Summertime compound heat wave and drought events in China: interregional and subseasonal characteristics, and the associated driving factors

This study investigates the characteristics of compound heat wave and drought events (CHDEs) across various subregions of China from 1961 to 2022 by utilizing a monthly probability-based index. The results uncover significant interregional and subseasonal variations. The trend analysis of CHDEs reveals statistically significant increases in most regions of China; however, there is no significant trend in the JiangHuai region throughout the entire summer season. The trends across regions exhibited subseasonal differences, especially in the eastern regions (Northeast China, North China, and South China (SC)). Furthermore, the occurrence of severe CHDEs (SCHDEs) in China has significantly increased in both frequency and extent since the 1990s. Southwest China and eastern Northwest China have witnessed the highest frequency of SCHDEs, while SC has remained relatively unaffected compared to other regions. The occurrences of SCHDE demonstrate a higher frequency occurred in June than in July and August, especially in the southern regions. The local driving factors are further explored. The incidence of CHDEs in eastern China is significantly influenced by anticyclonic circulation anomalies, which span from the upper to the lower troposphere. These anomalies are crucial in shaping the dynamic and moisture conditions necessary for CHDE formation. Their specific locations dictate the unique atmospheric conditions that lead to the regional characteristics of CHDEs across eastern China.


Introduction
With global warming, both the frequency and the intensity of extreme climatic events have increased substantially in most regions of the world in the past century (Wang et al 2017, Sheridan and Lee 2018, Li et al 2018a, 2019, Li et al 2023b, Chen et al 2020, IPCC 2021, Zhu et al 2022, Fan et al 2023, Sun et al 2023), causing severe impacts on society and human lives (Lesk et al 2016, Weilnhammer et al 2021, Liu et al 2022, Heino et al 2023, Chen et al 2023a, Chen et al 2023b, Li et al 2023a).Compared to the impact of an individual extreme climatic event, a combination of different climatic events, which might lead to a compound extreme event (AghaKouchak et al 2014, Chen et al 2022, Liu et al 2022, He and Chen 2023), can often cause profound disasters (Yuan et al 2016).In recent years, compound heat wave and drought events (CHDEs) have emerged as significant climatic phenomena affecting vast regions across China.For example, a severe CHDE (SCHDE) that occurred in northeastern China in summer 2016 caused economic losses of CNY 15.61 billion (Li et al 2018b).Besides, CHDEs have profound impacts on multiple sectors including vegetation growth, agriculture, water resources, and public health (Xu et al 2017, Liu and Zhou 2021, Zhang et al 2021, Gao et al 2023, Wu et al 2023).Because an individual climatic event might be aggregated by other potential climatic events, traditional criteria adopted for extreme events might underestimate the severity of CHDEs (AghaKouchak et al 2014, Li et al 2018bLi et al , 2020a)).Therefore, identifying and characterizing the variation in CHDEs across China is essential for implementation of global warming adaptation and government decision-making.
Over the past decades, numerous studies have investigated droughts and heat waves in China from the perspective of univariate variables (Zou et al 2005, Ding et al 2010, Wei and Chen 2011, Yu et al 2014).However, there is limited consensus on characterizing the heat waves and droughts that occur simultaneously.Recent studies have revealed the characteristics of compound heat waves and droughts, typically through quantitative identification methods.Some studies set criteria where the extremes of temperature and precipitation must simultaneously exceed or fall below predetermined thresholds (Ye et al 2019, Bevacqua et al 2022).Other studies identify CHDEs using statistical models based on meteorological indicators (Kong et al 2020, Yu and Zhai 2020, Zhang et al 2022).Some findings identify CHDEs through probability statistic derived from the joint distributions of multiple single variables.Researchers can examine the effects of both dependence and interaction among multiple extreme driving factors on compound extreme events (Hao et al 2018, 2022, Li et al 2022).Variations in droughts and heat waves across different regions and months in China are significant (Xie et al 2020, Han et al 2021a, Yu et al 2021, Long et al 2022).Yet, the characteristics and drivers of CHDEs on a subseasonal time scale, as well as regional disparities, remain underexplored.
Numerous studies have investigated the largescale atmospheric circulation patterns and the physical mechanisms relating to heatwaves and drought events across China (Wang 2001, Ma 2007, Gao and Yang 2009, Yang et al 2012, Wang and He 2015, He et al 2018, Garreaud et al 2020, Li et al 2020, Zhu et al 2020, Ullah et al 2023).For example, the Polar-Eurasian teleconnection pattern plays an important role in regulating the atmospheric circulation over NEC (Li et al 2020b), which might be largely regulated by variation in sea ice in the Barents Sea ice (Li et al 2018b, An et al 2020, Han et al 2021b, Yang et al 2021, Luo et al 2023).Wang and He (2015) indicated that the Pacific-Japan teleconnection pattern causes a precipitation anomaly (north-south wet and middle dry) across eastern China.Furthermore, the Silk Road teleconnection pattern, which is strongly correlated with precipitation and surface air temperature anomalies over Eurasia (Hong andLu 2016, Hong et al 2018), is also an important cause of summertime drought in NEC (He et al 2018, Li et al 2020b).Few studies have investigated the physical processes for the CHDEs across China.Hence, a comprehensive understanding of CHDEs and the associated local driving factors is crucial for adaptation strategy preparation.
The main objectives of this study are to investigate the month-by-month spatiotemporal characteristics of summertime CHDEs across eight subregions in China and to examine the associated atmospheric circulations using a multivariate probability-based index (PI).The remainder of this paper is structured as follows.Section 2 outlines the datasets and methods employed in this study.Section 3 presents the month-by-month spatiotemporal variations in summertime CHDE intensity in China, together with analysis of the interdecadal changes and variations in the frequency of SCHDEs.Section 4 explores the atmospheric circulation anomalies linked to CHDE occurrence in different regions of eastern China.Finally, a concise discussion and the conclusions derived from this study are presented in Section 5.

Data
In this study, monthly gridded meteorological datasets (hereafter, referred to as CN05), including precipitation and 2 m air temperature (spatial resolution 1 • × 1 • ) for summer (JJA) during 1961 − 2022, were used to analyze the CHDEs in mainland China.The CN05 dataset was generated using the thinplate smoothing splines method based on data  from more than 2400 stations distributed across China (Wu and Gao 2013), which were provided by the Climate Change Research Center of the Chinese Academy of Sciences (http://ccrc.iap.ac.cn/resource/detail?id=228).Additionally, monthly reanalysis datasets (horizontal resolution: 1 • × 1 • ) derived from the fifth generation reanalysis data provided by the European Centre for Medium-Range Weather Forecasts were also employed (Hersbach et al 2020).Here, the variables include surface pressure, meridional and zonal winds, vertical velocity, and geopotential height.
This study mainly focused on the characteristics of CHDEs over eight subregions of China (You et al

Method
The non-parametric Mann-Kendall (MK) trend test, which approved by the World Meteorological Organization, was used to examine the trend of time series and significance (Bari et al 2016).In this study, the MK test was utilized to analyze the CHDE time series.Trends with a p-value less than 0.05 were deemed statistically significant (Yang et al 2020).
In this study, the joint cumulative distribution of precipitation (X 1 ) and temperature (X 2 ) temporal series was used to calculate the PI (Li et al 2022).The cumulative distribution function of precipitation (x 1 ) and temperature (x 2 ) can be expressed as F 1 (x 1 ) = P (X 1 ⩽ x 1 ) and F 2 (x 2 ) = P (X 2 ⩽ x 2 ), respectively.The joint survival function F i can be used to identify a CHDE, which is defined as F i = 1 − F i .Thus, the PI can be used to represent the joint survival cumulative distribution function as Accordingly, the PI values range from zero to unity, with smaller values representing more SCHDEs, and vice versa.
To better understand the spatial distributions of the SCHDEs across China, a SCHDE was defined as an event with a PI value smaller than 0.1.Relatively low PI values can be observed in Qinghai Province and Sichuan Province throughout the summer, suggesting that these regions are more prone to severe compound heat wave and drought conditions (figures 1(g)-(i)).Moreover, the spatial distribution of the climatological PI differs among the three summer months (figures 1(g)-(i)).In June, the eastern parts of JH and NC are much drier and hotter than other regions (figure 1(g)).In July, relatively low PI values can be observed in NEC, indicating severe compound heat wave and drought conditions in this area (figure 1(h)).In August, the eastern parts of NEC and NC experience notably severe heat wave and drought conditions (figure 1(i)).

Interregional and subseasonal trends of CHDEs across China
Figure 2 illustrates the spatial distributions of statistical trends in summer monthly precipitation, 2 m air temperature, and the PI.The overlap of areas with decreasing precipitation and increasing temperatures leads to a trend towards reduced PI across most regions in China, indicating an increasing trend in the occurrence of CHDEs.Conversely, areas with positive PI trends do not demonstrate statistical significance.Furthermore, under the influence of global warming, the linear trends of the PI are primarily driven by 2 m temperature.In addition, CHDEs show remarkable spatial variations in different months.In June, statistically significant uptrends in CHDE occurrences are primarily observed in NWC, ENC, the TP, SWC and SC.In July, NEC, NE, ENC, NEC and the TP displayed more pronounced increases in CHDE occurrence trends than those depicted in figure 2(g).Additionally, SC shows a smaller value in negative PI trends in July than in June.In August, the spatial distribution is similar to July, whereas the western part of NWC reveals more positive PI trends, although they are not significant, compared to July.Moreover, western SC demonstrates more significant uptrends in CHDE occurrences than those shown in figure 2(h).It can be inferred that increasing occurrences of CHDE are predominantly found in SC, SWC, the TP, NWC and ENC in June, while there are more evident trends in NEC, NC, SWC, the TP, NWC, and ENC in July and August (figures 2(g)-(i)).
All subregions experienced negative trends in the PI, with regional differences across eight subregions revealed (Figure S1; Table 1).For instance, the TP region exhibited larger trends than other regions (−0.042, −0.055, and −0.064 per decade in June, July, and August).In contrast, the trends in JH were weak and almost identical (−0.007, −0.005, and −0.002 per decade in June, July, and August), indicating insignificant variations in CHDE occurrence during these months throughout the study period .Subseasonal disparities were also observed.Table 1 indicated that both the NEC and NC regions exhibited notable trends in July and August, while showing insignificant trends in June.Conversely, SC exhibited significant trends in early summer (June) and late summer (August), but showed an insignificant trend in mid-summer (July).

Interdecadal variations in CHDE across eight subregions of China
To further investigate interdecadal variations in CHDEs, the 10 year running mean temporal series of the PI across eight subregions in China are presented in figure S2.It is worth noting that the northern regions (NEC, NC, NWC, and ENC) witnessed a notable shift in the 1990s during June and July, while the southern regions (JH, SC, SWC, and the TP)  experienced a shift in the 2000s, but in August (table S1).The frequency of these abrupt change points also differed among the subregions.For example, SC encountered four change points in August, occurring in 1970August, occurring in , 1980August, occurring in , 1992August, occurring in , and 2002.Meanwhile, NC in August showed no significant change points during 1961-2022.Additionally, other three window lengths (5, 9, 11) are chosen to verify the above analysis, and similar results can be found concerning characteristics of interdecadal variations (Figures and Tables not  shown).
The spatial distribution maps in figure 3 reveal the trends in the occurrence of SCHDEs, highlighting significant interdecadal variations across different subregions and periods.In the 1960s, SCHDEs were more frequent in June, especially in NEC and JH (figures 3(a)-(c)).In the 1970s, the extent and frequency of SCHDEs in June diminished, while SCHDEs occurred more frequently in NEC in July and August compared to the 1960s (figures 3(d)-(f)).In the 1980s, JH experienced fewer SCHDEs in June (figure 3(g)), while NEC experienced an increase (Figures 3(h) and (i)).Into the 1990s, SCHDEs become more frequent and wide-ranging across China (figures 3(j)-(r)), particularly in ENC, NC, and NEC (figures 3(j)-(l)).In the 2000s, SCHDEs were mainly centered in JH in June (figure 3 The interdecadal analysis of SCHDEs reveals significant increasing trends in the frequency and extent of SCHDE across China, particularly in NEC, NC, ENC, NWC, the TP, and SWC.Table 2 indicates notable disparities in the interdecadal frequency of SCHDE occurrences among subregions, the bold values suggest where the SCHDE frequency exceeded two occurrences per decade.Overall, SCHDEs were more frequent in SWC in June and July, and ENC in July and August during the past decades, whereas SC keeps the lowest value in the average occurring frequency of SCHDEs among these regions during the whole summer.However, subseasonal differences for the occurring frequency of SCHDEs are observed in the southern China (JH, SC, SWC, and the TP), with highest value in June and smaller values in July and August.

Local driving factors in eastern subregions
To document the atmospheric circulation patterns relevant to CHDEs in eastern China, regression maps of the atmospheric circulation patterns in summer against the PI during 1961-2022 are shown in figures 4-5 and S3-S4.Since a smaller value of the PI indicates a more SCHDEs, the PI value was multiplied by −1 in the following analyses.
Figure 4 shows regression maps of the circulations with regard to PI during 1961-2022 over SC in JJA.When CHDEs occurred, SC and the surrounding regions experienced warm  (Figure 4(a)).Correspondingly, the negative precipitation anomaly and positive temperature anomaly are conducive to the maintenance of compound heat wave and drought conditions.
The regional thermodynamic and dynamic processes associated with CHDEs in other eastern regions were similar to those in SC.Specifically, when CHDEs occurred over NC in summertime, northern China was dominated by an anomalous positive geopotential height center, an anomalous anticyclonic center, and an anomalous downward atmospheric motion over the region (figures 5(d)-(g)).As a result, the anomalous moisture flux divergence (Figure 5(h)) resulted in precipitation deficiency (Figure 5(b)).Meanwhile, the significant decreases of cloud cover (Figure 5(i)) led to the anomalous increased net radiation flux (Figure 5(j)), thus providing dry and hot conditions for CHDE occurrences.Similar local dynamic processes can be found in NEC and JH regions (Figures S3 and S4), with differences in the location and intensity of the anticyclones relative to the target regions.Additionally, we compare the circulation patterns relevant to CHDEs in different summer months for each target region and find no significant discrepancies in patterns (figures not shown).

Conclusions and discussion
This study analyzed the spatiotemporal variations of summertime CHDEs across different subregions of China during 1961-2022, and further explored the associated local driving factors.We observed significantly increasing trends in CHDE occurrences across most regions of China, particularly in NEC and western China.However, there is no notable significance in the JH region (figures 2(g)-(i)), contrary to other regions of China, where negative trend prevailed in those regions.In JH, both temperature and precipitation showed increasing trends in summertime during 1961-2022 (figures 2(a)-(f)), thus leading to concurrent hot and humid trend in the region.The results align with previous studies by Wang et al (2012) and Li et al (2023b).Significant increases in the frequency and extent of SCHDEs in China have been observed since the 1990s, which is similar to previous studies (Wu et al 2019(Wu et al , 2020)).Compared with previous studies, our results highlight regional characteristics and subseasonal disparities in CHDE characteristics.CHDEs were intensified in the NEC and NC regions and showed significant trends in July and August, while exhibiting an insignificant trend in June.Conversely, the trend in SC is obvious in June and August, but is insignificant in July.Furthermore, in June and July, northern regions (NEC, NC, NWC, and ENC) experienced an interdecadal abrupt in the 1990s, whereas in August, most southern regions (SC, SWC, and the TP) witnessed an interdecadal abrupt in the 2000s.Additionally, SCHDEs are prone to occur in SWC and ENC, whereas the average occurring frequency of SCHDE in SC keeps the lowest level during the three summer months compared to other regions.However, subseasonal differences in the occurring frequency of SCHDEs are observed in southern regions of China, with the highest value in June and smaller values in July and August.
Moreover, this study further explored the local driving factors related to the CHDE occurrence in eastern China.Positive geopotential height anomalies, significant wind anomalies, anomalous anticyclonic circulation, and downward atmospheric motion are evident over the target region.This atmospheric circulation pattern creates conditions favorable for moisture divergence and precipitation deficiency, which are detrimental to local cloud formation, further increase surface radiation, and provide an environment suitable for CHDE occurrence.
Notably, the locations of the anticyclonic circulations relating to the target region show regional differences, suggesting that CHDEs in different regions may be driven by different teleconnection patterns.For instance, the SC region is likely regulated by the negative phase of the Eurasian (EU) teleconnection pattern when the CHDEs occurred in summer (figure 4).By contrast, the positive phase of the EU pattern may result in the occurrences of CHDEs in NC (figure 5).In addition, the occurrences of CHDEs in the JH region may be regulated by the PJ pattern teleconnections (figure S3), while the CHDE occurrences of NEC are likely regulated by the Polar-Eurasian (POL) teleconnections (figure S4), consistent with the studies of Li et al (2018b).Additionally, we also analyzed the atmospheric circulation patterns relevant to CHDEs in western China during summertime, the regression maps of anomalous atmospheric circulation patterns show that there is an anomalous anticyclonic center over target region, with anomalous descending atmospheric motion extending from the upper to the lower troposphere, provide favorable conditions for CHDE occurrence (figures S5-S8).Our research reveals interregional disparities in summertime circulation patterns.To further understand the formation and maintenance of these critical atmospheric circulations, the physical mechanisms of related large-scale atmospheric circulations and teleconnections deserve further investigation in future work.

Figure 1
depicts the spatial distribution of climatological summertime precipitation, temperature, and the PI across eight subregions in China during 1961-2022.Throughout the entire summer, average precipitation in China gradually decreases from the southeast toward the northwest (figures 1(a)-(c)).The climatological temperature conditions in southern and northwestern China are relatively high compared to other regions, with the TP region experiencing the lowest temperatures (figures 1(d)-(f)).

Table 1 .
Mann-Kendall (MK)trend test and significance coefficient (P) of the PI in summertime during 1961-2022 (10 −1 per decade) ( * indicates regression coefficient significant at the 95% confidence level based on the MK trend test; * * indicates regression coefficient significant at the 99% confidence level based on the MK trend test.).

Table 2 .
Interdecadal frequency of SCHDE occurrence in eight subregions of China over the previous 60 years during summertime (unit: number of SCHDEs/decade).
different drought indicators Int.J. Climatol.42 8133-45 Zhang Y, Mao G, Chen C, Shen L and Xiao B 2021 Population exposure to compound droughts and heatwaves in the observations and ERA5 reanalysis data in the Gan River Basin, China Land 10 1021 Zhu B, Li H, Sun B, Zhou B and Duan M 2022 Physical-empirical prediction model for the dominant mode of extreme high temperature events in eastern China during summer Front.Earth Sci. 10 989073 Zhu B, Sun B and Wang H 2020 Dominant modes of interannual variability of extreme high-temperature events in eastern China during summer and associated mechanisms Int.J. Climatol.40 841-57 Zou X, Zhai P and Zhang Q 2005 Variations in droughts over China: 1951-2003 Geophys.Res.Lett.32 L04707