An interdecadal decrease in extreme heat days in August over Northeast China around the early 1990s

Wenjun Liu , Ruin Chen , , , , Zhiping Wen ,

a Center for Monsoon and Environment Research, Guangdong Province Key Laboratory for Climate Change and Natural Disaster Studies, School of Atmospheric Sciences,Sun Yat-sen University, Guangzhou, China

b Department of Atmospheric and Oceanic Sciences, Institute of Atmospheric Sciences, Fudan University, Shanghai, China

c Southern Marine Science and Engineering Guangdong Laboratory (Zhuhai), China

d Jiangsu Collaborative Innovation Center for Climate Change, Nanjing, China

Keywords:Northeast China Extreme heat Interdecadal Variability

ABSTRACT Northeast China (NEC) witnessed an interdecadal increase in summer extreme heat days (EHDs) around the mid-1990s. The current study reveals that this interdecadal increase only occurs in June and July, while August features a unique interdecadal decrease in EHDs around the early 1990s. Plausible reasons for the interdecadal decrease in EHDs in August are further investigated. Results show that the interdecadal decrease in EHDs in August is due to the deceased variability of daily maximum temperature ( T max ). Overall, the variation of the T max over NEC in August is modulated by the Eurasian teleconnection pattern, Silk Road pattern, and East Asia-Pacific pattern. However, the influence of the Silk Road pattern dramatically weakens after the early 1990s because the meridional wind variability along the westerly jet significantly decreases. The weakened influence of the Silk Road pattern contributes to the decreased T max variability over NEC. Meanwhile, the convection over the western North Pacific, which accompanies the East Asia-Pacific pattern, presents a significant decrease in variance after the early 1990s, further decreasing the T max variability over NEC.

1. Introduction

Against the background of global warming, extreme heat days(EHDs) occur frequently and greatly threaten human health and societal development, which concerns both the public and government( Lu and Chen, 2016 ). China has experienced nationwide warming,but also shows distinct regional features, with larger trends in northern China than in southern China ( Qi and Wang, 2012 ). The trend of the frequency of EHDs (EHDF) during 1961-2008 is largest over Northeast China (NEC), compared to the different regions in eastern China ( Zhang et al., 2016 ). NEC is the “granary of China ”,and thus a frequent occurrence of EHDs would be hazardous for plant growth and lead to considerable economic losses. Therefore,it is of great importance to understand the variation of EHDF over NEC.

NEC experienced an interdecadal increase in EHDs around the mid-1990s ( Sun et al., 2011 ; Wei and Chen, 2011 ). Both the background global warming and the changes in atmospheric circulation are important for determining regional EHDs. The interdecadal change in summer circulation around the mid-1990s features a high pressure anomaly over Northeast Asia, which leads to anomalous subsidence and dry conditions and thus favors the occurrence of EHDs ( Wei and Chen, 2011 ).This anomalous high pressure over Northeast Asia is related to the increased precipitation over South China ( Chen and Lu, 2014a ) and the interdecadal change of the Silk Road teleconnection pattern over the Eurasian continent ( Hong et al., 2017 ; Wang et al., 2017 ).

Fig. 1. (a) Distribution of the stations (dots) and the 90th percentile of summer T max series (shading; units: °C) over NEC. Time series of EHDF in (b) summer,(c) June, (d) July, and (e) August during 1960-2016 (bars; units: days). The red and black curves denote the interdecadal components of EHDF and mean T max ,respectively. The black horizontal lines denote the mean of the time series. The dashed vertical lines denote the interdecadal change points.

Previous studies have revealed the interdecadal increase in summer EHDs over NEC, no matter whether the EHDs are identified by an absolute temperature threshold or a relative threshold based on the percentile of temperature series ( Sun etal., 2011 ; Wei and Chen, 2011 ).However, the subseasonal differences in the interdecadal change of EHDs have been overlooked. Other studies concerning the interannual time scale have indicated that the summer climate in China and its influencing factors demonstrate clear subseasonal differences ( Chen and Lu, 2014b ; He et al., 2018 ; Shen et al., 2019 ). It is possible that subseasonal differences also exist in the interdecadal change of EHDF over NEC. Our analysis shows that the interdecadal change of EHDF over NEC in June and July is similar to that in summer, while the EHDF in August features a unique interdecadal decrease around the early 1990s ( Fig. 1 ). The current study explores the plausible mechanisms responsible for the unique interdecadal decrease in EHDs in August over NEC.

Previous studies have explained the interdecadal change of EHDF over NEC mainly from the perspective of the change in mean state( Sun et al., 2011 ; Wei and Chen, 2011 ). However, the occurrence of EHDs is modulated by not only the mean state but also the variability, and their relative roles vary with the studied regions.Weaver et al. (2014) proposed that the recent increase in EHDs over North America is primarily due to the warming of the mean state. In contrast, Sch?r et al. (2004) emphasized the critical role of temperature variability in the occurrence of EHDs over central and eastern Europe.The roles of mean state and variability in the interdecadal change of EHDF over NEC are examined in the current study.

2. Data and methods

Homogenized daily maximum temperature (

T

) at 762 observation stations in China is provided by Li et al.(2018) (the data are available online at https://doi.org/10.11922/sciencedb.516 ). A total of 142 stations within the region (38°-55°N,114°-135°E) are used to represent NEC ( Fig. 1 (a)). The daily 500-hPa geopotential height, 200-hPa wind, and outgoing longwave radiation(OLR) are extracted from the Japanese 55-year Reanalysis (JRA-55)dataset, with a horizontal resolution of 1.25°×1.25°( Kobayashi et al.,2015 ). The studied period is the summers (June, July, and August)during 1960-2016.

Fig. 2. Spatial distribution of the interdecadal changes in mean (a) T max (units: °C) and (b) 500-hPa geopotential height (units: gpm) in August. The dots indicate significance at the 0.1 level.

Fig. 3. (a) Histograms (bars) and Gaussian fitting PDFs (curves) of the daily series of NEC- T max for August during P1 (black line) and P2 (red line). The vertical dashed lines denote the mean T max . The orange shading denotes the warm tail exceeding 31.37 °C. (b) Spatial distribution of the interdecadal change in T max variance in August (units: °C 2 ). The dots indicate significance at the 0.1 level.

An EHD over a station is defined as a day with

T

exceeding the 90th percentile of the summer

T

series during 1960-2016 over the station (5244 days in total). A similar definition was adopted in previous studies ( Chen and Lu, 2015 ; Li et al., 2015 ). The spatial distribution of the

T

threshold is shown in Fig. 1 (a). An EHD over NEC is further identified when more than one-third of stations meet the EHD definition, and the number of these days is calculated as the EHDF over NEC.We also adjusted the criterion of one-third to half of stations, or simply averaged the number of EHDs over all the stations, to calculate the EHDF over NEC, and gained similar results (not shown). In total, there are 469 EHDs over NEC in summer, with 154 days in June, 208 days in July, and 107 days in August.A probability distribution function (PDF) is used to analyze the distribution of

T

. A Gaussian distribution is fitted to the distribution of the daily series of the mean

T

over NEC (denoted by NEC-

T

hereafter), which is acceptable according to the Kolmogorov-Smirnov test ( Massey, 1951 ). We also employed the Kernel Density Estimation to fit the original distribution ( Engel et al., 1994 ), and obtained similar results (not shown). Based on the PDF, the probability of EHDs is calculated as the cumulative probability of the warm tail exceeding the mean 90th percentile of

T

over NEC, i.e., 31.37 °C. Regression analyses are performed to investigate the circulation anomalies associated with the variation of NEC-

T

. The significance of the regressed anomalies is validated by the Student’s

t

-test. The significance of the interdecadal changes in the mean value and variance is examined by the Student’s

t

-test and

F

-test, respectively.

3. Unique characteristics of the interdecadal change in EHDF in August over NEC

Fig. 1 (b-e) display the time series of EHDF over NEC in summer and each month. According to the interdecadal components extracted by the nine-year running mean (red curves), the EHDF presents an interdecadal increase around 1993/94 for summer, June, and July, while an interdecadal decrease around 1991/92 for August. All these interdecadal changes are significant according to a sliding

t

-test.The interdecadal components of the summer mean and the monthly means of NEC-

T

are also analyzed (black curves). For summer, June and July, the mean

T

presents an interdecadal increase around 1993/94, indicating the interdecadal increase in EHDs is closely related to the increase in mean

T

. For August, although with smaller amplitude compared to June and July, the mean

T

also increases since the 1990s, which is inconsistent with the interdecadal decrease of EHDs.Therefore, the interdecadal decrease of EHDs in August might result from the change in

T

variability rather than mean

T

. The current study focuses on the unique feature of August and explores the plausible reasons for the interdecadal decrease in EHDs between 1980-91 and 1992-2006 (hereafter abbreviated as P1 and P2, respectively).

Fig. 4. V200 anomalies regressed onto the NEC- T max for August during (a) P1 and (b) P2 (units: m s ? 1 ). The purple dashed arcs denote the wave trains. (c) Interdecadal change in the V200 variance in August (units: m 2 s ? 2 ). The black contours denote the westerly jet with zonal wind speed greater than 20 m s ? 1 in (a) P1, (b) P2, and(c) P1 to P2. The dots indicate significance at the 0.1 level.

The spatial distribution of the interdecadal change in the mean

T

in August is shown in Fig. 2 (a). A positive (negative) anomaly occurs over western (eastern) NEC, resulting in a slight interdecadal change for the regional-mean

T

. The interdecadal change in monthly mean

T

could be explained by the large-scale circulation. The interdecadal change in 500-hPa geopotential height demonstrates abnormally high pressure centered over Lake Baikal and low pressure centered over the Sea of Okhotsk ( Fig. 2 (b)). A significant high-pressure anomaly occurs over western NEC, while eastern NEC is dominated by a weak and even low-pressure anomaly, favoring the warming and cooling over western and eastern NEC, respectively. In contrast, the significant high-pressure anomaly extends eastwards and covers eastern NEC in June and July, resulting in wider and stronger interdecadal warming compared to August(not shown).The PDFs of the daily series of NEC-

T

for August during P1 and P2 are illustrated in Fig. 3 (a). The PDF shifts towards the right slightly and gets narrower in P2 compared to P1, indicating a warmer mean state but smaller variability for the

T

in P2. The interdecadal increase in the mean

T

is even smaller than that of its interdecadal component as shown in Fig. 1 (e). In contrast, the interdecadal decrease in NEC-

T

variance is significant at the 0.01 level according to the

F

-test. The decreased

T

variability inhibits the occurrence of EHDs and overwhelms the effect of the increased mean

T

. Consequently, the probability of EHDs reduces from 1.49% in P1 to 0.69% in P2, which is calculated as the cumulative probability of the warm tail. Therefore,it is confirmed that the decreased

T

variability is critical for the interdecadal decrease in EHDs in August. The spatial distribution of the interdecadal change in

T

variance shows that a negative anomaly occurs over most parts of NEC, except the northern corner and a few stations in the south ( Fig. 3 (b)), with 116 stations out of 142 stations featuring decreased variance.

4. Circulation anomalies responsible for the interdecadal decrease in NEC- T variability in August max

In this section, the circulation anomalies associated with the variation in NEC-

T

in August during P1 and P2 are compared, so as to investigate the mechanisms responsible for the interdecadal decrease in

T

variability over NEC. All the analyses are performed on the daily series in August.

Fig. 5. OLR anomalies regressed onto the NEC- T max for August during (a) P1 and (b) P2 (units: W m ? 2 ). (c) Interdecadal change in the OLR variance in August(units: W 2 m ? 4 ). The dots indicate significance at the 0.1 level.

Fig. 4 (a,b) show the 200-hPa meridional wind (V200) anomalies regressed onto the NEC-

T

during P1 and P2, respectively. For P1,there are two wave trains over the mid-high latitudes ( Fig. 4 (a)). The wave train in the north is similar to the Eurasian teleconnection pattern ( Wallace and Gutzler, 1981 ; Wang and He, 2015 ), which propagates southeastwards from northern Europe to Northeast Asia. The wave train in the south is similar to the Silk Road pattern ( Lu et al., 2002 ;Enomoto, 2004 ), which propagates along the westerly jet around 40°N.The Eurasian teleconnection pattern and the Silk Road pattern jointly induce an anomalous anticyclone over NEC and modulate the NEC

T

. In comparison, for P2, the variation of NEC-

T

is only associated with the Eurasian teleconnection pattern, while the Silk Road pattern disappears ( Fig. 4 (b)). The Eurasian teleconnection pattern alone induces an anomalous anticyclone over NEC during P2, but the V200 anomalies around NEC shrink northwards in P2 compared to P1. It is indicated that the influence of the Silk Road pattern on the NEC-

T

dramatically weakens in P2, contributing to the decrease in NEC-

T

variability.

Fig. 4 (c) further displays the interdecadal change in V200 variance,so as to examine the interdecadal change in the teleconnection activity.It turns out that most regions along the westerly jet over the Eurasian continent present significant interdecadal decreases in the V200 variance. Since the Silk Road pattern is characterized by alternate southerly and northerly anomalies along the westerly jet, it is suggested that the Silk Road pattern has much smaller variability and thus is less active in P2 compared to P1. The less active Silk Road pattern could weaken its effect on the climate variation over the downstream regions including NEC.

Apart from the influence of the mid-high latitude teleconnection, the variation in NEC-

T

might also be modulated by the tropical circulation. Fig. 5 (a,b) shows the OLR anomalies regressed onto the NEC-

T

during P1 and P2, respectively. For both P1 and P2, a meridional seesaw pattern occurs over the western North Pacific and East Asia, resembling the East Asia-Pacific pattern ( Nitta, 1987 ; Huang, 1992 ). The abnormal convection over the western North Pacific favors the anomalous subsidence over NEC and thus modulates the NEC-

T

.Fig. 5 (c) exhibits the interdecadal change in OLR variance, so as to examine the change in the tropical convection activity. The OLR variance shows a significant interdecadal decrease over a large domain of the western North Pacific. The decrease in convection variability extends from the Philippines to the south of Japan, covering most of the tropical regions that have an impact on the NEC-

T

. Therefore, the interdecadal decrease in NEC-

T

variability is also contributed by the decreased variability of the convection over the western North Pacific.

5. Conclusions

Previous studies have revealed that the summer EHDs over NEC experienced an interdecadal increase around the mid-1990s. The current study points out that this interdecadal increase only occurs in June and July, while the EHDs in August feature a significant interdecadal decrease around the early 1990s. The unique interdecadal decrease in EHDs in August is due to the decrease in

T

variability, while the mean

T

shows a slight increase.The interdecadal decrease in the NEC-

T

variability in August results from the modulation of both the mid-high latitude and the tropical circulation. During P1, the NEC-

T

is influenced by the Eurasian teleconnection pattern and the Silk Road pattern over the mid-high latitudes and the East Asia-Pacific pattern originating from the tropics.During P2, the NEC-

T

is influenced by the Eurasian teleconnection pattern and the East Asia-Pacific pattern. The disappearance of the connection with the Silk Road pattern in P2 is due to the significant decrease in V200 variability along the westerly jet, indicating a less active Silk Road pattern. The dramatically weakened influence of the Silk Road pattern contributes to the decreased NEC-

T

variability. Meanwhile,the convection over the western North Pacific, which accompanies the East Asia-Pacific pattern, shows significantly decreased variability in P2 compared to P1, further decreasing the NEC-

T

variability.

The change in summer EHDs in China against the background of global warming has been explored by many previous studies, and the role of the shift in the mean state has been emphasized ( Wei and Chen, 2011 ; Su and Dong, 2019 ). The current study indicates that the change in regional EHDs might present clear subseasonal differences,and the role of variability and the changes in the influencing circulation patterns should be considered. Understanding the subseasonal differences in the change of regional EHDs is helpful for the planning of farming in agricultural regions such as NEC.

Funding

This study was supported by the National Key R&D Program of China [grant number 2016YFA0600601], the Guangdong Basic and Applied Basic Research Foundation [grant number 2020A1515011572],and the National Natural Science Foundation of China [grant number 41605027 ].

Acknowledgment

We sincerely thank the anonymous reviewers for their comments and suggestions, which helped to improve the paper.