More frequent and widespread persistent compound drought and heat event observed in China

Compound drought and heat event (CDHE) causes severe impacts on agriculture, ecosystem, and human health. Based on daily maximum surface air temperature and meteorological drought composite index data in China, changing features of CDHEs in warm season from 1961 to 2018 is explored at a daily time scale based on a strict and objective definition in this study. Results reveal that CDHEs have occurred more frequently and widely in China, especially since the late 1990s. Notably, such changes are more obvious in Southwest China, eastern Northwest China, northern North China, and the coastal area of southeastern China. A prominent feature is that persistent CDHEs on a daily scale have increased significantly. To better understand climate change of compound extreme events, further studies on the physical mechanism, especially attribution analyses at a regional scale, are urgently needed.


changes in compound drought and heat events
Based on the long-term time series of daily MCI and daily maximum surface air temperature, the trends in drought (Fig. 1a) and heat events days (Fig. 1b) are separately analyzed firstly. Results indicate that number of drought days illustrates a decreasing trend in southeastern China and Northwest China from 1961 to 2018, while it shows an increasing trend in Southwest China and North China (Fig. 1a). Such a changing pattern is generally consistent with that the conclusion on changes in precipitation by Zhai et al. 34 . As for Southwest China and North China, dry-spell has prevailed in recent years, thus triggered drought events frequently happened 33 . For the heat events, Fig. 1b reveals that the number of heat events days shows a widespread increasing trend in China. Different from drought, it has increased significantly in Northwest China, southeastern China, South China, and the Yangtze River Valley. However, there exists a negative trend in the number of heat event days in southern North China, where the downtrend is possibly reflected as a regional response to inter-decadal and inter-annual variability 35,36 . Also, some other studies suggested that such a unique regional changing feature under global warming is possibly attributed to changes in aerosol, and regional land use and management 1 .
As for drought and heat events, Fig. 2a displays the spatial distribution of the average number of CDHE days in the warm season. As illustrated, CDHEs increase in frequency from south to north in China. Notably, the average increase is more than 10 days in northern North China, Northwest China, and Southwest China. It is more frequent in the arid or semi-arid climate zones. Figure 2b provides the trend distribution for the number of CDHE days. The results indicate that the number of CDHE days has increased in most parts of China with a trend of more than 0.8 day/10a. However, it has decreased in southern North China and some eastern parts of Northeast China. Similar evidence is also identified by some regional research studies in China [37][38][39] .
To further understand why some regions in China show decrease trends in CDHEs, we have compared with evidence shown in Fig. 1a,b. Obviously, the decreasing trend in the number of CDHE days in southern North China is mainly because of decrease in the number of heat event days, that in the eastern part of Northeast China  www.nature.com/scientificreports/ and the Middle and Lower Reaches of the Yangtze River is mainly due to increase in precipitation. Nevertheless, China has experienced more frequent and widespread CDHEs in general. Additionally, the time series of the annual number of CDHE days in the warm season is shown in Fig. 3. For China as a whole, the frequency of CDHEs has increased with a trend of 5%/10a, indicating that CDHEs has happened more frequently. The occurrence of CDHEs is relatively lower during the early 1970s to mid-1990s, but it has increased quickly since the late 1990s. Regionally, as shown in Fig. 2b, the increased frequency of CDHEs is mainly found in Southwest China, eastern Northwest China, northern North China, and the coastal area of southeastern China. To further investigate regional change features, time series in the number of CDHE days for the stations with an identified positive trend in Fig. 2b (regard as selected area) is established. Results show that the trend in the occurrence of CDHEs in such stations is about 31%/10a, which is 5-time larger than the whole China average. Moreover, a greater increase in CDHEs frequency is observed since the late 1990s with the top 10 highest frequencies mostly in the twenty-first century. It indicates that compound events happen more frequently and significantly in Southwest China, eastern Northwest China, northern North China, and the coastal area of southeastern China, especially since the late 1990s.
To further analyze the change in spatial coverage of CDHEs, the time series of spatial area for CDHEs of different duration are provided in Fig. 4. As shown, the spatial extension of CDHEs is smaller than normal in the 1970s to mid-1990s, but larger in 1960s and late 1990s to 2010s. There is a non-significant increase trend (0.1%/10a) in spatial area of CDHEs. However, the increasing trend is significant for persistent CDHEs. Specifically, the trend of the area coverage for persistent CDHEs is about 1.2%/10a in China from 1961 to 2018. It implies that persistent CDHEs are occurring in more regions in China. Based on Fig. 3 and 4, it can be concluded that CDHEs have happened more widely and persistently in China, especially since the late 1990s.
As supported by the the previous studies 40 , the change in CDHE frequency and spatial extension before and after the late 1990s is likely related to the significant increase in both heat events and drought events in China since the late 1990s. The possible causes of the increase in China's heat events and drought events have been explained to be related to anthropogenic global warming and internal variabilities, such as the Pacific Decadal Oscillation and Atlantic Multidecadal Oscillation [41][42][43][44] .
Furthermore, the duration structure of CDHEs has changed since the late 1990s (Fig. 5). The change in the fractional contribution of non-persistent CDHEs shows more than normal before the 1990s (Fig. 5a). However, it has become less thereafter. Meanwhile, the fractional contribution of persistent CDHEs is less than normal before the 1990s and then it has increased since late 1990. As it is shown in Fig. 5b, the increasing trends for persistent CDHEs between 6 and 10 days are significant from 1961 to 2018. It can be concluded that fractional contribution has decreased for non-persistent CDHEs, while it has increased for persistent CDHEs. In general, CDHEs in China tend to happen more persistently in the warm season from 1961 to 2018.

conclusion and discussion
In this study, a daily index reflecting the compound drought and heat event (CDHE) has been defined for the warm season based on the most updated daily temperature and drought index used operationally in China. This is different from the past studies which were based on a longer timescale drought background. It allows to more objectively reflect the changes in CDHEs and their persistence at daily timescale. Results reveal that CDHEs in China have become more frequent and more widespread, especially since the late 1990s. The increasing trend in www.nature.com/scientificreports/ the frequency of CDHEs in China is 5%/10a. We have also seen distinct regional features in the change of CDHEs in China as the increased frequency of CDHEs is more obvious in Southwest China, eastern Northwest China, northern North China, and coastal area of southeastern China, with a trend about 31%/10a. Such changing feature seems to link to changes in both heat events and drought events. Generally, either heat events or drought events in China has been increased since the late 1990s, possibly due to the combined effects of long term trend of global warming and multi-decadal/multi-annual variability (e.g. Pacific Decadal Oscillation, Atlantic Multidecadal Oscillation, and El Nino) 42,43,45 . As a result, CDHEs have increased spatially and temporally.
Furthermore, CDHEs are found to have become more persistent. As CDHE is a concurrent event of drought and heat, some studies have already found that both heat and drought events in China have been prolonged during the past decade 43,46 mainly because of global warming, the influence of ocean conditions, and regional land-use and land management changes. Noticeably, there may exist positive feedback between heat events and drought events 47,48 , especially under the global warming background. Self-intensification and self-propagation of compound events will exist with the aid of local and teleconnection feedback 49 . Possibly due to such reasons, CDHEs tend to happen more persistently under the background of increased drought and heat events since the late 1990s. Besides, soil moisture deficits before the onset of concurrent heat and dry extreme may also play an important role 17 . Understanding of these mechanisms will greatly support explain why CDHEs have increased and prolonged in China, especially in arid and semi-arid regions in China.
However, the interplay of multi-time scale variability and anthropogenic forcing on CDHEs is a complicated issue. Attribution study for change in compound extreme events such as CDHEs is still a challenging research topic. Meanwhile, a comprehensive understanding of the feedback between heat events and drought events under  www.nature.com/scientificreports/ the circumstance of CDHEs in China is insufficient and unclear at present. All these related studies in the future will improve the understanding of change in CDHEs at a regional scale and support climate change projection for extreme events, especially compound extreme events.

Data and methods
Data and definition. A long-term daily surface air temperature dataset consisting of 2,419 stations from 1961 to 2018 is obtained from China National Meteorological Information Center (NMIC). This dataset has been conducted rigorous quality control 50 . It has been widely used in the previous studies 51,52 . Additionally, an established drought index, the MCI dataset of 2,423 stations from 1961 to 2018 is provided by the National Climate Center of China. Different from some widely used drought index, such as standardized precipitation index (SPI) 53 and Palmer drought severity index (PDSI) 54 , MCI provides a more objective drought index of daily scale and has been operationally used for drought monitoring in China for over 10 years 30,31 . Furthermore, some stations are not taken into account if: (1) they do not cover for the whole period from 1961 to 2018, (2) stations have experienced relocation with their horizontal change beyond 20 km or vertical change over 50m 33 , (3) total percentage of missing data during the period is over 5%, (4) stations only include temperature data or MCI data. After those strict procedures, 1,250 stations are selected for this study. Based on MCI and maximum surface air temperature from 1961 to 2018, CDHE days are defined objectively as concurrent events of heat days and drought days (Related definitions are illustrated in Table 1).
Here, to derive a universal definition and to easily conduct spatial comparison for entire China, 95 percentile of daily maximum air temperature (Tmax 95 ) in the warm season of 1961-1990 is used to define a heat day event. Based on this threshold, Tmax 95 in about 90% stations is higher than 30 °C and in about 45% stations over 35 °C. Only in few stations in the places over 3000 km elevation of the Tibetan Plateau region, Tmax 95 is slightly below 25 °C. Nevertheless, such threshold has been also used to define high temperatures for a heat event locally 55 . Also, the percentile threshold definition for heat day is used in many recent studies related to compound event 28,56 . Methods. To analyze the persistence change of the compound event, CDHEs are divided into 14 categories based on the number of persistent days from 1 to 14 days. CDHE lasting for more than 14 days is allocated into the category of 14-day CDHE. Additionally, for a given category K, the fractional contribution is calculated as the ratio of the total number of K-day CDHEs to the total number of CDHEs.
Following Zolina et al. 32 , normalized anomalies of the fractional contribution of CDHEs with different duration to the total number of CDHE days are calculated, derived from all selected stations in the warm season and smoothed with a 5-year running mean. If the fractional contribution of CDHEs with longer (shorter) duration is found to increase, it means that CDHEs tend to happen in a more persistent (non-persistent) way.
Additionally, the in situ observation network is dense in eastern China but scarce in western China. To avoid spatial sampling bias in whole country statistics, the percentage of the spatial coverage of CDHEs is calculated as a ratio of areal coverage of CDHEs to the valid observation area in China based on averages inside 1° × 1°grids. The same approach also applied for persistent CDHEs's change study. www.nature.com/scientificreports/