4/50.6 times in China at GW1.5/2, respectively. The greatest increase in the frequency of extremely high temperatures is over the Tibetan Plateau, northwestern China, and south of the Yangtze River84. In terms of stabilized global warming scenario, an extreme hot event with a return period of 100 years in the present climate becomes an event occurring every 4.79 (GW1.5) and 1.56 years (GW2), an extreme cold event with a return period of 10 years becomes event occurring every 67 years at GW1.5 and is unlikely to occur under GW285. Additionally, since East Asia is one of the most densely populated regions in the world, enhanced extreme temperatures associated with global warming would expose more people to danger. It is believed that if global warming is restricted well below 2 °C, the avoided impacts in population exposure are prominent for most regions in East Asia86,87.
Mean precipitation and extreme precipitation
Based on CMIP6 models, precipitation in East Asia will increase in the future (high confidence) (Fig. 5a–c). At GW1.5/GW2 relative to 1975–2005, the global mean precipitation will increase by 1.7–1.9%/2.5–2.7%, while the regional mean precipitation in China will increase by 2.1–2.9%/3.2–5.3%, suggesting that regional precipitation changes are larger than the global mean88. There are regional differences in the projected change of annual mean precipitation, and the percentage change tends to increase from southeast to northwest at GW1.580. The percentage change of precipitation in each season is similar to the annual average, which increases in the north of China but decreases in most parts of southern China28. Changes in summer precipitation show intra-seasonal contrasts, and are projected to decrease over 30-40°N in June and over the Meiyu belt in July, while on the other hand August rainfall is projected to increase in the high latitudes of East Asia89. Winter precipitation is also projected to increase in East Asia at GW1.5, especially in the northern part of the continent where positive anomalies exceed 12% and such a pattern remains evident at GW2 but with a stronger magnitude65. On average, winter precipitation is projected to increase by 4.9% at GW1.5, and the additional GW0.5 for GW2 induces a further 4% increase in the east Asian continent65. However, the regional average signal-to-noise ratios of annual and seasonal precipitation changes at GW1.5 are only 0.1 and 0.01–0.280, indicating that the model projection precipitation with great uncertainty, which may be related to the poor ability of global climate models to simulate the East Asian monsoon precipitation90.
The intensity and frequency of extreme precipitation events are projected to increase over most of East Asia at GW1.5 and GW2 (high confidence) (Fig. 5). Table 2 summarizes the changes in projected precipitation extremes at GW1.5/2. At GW1.5/GW2 relative to 1986–2005, the regional average extreme precipitation intensity in China increases by 6%/11%, respectively, and the extra GW0.5 increases extreme precipitation intensity by 4%91. The recurrence interval of a typical 100-year event reduces to 63/42 years at GW1.5/2, respectively91. The maximum 5-day precipitation total increases by 3.98% (7.63%), precipitation on very wet days increases by 19.41% (34.42%), and the annual total wet-day precipitation increases by 3.89% (8.23%) at GW1.5 (GW2) in China suggesting that the extra GW0.5 can double the increase in precipitation extremes92. However, there are regional differences. The maximum 5-day precipitation total is expected to increase by approximately 8.9% in northwestern China, 3.8% in northern China, and 2.3% in southern China at the extra GW0.593. At GW1.5/2 related to 1971–2000, precipitation on extremely wet days is projected to increase by 69%/87%, and the heavy precipitation days will increase by 25%/40%94. As for the difference between CMIP6 and CMIP5 models, the future response of extreme precipitation to warming in CMIP6 models is larger than in CMIP5 models. With respect to 1985–2005, precipitation on very wet days will increase by 5.3%/8.6%/16.3% at GW1.5/2/3 in CMIP6 models under SSP5–8.5, while the counterpart figures for CMIP5 models under RCP8.5 are only 4.4%/7%/12.8%3. Meanwhile, on the south bank of the Yangtze River and for most regions around 40°N, CMIP6 models show even higher increases for precipitation recorded on wet or very wet days3. Although the intensity and frequency of extreme precipitation events are both projected to increase in most areas in East Asia at GW1.5 and GW2, avoiding the additional GW0.5 helps reduce the risks of extreme precipitation frequency and intensity by 26–31%, and the greatest reductions are found in Japan with 38–54%49. Meanwhile, the population exposure would increase more rapidly at GW2 compared to GW1.5, especially under the RCP8.5 scenarios95. Realizing the GW1.5 will robustly reduce the areal and population exposures to dangerous extreme precipitation events for populous East Asia (reduce 35%/29% for RX5day events that exceed the baseline 10-year/20-year return values)46.
Drought is an important natural hazard and its risk and severity are expected to increase in a warmer climate, but the projected drought changes depending on region, season, and drought metrics96,97. Based on both Palmer Drought Severity Index (PDSI) and the Standardized Precipitation Evapotranspiration Index (SPEI) which are widely used drought indices, drought risks in China are projected to increase in many strong warming scenarios, probably attributable to an enhancement in evapotranspiration98,99. Projected changes in PDSI and SPEI show that the area of drought in northwestern China will increase100 and that most areas of southern China will be drier in a warming future98. The future drought frequency in CMIP6 models is projected to be less serious than that in CMIP5 models101. Drought frequency is expected to increase by 28 ± 4% at GW1.5 based on CMIP5 models, but to increase by only 12 ± 4% based on CMIP6 models101. This is due to higher precipitation and potential evapotranspiration in CMIP5 models than in CMIP6 models. Furthermore, the projection of drought is also dependent on future emission pathways. Based on PDSI, less/more drought is projected under a low/high emission pathway for the same warming level in East Asia at GW1.5102, which is likely related to contrasting precipitation responses to regional aerosol loading99. Compared to GW1.5, the 0.5 °C additional warming at GW2 will account for approximately 9% of the increase in the drought occurrence in China and ~8% of extreme droughts103. Thus, the additional warming would lead to a significantly higher drought impacts104,105,106. It is shown that the population exposure to extreme droughts is projected to increase by about 17% due to the 0.5 °C additional warming103, and the drought loss caused at GW2 would increase by tens of billions of dollars compared with that at GW1.598.
Local and regional changes and hotspots
Importance of local and regional climate change
Although consequences of climate change vary with regions, climate response at the local and regional levels plays a critical role in regional climate variability1. This is because, in addition to global forcings (e.g., greenhouse gases), regional climate change is also driven by regional forcings, including aerosol emissions, land use, and land cover changes, as well as changing synoptic climatology which drives regional climate variability1. For example, cooling effects of heavy aerosol load dominate historical temperature change in central-eastern China, in some cases masking or even reversing warming trends caused by greenhouse gases107. Moreover, regional climate change is also modulated by regional physical processes such as snow/ice-albedo feedback and soil moisture–temperature–precipitation feedback. For instance, the Tibetan Plateau shows stronger temperature responses to GW1.5 and GW2 than surrounding regions, due in part to local snow/ice-albedo feedback108,109,110. More importantly, predicting regional climate change is highly relevant to policy-making related to regional climate mitigation and adaptation. Thus, more attention is necessary to downscale local and regional climate changes from predictions made at a broader scale. It is often local changes that have disproportionate impacts on humans and ecosystems.
Approaches on local and regional climate projection
Both general circulation models (GCMs) and regional climate models (RCMs) have shortcomings in attempting to simulate the fine spatial structure of climate change, for example, both GCMs and RCMs generally fail to reproduce mean and extreme precipitation at the local level compared with observations111. Dynamical and statistical downscaling techniques are important approaches to improve model simulations, and statistical downscaling can achieve more realistic local and regional climate change projections by transforming outputs and simulations of both GCMs and RCMs to the local scale, whereas dynamic downscaling is often carried out using RCMs which have a finer resolution and consider more local effects than the GCMs1.
A comparison of statistical and dynamical downscaling of extreme temperature and precipitation events in China at GW1.5 and GW2 found the results are similar and method-independent112. Recent studies show that statistical downscaling approaches such as a non-homogeneous hidden Markov model113 and self-organizing maps114 can add values to GCMs in simulating climate extremes and show good success in simulating local probability distribution function and spatial distribution of climate extremes in eastern China.
Identification of hotspots and sensitive regions
Identifying both climate change “hotspots” and sensitive regions of climate change is critical for effective mitigation and adaptation activities. Climate change hotspots can be identified by either (a) the magnitude of physical climate change, (b) the vulnerability to climate change impacts, or (c) both (a) and (b). There are many methods designed to identify the most responsive regions to climate change. The most widely used method is focused on constructing metrics quantifying aggregate (positive and negative) changes in several variables. Different variables may have different scales and units. For example, a Regional Climate Change Index (RCCI) was developed based on regional mean precipitation change, mean surface air temperature change, and changes in the interannual variability of precipitation and temperature. These changes were combined to quantify the aggregate response in multi-dimensional climate space between the present period and the future simulation115. Other aggregations of multi-dimensional climate change include (a) standard Euclidean distance which measures the distance traveled in multivariable climate space116, (b) summing the number of seasons exceeding the different temperature and precipitation thresholds, and (c) measures of extreme seasonal temperature and precipitation. Evaluating the changes in future risk varied strongly by geographical location117,118.
There is increasing interest in identifying hotspots and sensitive regions in East Asia, and a desire to understand how these may vary for different levels of global warming. The Tibetan Plateau, north-western China (arid and semi-arid region), the region south of the Yangtze River (including Jianghuai), and urban areas have been variously identified as regions with the strongest responses to global warming119. They are also exposed to greater risk than other regions suffering from climate extremes84. The subsequent section discusses the enhanced changes experienced in such regions.
Climate change and associated physical mechanisms in hotspot regions
The Tibetan Plateau contains the largest cryospheric region outside the polar regions and is undergoing cryospheric change and rapid warming largely in phase with global trends but at a higher magnitude108,120. Both GCMs and RCMs consistently project an amplified increase in mean temperature over the Tibetan Plateau at GW1.5 and GW2, which is even more obvious for temperature extremes, such as the hottest day representing the annual maximum value of daily maximum temperature109. Robust elevation-dependent warming over the Tibetan Plateau at both GW1.5 and GW2 is projected, and greater amplification appears in the higher GW2 scenario. Understanding such amplification is essential for the management of the sustainability of water resources over the Tibetan Plateau110. The amplified warming and elevation-dependent warming over the Tibetan Plateau can be attributed to multiple physical processes121,122 (Fig. 6), mainly snow/ice-albedo feedback and cloud–water vapor–radiation interactions121, supplemented by other local forcing and feedback processes (e.g., deposition/transport of black carbon)108,110,120. At moderate to high elevations, especially between 3500 and 4000 m where snow cover is rapidly decreasing, snow/ice-albedo feedback and changes in solar radiation receipt greatly contribute to warming over the Tibetan Plateau110.
North-western China is characterized by arid and semi-arid climates, and semi-arid regions in China are projected to continuously expand in the 21st century123, which will increase the risk of land degradation and desertification in the future. This vulnerability is a feature at both GW1.5 and GW214,124, North-western China shows the strongest increase in both drought area and drought intensities in the future98, and future temperature extremes in North-western China show strong changes at GW1.5 and GW2. The enhanced responses of the temperature in this region are mainly attributed to strong sensible heating and suppressed evapotranspiration as a result of limited soil moisture and vegetation14. The additional GW0.5 in North-western China is expected to increase the number of summer days, warm spell duration, warm days, and warm nights by up to 15 days, 30 days, 10% and 15%, respectively124. In terms of precipitation, north-western China will experience the largest magnitude of ~3.2%(−4.7–12.4%) from GW1.5 to GW2125. Meanwhile, arid and semi-arid regions are projected to experience threefold higher risks of extreme precipitation due to the additional GW0.5126. Enhanced surface warming in North-western China can be explained by surface and atmospheric processes controlled by low soil moisture and sparse vegetation in this region14. Other mechanisms such as dust–cloud interactions and large-scale atmosphere–ocean interactions, also tend to amplify/dampen climate change in North-western China14,124.
The Jianghuai region in eastern China which covers most of the plain between the Yangtze and Huai rivers (e.g. Anhui and Jiangsu provinces) is strongly influenced by the Asian monsoon, and because it contains a high population and many cities (e.g. Nanjing, Hefei) it suffers frequent extreme climate events causing serious loss of life and property. For precipitation and precipitation extremes, the future projection of precipitation over this region based on the nonhomogeneous hidden Markov model and self-organizing map statistical downscaling methods shows that precipitation increases over the eastern Jianghuai region while it decreases over western parts of the Jianghuai region at GW1.5. Any enhancement of precipitation intensity is more significant in the southern and western parts of the region at GW2. Meanwhile, GW0.5 (GW2–GW1.5) increases the number of heavy precipitation days by 12.1% (10–13%) over eastern regions but decreases it by 3.7% (3.3–4.3%) over the western region113,127. In the Jianghuai region as a whole, changes in future precipitation are mainly attributable to the variable occurrence of different synoptic regimes, and the overall intensification of precipitation in the region in warmer simulations is linked to a rise in the occurrence of a wet regime, which brings the South Asia High eastward and pushes the WNPSH westwards114. In future studies, more experiments specifically designed at GW1.5 and GW2 are needed to better understand future changes in climate extremes in the Jianghuai region.
Urban areas (including urban agglomerations) contain a high proportion of the population and are increasingly vulnerable to the risks of extreme weather and climate events128. The Yangtze River Delta (YRD), Beijing–Tianjin–Hebei region (BTH), and the Pearl River Delta (PRD) have been developed as three world-class urban agglomerations, and this means that more research on their vulnerability to climate change is required. Based on downscaled CMIP5 simulations129, the additional half degree of warming (GW2 compared with GW1.5) relative to 1986–2005 will lead to an increased risk for China’s urban agglomerations concerning extreme maximum temperature, total precipitation on very wet days and maximum 5-day precipitation totals by 4.1, 1.8, and 1.0 times, respectively. Both GW1.5 and GW2 show the largest increases in extreme risks in the YRD region129. However, compared to GW2, the occurrence probability of historically unprecedented compound heatwaves at GW1.5 will reduce by more than half in both BTH and YRD130, and the increasing rate of future compound heatwave days in PRD will be much larger (for GW2 and GW1.5) in comparison to BTH and YRD. Dynamical downscaling results show that BTH, YRD, and PRD all experience a more significant increase in summertime extreme maximum temperature than in mean temperature131. The increase rate in intensity, duration, and frequency of heatwaves in BTH, YRD, and PRD increases by more than 50%/70% at GW1.5/2, respectively, and YRD will receive the strongest increase in heatwave risk. This is in part due to strong but localized warming effects produced by larger concentrated urban aggregations131. The combined effects of increasing climate extremes and localized urbanization effects mean that these urban agglomerations are particularly sensitive to future warming and require much more investigation in this regard.
Scientific highlights and future research prospects
Divergent responses to GW1.5 and GW2 for transient vs. stabilized scenarios
The Paris agreement attempted to limit global warming to the 1.5/2 °C target with quasi-stabilized scenarios in this century. For a long time, there was a lack of specific GW1.5/2 coupled model simulations, and studies investigated the climate impacts of GW1.5/2 as a transient phase in CMIP5/6 model simulations which eventually warm much more than this1,7. There was an implicit assumption that there would be similar climate impacts for transient warming of 1.5/2 °C as for a stabilized GW1.5 or GW2 scenario. However, this hypothesis has been challenged by recent studies that show remarkable differences in climate response between transient and stabilized scenarios at the same temperature threshold132,133.
As different carbon emission strategies are applied, transient global warming tends to produce greater Northern Hemisphere (NH)–Southern Hemisphere (SH) thermal contrast, greater land–ocean temperature difference, weaker warming in high-latitude oceans, higher sea level rise, and less sea ice coverage compared to stabilized scenarios132,134. Differences in ocean heat storage and transport are critical to explaining the distinct climate differences between the transient versus stabilized scenarios. Ocean and atmospheric heat transport are both symmetric about the equator under the transient scenario135. In contrast, heat transport shows a significant difference between the two hemispheres in the stabilized scenario135. The divergent response of ocean heat transport between the two scenarios is caused by distinct differences in heat storage in the Southern Ocean and in the strength of the Atlantic Meridional Overturning Circulation135. Based on the comparison of results from the transient (CESM Large Ensemble, CESM-LE) experiment versus the quasi-equilibrium (CESM Low Warming, CESM-LW) experiment at GW1.5, CESM-LW exhibits the most warming in the ocean at GW1.5 at the end of the 21st century, while in the transient simulations of CESM-LE the land warms up more quickly to achieve global GW1.5 by the 2020s119. The resulting imbalance in warming between the northern and southern hemispheres and the focus of warming over land in the transient scenario leads most of the world’s population to suffer a more extreme local climate and more extreme hot events in the transient model (in comparison with the stabilized scenario), especially in summer132. Such imbalanced temperature patterns accelerate across-equatorial moisture transport, thus strengthening the boreal summer Hadley circulation and Inter-tropical Convergence Zone, as well as increasing moisture convergence in the NH land monsoon region, resulting in a more vigorous NH land monsoon under the transient scenario134.
On a regional scale, significant differences in drought intensity, dryland coverage, and temperature response are also detected between transient and stabilized global warming scenarios136,137. In East Asia, precipitation increases more significantly in south-eastern China in the stabilized scenario compared with the transient one (Fig. 7). The difference is primarily attributable to dynamic effects related to circulation changes and is again driven by differences in the land–ocean thermal contrast in East Asia between the two scenarios. The stabilized warming scenario favors larger ocean warming and smaller land warming. Enhanced warming in the sea to the south of China leads to an enhanced meridional temperature gradient and ultimately speeds up the jet stream in East Asia. The formation of ageostrophic wind at the entrance of the accelerated jet (upstream) strengthens ascending motion and therefore precipitation along the coast of Asia from southeastern China to India138. Additionally enhanced heating over the land mass of India further induces a Kelvin-wave response with the formation of an anomalous anticyclone contributing to a westward shift and intensification of the WNPSH. Southerly winds on the western flank of the anticyclone transport moisture northward from tropical oceans, and strengthen precipitation in south-eastern China138. In addition, contrasting responses of regional-scale hydrology may create differences in the biosphere and land–atmosphere interaction between transient and stabilized scenarios133,137. Thus, multiple lines of evidence confirm that the responses of the earth’s climate could be different in transient versus stabilized scenarios and that this needs to be considered when examining the global/regional impacts of GW1.5/2.
Sources of uncertainties in climate projections
There are three different sources of uncertainty for future projections using multi-model approaches: (1) uncertainty due to the future scenario of external forcings (scenario uncertainty), (2) uncertainty from climate sensitivity to external forcings (model uncertainty), and (3) natural climate variability (internal uncertainty)5,139,140. As shown in Fig. 8, model uncertainty and scenario uncertainty are the major sources of future uncertainties in temperature projections for CMIP6 model simulations across China throughout the 21st century. The contribution to the total uncertainty of internal variability and model uncertainty both decrease over time while that of scenario uncertainty increases, which can be explained by the increasing importance of anthropogenic forcing as time progresses. Model spread is the largest contributor to overall uncertainty at GW1.5 and GW2, and accounts for 62.18%/61.08% and 56.52%/53.80% of the total uncertainty under the SSP2–4.5/SSP5–8.5 scenario. Internal variability accounts for 25.21%/23.60% and 18.35%/15.95% of the total at GW1.5 and GW2, respectively. Scenario uncertainty is the smallest contributor, only accounting for 12.62%/15.32% and 25.13%/30.25% respectively. Similar results are found in CMIP5 models according to Wu et al. (2021), which indicated the largest total uncertainty is found over the Tibetan Plateau and in southern China, dominated by the pattern of model uncertainty. However, model uncertainty decreases slightly in the south of China at GW2 compared with GW1.5. Internal variability and scenario uncertainty are consistently small in the four sub-regions of China. Thus, reliable projection of surface mean temperature in China at GW1.5 mainly depends on reducing model uncertainty.
In terms of the temporal evolution of the contribution of the three uncertainties to the total uncertainty for precipitation, internal uncertainty is projected to decrease with time, while model uncertainty and scenario uncertainty are projected to increase with time. At GW1.5, internal uncertainty is the dominant source (>50% of total) for annual total wet-day precipitation and precipitation on very wet days. Uncertainty for consecutive dry days is an exception with model uncertainty accounting for more than 50% of the total uncertainty at GW1.5141. Projections of the East Asian monsoon described above (see the section “East Asian monsoon circulations”) demonstrate that large inter-model differences were found at GW1.5 compared to GW2, and this shows that model uncertainty plays a dominant role in the projection of the EASM at GW1.5141. Overall more work is required to assess the source and magnitude of uncertainties in climate projections at GW1.5 and GW2, and more efforts should be put into understanding internal climate variability on decadal time scales as part of future climate projection. In addition, it needs to be clarified that the uncertainty analysis method above only focuses on the three sources of uncertainty in transient simulations, and further research is needed on the estimated uncertainty of stabilized simulations.
Methods required to reduce future projection uncertainty
Climate model uncertainty is one of the important sources of uncertainty for future projection of regional climate, and it is of great necessity to reduce it. Many approaches can be used including climate model evaluation, multi-model ensembles (MMEs), GCM bias correction, emergent constraints, and machine learning. Although no climate model performs best in all aspects, the “best model” is usually determined by a multi-variable integrated evaluation method, which can summarize the overall performance of each climate model in simulating multiple variables142. One can evaluate model performance in terms of multi-variate climatological means and interannual variability143. Moreover, one can select a subset of climate models to reduce the uncertainty in climate projections, based on the relationship of each model’s ability to correctly simulate historical and future climate144.
The selection of an MME based on climate model evaluation is regarded as an important method to reduce model uncertainty. A weighted MME based on the performance of individual models in simulating the climatology, and internal variability of target variables, is commonly used145,146,147. Many weighting schemes have been proposed in recent years, including those based on the reliability of the ensemble average, the rank weighting method, and a weighting scheme jointly considering model performance and independence148. A further improvement takes into account the area weighting, the combination of multiple scalar and vector fields, and the use of a normalized multi-variable integrated skill score149. GCM bias corrections, together with MMEs, have great potential to improve dynamical downscaling simulations and reduce uncertainty in future climate projections150. Therefore, many GCM bias correction methods have been developed to reduce the impact of GCM bias on dynamical downscaling simulations151. These GCM bias correction methods include, but are not limited to, GCM mean bias correction, mean and variance bias correction152,153, quantile-quantile bias correction154, trend-preserving bias correction, nested bias correction155, and MME-based GCM bias correction156. Together such studies suggest that GCM bias correction can significantly improve dynamical downscaling simulations compared with a traditional dynamical downscaling approach150,152.
An emergent constraint method is a new approach to reducing future uncertainty and has received more recent attention157. Numerous emergent constraints have been identified, including studies of equilibrium climate sensitivity158, cloud feedback159, carbon cycle feedback160, the hydrological cycle161, and Arctic warming162. The proposed observable constraints involve historical trends, seasonal cycles, interannual variability, and climatological biases. Moreover, most current research generally focuses on global or hemispheric scale issues and pays less attention to regional scale or high-impact climate events.
All the above methods including climate model evaluation, MMEs, emergent constraint, and GCM bias correction rely on the performance of individual models and the concept of quantifying specific relationships or indices (usually through linear regression), potentially neglecting useful information. Machine learning is a useful tool to extract more information from multi-model data to reduce model uncertainty163. Since this approach is still in its infancy, a lot more work is required to realize its potential.