Fei WANG, Lifang SHENG*,3, Xiadong AN, Haixia ZHOU, Yingying ZHANG,Xiaodong LI, Yigeng DING, and Jing YANG

1Department of Marine Meteorology, College of Oceanic and Atmospheric Sciences,Ocean University of China, Qingdao 266100, China

2Key Laboratory of South China Sea Meteorological Disaster Prevention and Mitigation of Hainan Province, Haikou 570000, China

3Ocean-Atmosphere Interaction and Climate Laboratory, Key Laboratory of Physical Oceanography,Ocean University of China, Qingdao 266100, China

4Sansha Meteorological Bureau of Hainan Province, Sansha 573199, China

5Professional Meteorological Science and Technology Service Center of Hainan province, Haikou 570000, China

ABSTRACT This study reveals a significant positive connection between autumn non-tropical-cyclone heavy rainfall on Hainan Island and the intensity of Eastern Pacific (EP) El Niño events. That is, the amount of rainfall in super EP El Niño years is more than that in normal EP El Niño years. Comparing EP El Niño years of different intensities, the cooler sea surface temperature in the northwestern Pacific during super EP El Niño years stimulates a negative surface latent heat flux (LHF)anomaly and abnormal anticyclonic circulation at 850 hPa. Under these conditions, an abnormal zonal vertical circulation develops in the northern South China Sea once a positive LHF anomaly and abnormal cyclonic circulation (ACC) at 850 hPa occur in the Beibu Gulf. The abnormal zonal vertical circulation further strengthens the ascending motion over Hainan Island, as the critical factor that leads to excessive rainfall. Further analysis shows that the positive LHF anomaly, which can be attributed to the increased latent heat transfer which resulted from the increased surface wind speed, is an important trigger for the ACC. However, the ACC is also the supplier of favorable moisture conditions because it intensifies vapor convergence over Hainan Island and meridionally transports moisture from the South China Sea to northeastern Hainan Island, thereby generating heavy rainfall. This paper emphasizes that the impact of El Niño events, especially super El Niño events, on rainfall over Hainan Island cannot be ignored, even if the traditional view is that frequent rainfall occurs mainly in La Niña years.

Key words:rainfall, super El Niño, abnormal zonal vertical circulation, Hainan Island

1.Introduction

Autumn rainfall on Hainan Island that is not associated with a tropical cyclone (TC), i.e., non-tropical-cyclone (non-TC) rainfall, has attracted a considerable amount of research interest. This non-TC rainfall accounts for ＞60% of total autumn rainfall on Hainan Island (Ren et al., 2006; Jiang and Zipser, 2010; Feng et al., 2017; Wang et al., 2021), and non-TC rainfall intensity is sometimes greater than TC rainfall intensity. Previous researchers have developed some understanding of the formation mechanism of autumn non-TC rainfall at the synoptic scale. They suggest that the cold continental high over China and warm intertropical convergence zone (ITCZ) over the South China Sea (SCS) are responsible for the formation and development of non-TC rainfall (Yen et al., 2011; Feng et al., 2017; Wang et al.,2021). The northeastern cold air carried by the Chinese continental high not only converges with the warm air carried by the ITCZ, resulting in frontogenesis and enhanced instability(Yokoi and Matsumoto, 2008; Srock and Bosart, 2009), but also further strengthens the ITCZ to form a vortex in the northern SCS (Chang et al., 2005). These conditions provide favorable dynamic and thermal fields for the generation of rainfall over Hainan Island. In addition, in autumn, under the combined influence of the Chinese continental high,ITCZ, and western Pacific subtropical high (WPSH), a lowlevel easterly jet becomes established over the northern SCS,and this is an important factor for the triggering of autumn rainfall on Hainan Island (Liu et al., 2010; Feng et al.,2016a). Nevertheless, there is insufficient understanding of the interannual variability of autumn non-TC rainfall on Hainan Island.

El Niño-Southern Oscillation (ENSO) plays a profound and complex role in the interannual variability of autumn rainfall in China (Xiao et al., 2015; Xu et al., 2016; Zhu et al.,2020). Much research has demonstrated that an anomalous southwesterly wind with strong vapor transportation caused by the “subtropical high-rainfall-anticyclone” feedback mechanism during El Niño events is the primary driver of autumn rainfall over southern China (Wang et al., 2000,Wang et al., 2020a; Wu et al., 2003; Xu et al., 2016; Wen et al., 2019), particularly in Guangdong, Guangxi, Fujian, and Jiangsu provinces. However, the rainfall over Hainan Island is obviously different from that in southern China. A traditional conclusion supported by many studies is that excessive rainfall over Hainan Island occurs during the autumn of La Niña years and that the increase in TC activity over the Pacific at this time is an important factor (Chen et al., 2004;Feng et al., 2013; Hu et al., 2020). Meanwhile, numerous other factors, including an abnormal cyclone in the northern SCS, a long-lived summer monsoon, the activated quasibiweekly oscillation (Hu et al., 2020) and Madden-Julian oscillation (MJO; Feng et al., 2013), the frequent northward movement of the ITCZ (Wang et al., 2021), and an enhanced WPSH (Niu and Li, 2008) can also increase non-TC rainfall amounts in La Niña years. Consequently, it seems possible that there is less rainfall on Hainan Island in the autumn of El Niño years because of the absence of these aforementioned factors that promote rainfall generation. Nevertheless, a consensus has yet to be reached on whether El Niño events are conducive to autumn rainfall on Hainan Island. For instance, Li et al. (2012) suggested that a warm pool would form in the Beibu Gulf because less cold air is supplied from the continent in the autumn and winter of El Niño years, and this might be conducive to rainfall generation over Hainan Island. Furthermore, the statistics related to non-TC rainfall in autumn from 1979 to 2020 show that there was also excessive rainfall during several El Niño events, including 1997, 2002, 2009, and 2015 (not shown).Therefore, the impact of El Niño events on autumn rainfall over Hainan Island remains elusive.

Aside from the differences in rainfall patterns between La Niña and El Niño events, the characteristics of regional rainfall also vary with the different types (i.e., the Central Pacific (CP) and Eastern Pacific (EP) El Niño events) and the intensities of the El Niño events (Feng and Li, 2011,2013; Zhang et al., 2011; Xie et al., 2012; Heng et al., 2020).Feng et al. (2016b) investigated the impacts of the two types of El Niño events on rainfall over southern China and found that rainfall increased significantly in southern China during the autumn of the developing phases of EP El Niño events.This increase in rainfall was the result of the anomalous southwesterly winds in southern China induced by the westward extension of the WPSH. Notably, among EP El Niño events,so-called super El Niño events have an extremely warm sea surface temperature (SST) anomaly in the tropical Pacific and are of considerable public concern because they can lead to more severe global disasters when compared with normal El Niño events (Smith et al., 1999; Kumar and Kamra,2012; Zhang et al., 2016). Since the late 1970s, three super El Niño events (i.e., 1982/83, 1997/98, and 2015/16) have been well documented. Lei et al. (2021) pointed out that there was enhanced convection over the western Pacific during the boreal winter of the 2015/16 super El Niño event,despite El Niño events usually being accompanied by suppressed convective activity over the western Pacific owing to the strong descending motion associated with a weakened Walker Circulation. Unfortunately, whether or not these super El Niño events cause excessive autumn rainfall, especially non-TC rainfall over Hainan Island, remains enigmatic. In addition, it is yet to be clarified whether a local circulation affects autumn rainfall in super El Niño years (e.g.,a local meridional circulation formed by an upward stream on the south coast of China and a downward stream in the Southern Hemisphere that can intensify rainfall over southern China) (Wang et al., 2020b; Lau and Nath, 2006; Wu et al.,2010, Gu et al., 2015, Qin and Wang, 2015). The present study is motivated by the abovementioned questions and aims to determine the various characteristics of autumn non-TC rainfall over Hainan Island during EP El Niño periods of different intensities and identify the causes of excessive autumn rainfall during super El Niño events.

2.Data and methods

2.1.Data

The EP El Niño events discussed in this paper refer to the statistical assessment of ENSO historical events by the National Climate Center (https://cmdp.ncc-cma.net/pred/cn_enso_index.php). The ENSO events are divided into weak, medium, strong, and super events, based on their peak intensities. The statistical results indicate that there have been two weak, two medium, and three super EP El Niño events since 1979. We regard the weak and medium events as normal events. So, three super (i.e., 1982/83, 1997/98, and 2015/16) and four normal (i.e., 1979/80, 1987/88,1991/92, and 2006/07) EP El Niño events were used in this study to investigate the relationship between EP El Niño events and autumn rainfall on Hainan Island.

Daily rainfall data from 18 national stations on Hainan Island provided by the Hainan Meteorological Bureau were used to explore the linkage between autumn rainfall on Hainan Island and EP El Niño events and to compare the characteristics of non-TC rainfall during super and normal EP El Niño periods. Here, selection of non-TC events was based on the TC best-track data provided by the China Meteorological Administration (Lu et al., 2021) and method provided by Wang et al. (2021). For this study, we limited the region within which TCs were assumed to affect Hainan Island to the area bounded by 14°-25°N and 100°-120°E.

Reanalysis data, including wind, geopotential height,air temperature, and specific humidity, with a temporal resolution of 1 d and a spatial resolution of 2.5° × 2.5° were used to analyze weather conditions. The dataset was obtained from the National Centers for Environmental Prediction(NCEP; Kalnay et al., 1996). The monthly mean Niño-3 index, provided by the Climate Prediction Center (CPC),was used to calculate the correlation coefficient between autumn rainfall and the intensity of the EP El Niño events.

2.2.Methods

We used the percentile method described by Bonsal et al. (2001) to sift the heavy rainfall events. Daily rainfall data from the 18 stations on Hainan Island for September and October over the period 1979-2019 were first ranked in ascending orderX1,X2,...,XN. The probabilitypthat a random value is less than or equal to the rank of that valueXMis estimated as:

We regarded the value at the 90th percentile as the threshold for heavy rainfall (corresponding top=90%).Nis the number of this order, which contains 40 112 members after omitting the missing value. The mean value (i.e., 32 mm)between the 36 101st-ranked and 36 102nd-ranked values was determined as the threshold becauseMwas 36 101.45.A heavy rainfall event was defined as one in which at least two stations recorded rainfall amounts of ＞32 mm. In total,59 and 46 days with non-TC heavy rainfall events were identified in super and normal EP El Niño years, respectively.

The moisture flux anomaly (Vq)′is decomposed into the following components (Wang et al., 2013), rather than usingV′q′directly:

Here,Vandqrepresent the wind vector and specific humidity, respectively.Vcontains the zonal (u) and meridional(v) winds. thesubscriptM istheclimatologicaldaily mean, andtheV′=V-VM,q′=q-qMrepresentthevariation from the climatological daily mean.

Linear baroclinic models (LBMs; Watanabe and Kimoto, 2000) have been used in previous studies to examine the atmospheric response to idealized diabatic heating (e.g.,Cui et al., 2015; Hu et al., 2020). In this study, the background of the model consisted of the primitive equations linearized with respect to the autumn climatology from the NCEP reanalysis for 1981-2010. There were 20 sigma levels set in the LBM. We aimed to explore the impact of the surface latent heat flux (LHF) on the abnormal cyclonic circulation (ACC)near Hainan Island. Therefore, the lower level was heated(level = 1000 hPa). With the dissipation terms used, the model response took 30 days to approach a steady state.

To identify the main source of moisture to Hainan Island, the average moisture flux anomaly (Q) at the four boundaries of the Hainan Island region (18°-21°N, 108°-111.5°E) was calculated following the equations from Sun et al. (2011):

3.Salient features of rainfall over northeastern Hainan Island in super EP El Niño years

Between 1979 and 2020, the variation in rainfall over Hainan Island and the Niño-3 index during the autumns of all EP El Niño years followed similar trends. The correlation coefficient between them is 0.72, which indicates that the stronger the EP El Niño events, the more autumn rainfall there was on Hainan Island (Fig. 1). We also separated total rainfall into TC and non-TC rainfall categories to investigate the relationship between the intensity of the EP El Niño events and these two types of rainfall. This showed that there was no significant correlation between TC rainfall and the Niño-3 index, but the correlation coefficient between non-TC rainfall and the Niño-3 index was 0.84 (Fig. 1). In other words, the increased rainfall amounts during the autumns of super EP El Niño years (i.e., 1982, 1997, and 2015) can be attributed mainly to the contribution from non-TC rainfall. This consequence is still valid after removing the long-term trend of the autumn rainfall on Hainan Island.To verify the reliability of this conclusion, we calculated the correlation between the daily precipitation data obtained from the CPC and the Niño-3 index, and the results were consistent with those based on the station data.

To investigate the characteristics of the non-TC rainfall associated with EP El Niño events of different intensity, as well as the horizontal distribution of non-TC rainfall, Fig. 2 shows a composite of the 59 and 46 heavy rainfall events from super and normal EP El Niño years, respectively. We found that more rainfall occurred in northeastern Hainan Island during super EP El Niño years (Fig. 2a), which is an unusual distribution pattern for autumn rainfall on Hainan Island. In contrast, during normal EP El Niño years, autumn rainfall was concentrated over southeastern Hainan Island(Fig. 2b), which is consistent with the typical pattern of rainfall caused by the convergence of the easterly winds on the east side of Wuzhi Mountain in autumn (Feng et al., 2016a).Their differences in rainfall distribution further suggest that excessive rainfall in super EP El Niño years occurs mainly in northeastern Hainan Island, which passes the 90% significance test (Fig. 2c). Therefore, the question arises as to what causes the increase in non-TC rainfall over northeastern Hainan Island during the autumn of super EP El Niño years.To address this, atmospheric instability and moisture transportation are discussed below.

Fig. 1. Variation of non-TC rainfall (blue bar, units: mm), TC rainfall (grey bar, units: mm),and Niño-3 index (line and labels, units: °C) in September and October of the EP El Niño years. 0.72, 0.84, and -0.07 are the correlation coefficients between the Niño-3 index and total, non-TC, and TC rainfall, respectively. The correlation coefficients pass 90%significance test.

Fig. 2. Distribution of composited heavy non-TC rainfall (units: mm) over Hainan in autumn of (a) super and (b) normal EP El Niño years. (c) Rainfall difference between (a) and (b). White dots indicate that the area passes 90% significance test.

4.Impact of an abnormal vertical circulation in the northern SCS on rainfall and its formation mechanism

4.1.Abnormal zonal vertical circulation further enhances the ascending motion

Atmospheric instability plays a crucial role in rainfall.Here, a strong ascent speed was used to recreate this dynamic instability. By depicting the average vertical section of the vertical velocity anomaly between 18°N and 21°N under different EP El Niño backgrounds (Fig. 3), we found that the abnormal ascent speed between 105°E and 115°E in super EP El Niño years was stronger than that in normal EP El Niño years once non-TC rainfall occurred. This was associated with an abnormal zonal vertical circulation established in the northern SCS during super EP El Niño years (Fig. 3a),which helped to further intensify the ascending motion near Hainan Island. The abnormal zonal vertical circulation was formed by the combination of the rising branch between 105°E and 115°E and the sinking branch between 120°E and 130°E. However, the abnormal zonal vertical circulation was absent due to the lack of descending motion around 120°-130°E in normal EP El Niño years (Fig. 3b). Thus, the question regarding the cause of the increased rainfall during super EP El Niño years can be refined to one of investigating how the abnormal zonal vertical circulation becomes established when non-TC rainfall occurs.

4.2.Establishment mechanism of the abnormal zonal vertical circulation

To explore the establishment mechanism of the abnormal zonal vertical circulation in the northern SCS during super EP El Niño periods, we analyzed the surface LHF anomaly,the wind and geopotential height anomalies at 850 hPa, and the divergence and divergent wind anomalies at 100 hPa.Figure 4 shows a dipole mode in the northern SCS causing the abnormal zonal vertical circulation to establish a positive surface LHF anomaly and an ACC at 850 hPa, accompanied by the divergence anomaly at 100 hPa, which enhances the uplifting flow via the pumping effect near the Beibu Gulf.Meanwhile, a negative surface LHF anomaly and an abnormal anticyclonic circulation (AAC) at 850 hPa, accompanied by the convergence anomaly at 100 hPa, cause atmospheric subsidence around the Bashi Strait, which provides the background for the formation of the abnormal zonal vertical circulation.

The reason why the downdraft around the Bashi Strait provides the background for the formation of the abnormal zonal vertical circulation is that the negative LHF anomaly and AAC at 850 hPa, as a Rossby wave response to the cooler SST in the northwestern Pacific, keep influencing the northern SCS in autumn, which is a salient feature in EP El Niño years (Wang et al., 2000; Feng and Li, 2011). However, after careful scrutiny, we found that the aforementioned negative LHF anomaly and the AAC developed mainly during the autumn of super EP El Niño years but were not obvious during normal EP El Niño years (Fig. 5). Therefore, the abnormal zonal vertical circulation could (could not) be established in the northern SCS with (without) the formation of a sinking branch around the Bashi Strait during super (normal) EP El Niño years.

The above analysis confirms that the positive surface LHF anomaly and the ACC at 850 hPa near the Beibu Gulf,which together cause the vertical uplift, are the pivotal elements that trigger an abnormal zonal vertical circulation in northern SCS during super EP El Niño years. To investigate the formation and development of the positive LHF anomaly and the ACC, Figs. 6a-d show the composite of the surface LHF and wind field anomalies before rainfall.The reference time for the composite (0 d, Fig. 6e) was the day when non-TC heavy rainfall occurred. We found that an AAC, the typical circulation during the autumn of super EP El Niño years, controlled the northern SCS at first (Figs. 6a,b). As the southwesterly wind in the northern SCS was enhanced by the strengthened Somali cross-equatorial flow and converged with the northeasterly wind from the Chinese continent, an ACC formed near the Beibu Gulf, which affected the whole of Hainan Island. The ACC not only strengthened the local vertical upward motion, but also advanced the development and eastward movement of the AAC, promoting an abnormal zonal vertical circulation to develop before the heavy rainfall (Figs. 6c, d). In fact, a weaker positive LHF anomaly and an ACC would also be formed with strong upward motion during non-TC heavy precipitation in normal EP El Niño years (Fig. 6f). However,the center of the ACC is situated on the south side of Hainan Island under the suppression of an AAC over China(Fig. 6f; Wang et al., 2021), resulting in a weak dynamic field over Hainan Island, especially northeastern Hainan Island, and concentrating the rainfall over southeastern Hainan Island.

Fig. 3. Average vertical cross section of vertical velocity (shading, units: Pa s-1) and wind anomalies (vector, units:m s-1) between 18°N and 21°N in (a) super and (b) normal EP El Niño years. White dots indicate that the area passes 90% significance test.

Fig. 5. Composite of the abnormal surface LHF (shading, units: W m-1) and wind (vectors,units: m s-1) at 850 hPa in September and October of all super EP El Niño years. White dots indicate that the area passes 90% significance test.

Fig. 6. (a)-(d) Variation of abnormal LHF (shading, units: W m-2) and wind (vector, units: m s-1) at 850 hPa for the four days before non-TC heavy rainfall in super EP El Niño years. (e)-(f) As (a)-(d), but showing the characteristics for non-TC heavy rainfall in super and normal EP El Niño years. Blue and red arrows indicate that the abnormal zonal vertical circulation has been established in the northern SCS. White dots indicate that the area passes 90% significance test.

The positive LHF anomaly near the Beibu Gulf is largely influenced by the increased latent heat transfer which resulted from the increased surface wind speed in autumn non-TC rainfall events (not shown). The impact of wind speed on LHF has been presented in previous research(Li et al., 2012). And the positive LHF anomaly near the Beibu Gulf was stronger due to greater surface wind speed in super EP El Niño years than it was in normal EP El Niño years (Figs. 6e, f). However, it is worth noting that the formation of the positive LHF anomaly predated the appearance of the ACC near the Beibu Gulf. Hence, the increased LHF might play a role in the formation of the ACC. To test this hypothesis, we used an LBM simulation to verify the response of the wind at 850 hPa to the surface heating in the Beibu Gulf, which was consistent with the location of the positive LHF anomaly. Above the idealized diabatic heating (centered at 19.5°N, 108°E with the maximum at 1000 hPa), one ACC appeared near the Beibu Gulf (Fig. 7), which was consistent with the result of the statistical analysis based on the reanalysis data (Fig. 4). Thus, the occurrence of the positive LHF anomaly is another vital element in the formation of the ACC, in addition to the converge of the strengthened southwesterly wind in the northern SCS and the northeasterly wind from China.

5.Role of the ACC in moisture convergence and transportation

Figures 8a and b show the horizontal distribution of the vapor flux and vapor flux divergence anomalies at 925 hPa under different backgrounds. The 925-hPa level was chosen because the maximum moisture was located near 925 hPa,as shown in the profile in Fig. 8c. The vapor flux value across each of the four boundaries of the Hainan Island region was calculated to analyze the source of the moisture and the net vapor flux in Hainan Island.We found that during normal EP El Niño years, a vapor flux anomaly center covered Hainan Island and the net vapor flux reached 0.022 kg m-1s-1hPa-1. However, a robust vapor flux convergence anomaly, which appeared only below 850 hPa, was located only on the south side of Hainan Island and so influenced only southeastern Hainan Island (Figs. 8b, c). In contrast, during super EP El Niño years, an intense center of vapor flux convergence anomaly controlled Hainan Island and was maintained from 1000 hPa to 500 hPa (Figs. 8a, c), resulting in increased rainfall over Hainan Island even though the net vapor flux (0.019 kg m-1s-1hPa-1, Fig. 8a) was less than that during normal EP El Niño years. This seems to indicate that the impact of the vapor flux convergence is more important than the vapor flux itself for the generation of heavy rainfall on Hainan Island. Furthermore, combined with the analysis above, we can conclude that the robust vapor flux convergence anomaly was induced by the ACC near the Beibu Gulf.

Fig. 7. (a) Heat forcing at 1000 hPa (shading; units: K d-1) and the steady response of wind at 850 hPa (vectors; units:m s-1). (b) Heating profile at center point (blue star, 19.5°N, 108°E), which is consistent with the location of the positive LHF anomaly.

Fig. 8. Horizontal distribution of vapor flux (shading, units: kg m-1 s-1 hPa-1) and vapor flux divergence (contours,only shows the area of robust convergence, units: 10-6 kg m-1 s-1 hPa-1) anomalies at 925 hPa in (a) super and (b)normal EP El Niño years, and (c) their profiles. Arrows in (a) and (b) represent vapor flux (kg m-1 s-1 hPa-1) across the four boundaries of Hainan region (18°-21°N, 108°-111.5°E). Green and yellow hollow arrows indicate inflows and outflows, respectively.

Fig. 9. Schematic diagrams of rainfall mechanisms in (a) super and (b) normal EP El Niño years. Grey arrows and orange and purple shading ellipses in (a) indicate the abnormal zonal vertical circulation in the northern SCS, surface heating in the Beibu Gulf, and cooling in the Bashi Strait, respectively. Blue and red circles represent ACC and AAC at 850 hPa, respectively. Blue shading arrows show vaper transport.

In addition, we noted that the sources of moisture entering Hainan Island were distinct under different backgrounds.In super EP El Niño years, the moisture comes mainly from the south, as a result of the southerly wind anomaly on the east rim of the ACC (Figs. 6d and 8a). This leads to increased rainfall over northeastern Hainan Island. However,during normal EP El Niño years, the moisture is transported by the easterly wind anomaly (Figs. 6f and 8b), which results in the moisture converging and being raised on the east side of Wuzhi Mountain, causing rainfall over southeastern Hainan Island (Fig. 2b).

6.Summary and discussion

There is a close relationship between heavy autumn rainfall, especially non-TC heavy rainfall, and the intensity of EP El Niño events. This paper focused on investigating the cause of excessive rainfall in super EP El Niño years.

In the autumn of super EP El Niño years, the cool SST in the northwestern Pacific induces a negative LHF anomaly and excites the Rossby waves to further form an AAC, resulting in the descending motion around the Bashi Strait. Under this background, when a positive surface LHF anomaly and an ACC at 850 hPa occurs near the Beibu Gulf,an abnormal zonal vertical circulation is established in the northern SCS, which further strengthens the ascending motion over Hainan Island and leads to excessive rainfall.Therefore, the formation of the positive LHF anomaly and the ACC are the critical factors for initiating the non-TC rainfall and the abnormal zonal vertical circulation in the northern SCS. Further analysis showed that the positive LHF anomaly, which was generated by the increased latent heat transfer which resulted from the increased surface wind speed, was not only conducive to local vertical upward movement, but also triggered the upper ACC. The ACC was also the supplier of favorable moisture conditions in addition to providing the strong dynamic field. This induced the robust vapor flux convergence, which played a more important role in rainfall over Hainan Island than the vapor flux itself,and transported the moisture from south to north causing the increased rainfall over northeastern Hainan Island (Fig. 9a).

In the autumn of normal EP El Niño years, the ascending motion over Hainan Island cannot be further strengthened because the abnormal zonal vertical circulation in the northern SCS is absent due to the lack of the negative LHF anomaly and the AAC at 850 hPa around the Bashi Strait. This is the reason for the reduction in rainfall in normal autumns,although an ACC would still form near the Beibu Gulf. In addition, the ACC near the Beibu Gulf is located on the south side of Hainan Island, owing to the limitation imposed by an AAC over China, inducing the robust vapor flux convergence that effects only southern Hainan Island.Combined with the effects of the mountainous topography,this scenario ensures that the rainfall is concentrated mainly over southeastern Hainan Island in normal EP El Niño years(Fig. 9b).

In the present study, the ACC near the Beibu Gulf provides the favorable dynamic field and moisture conditions that promote heavy rainfall over northeastern Hainan Island in super EP El Niño years. However, we were unable to fully explain why the rainfall is concentrated over northeastern Hainan Island (Fig. 2a). This restriction may be associated with the local boundary layer processes. We will investigate the mechanism of this phenomenon in a future study.

Acknowledgements. The research is supported by the National Natural Science Foundation of China (Grant No.41975008) and Key Laboratory of South China Sea Meteorological Disaster Prevention and Mitigation of Hainan Province (Grant No.SCSF201906). The authors gratefully acknowledge the China Meteorological Administration and NOAA Physical Sciences Laboratory for providing the tropical cyclone best-track data (tcdata.typhoon.org.cn) and NCEP Reanalysis data (https://psl.noaa.gov/data/gridded/reanalysis/).