As one of the most active components in theglobal climate system, the East Asian winter monsoon(EAWM) is an important climate feature over EastAsia in boreal winter, and one of the most important factors influencing the winter climate in China.EAWM often brings about various kinds of hazards, such as cold surge, low temperature, snow and iceweather, etc.(Ding, 1990; Chen et al., 1991; Huang et al., 2008) . Most previous studies have focused onthe temperature variations in winter and the associated physical mechanisms(Tao, 1957; Li, 1989; Ding, 1990; Wu et al., 1999) . Few studies have concerned about the winter precipitation anomaly features, probably because the regions controlled by the winter mon-soon are often cold and dry, and winter precipitationaccounts for only a small proportion of the annual precipitation(Chen et al., 2008) . On the other h and , win-ter precipitation anomaly may not cause huge floodslike the East Asian summer monsoon precipitationdoes. However, winter precipitation shows an interannual and interdecadal variability over different regionsin China(Xu et al., 1999; Xu and Chan, 2002). Excessively more winter precipitation, in the form of snowor freezing rain, not only brings inconvenience to people's travel, but also causes threat to people's life and property. During some other years, excessively lesswinter precipitation may also induce low soil moisture and drought, and seriously affect the crop production.For example, during mid January and early February 2008, an unprecedented disaster of low temperature, persistent rain, snow and ice storms occurredin the Yangtze River basin and South China, whichkilled at least 120 people and caused direct economiclosses of over 150 billion yuan(Li et al., 2008; WangZunya et al., 2008; Wen et al., 2009). However, duringlate 2008 and early 2009, the severe drought in northern China affected 10 million hectares of crops, and resulted in 4 million people short of drinking water.The direct economic losses were as high as 1.6 billionyuan in only Anhui Province(Gao and Yang, 2009) .Therefore, with the social and economic development, the impact of winter climate on the agriculture, energy, and water balance in China is becoming more and more important. Forecasting of the winter precipitation anomaly has also become an essential workin the meteorological operation(He et al., 2006).

On the interannual timescale, the major mode ofwinter precipitation over China is the consistent variation of rainfall in the area south of the Yangtze Rivervalley(YRV), which accounts for 50% of the total variance and shows a significant interannual variabilitywith 2-4-yr period(Wang and Feng, 2011; Li and Ma, 2012) . This mode is closely related to the intensityof the EAWM and the El Niño-Southern Oscillation(ENSO). Strong EAWM gives rise to a dry conditionin southern China, while weak EAWM results in moreprecipitation there(Zhou et al., 2010; Zhou, 2011) .The interannal variability of winter monsoon also contains some ENSO signals(Mu and Li, 1999) . Previous studies have revealed a significant inverse relationship between EAWM and ENSO. During El Niño winter, the EAWM is weakened, with a weak East Asiantrough(EAT) and low-level southerly anomalies overthe eastern coast of China. During La Niña winter, however, the EAT is deeper than normal, and low-levelnortherly anomalies occur over eastern China, favoring the southward outbreak of strong cold air, and the EAWM is largely intensified(Li, 1988, 1990; Webster and Yang, 1992; Tao and Zhang, 1998; Chen and Graf, 1998; Chen et al., 2000; Lau and Nath, 2000; Chen, 2002) . On the other h and , during the maturephase of El Niño, an anomalous low-level anticycloneappears around the Philippine Sea, which is excitedthrough the Rossby wave response. The anomaloussoutherly winds to the northwest side of the anticyclone not only decrease the EAWM, but also bringmore moisture from the tropical ocean to southernChina(Zhang et al., 1996, 1999; Wang et al., 2000; Wang and Zhang, 2002; Zhang and Sumi, 2002; Wu et al., 2003) . Therefore, during the winter of warm(cold)ENSO, the EAWM tends to be weaker(stronger)thannormal, and the water vapor from the Bay of Bengal and the South China Sea would converge(diverge)insouthern China, favoring more(less)precipitation inthis region(He et al., 2006; Zhou et al., 2010; Wang and Feng, 2011; Li and Ma, 2012) .

Nevertheless, some studies have also proposedthat the climate impact of La Niña shows a significantasymmetry with that of El Niño(Deser and Wallace, 1990; Hoerling et al., 1997; Wu et al., 2010; Wang et al., 2012) . Zhang et al. (1996)found that the effect ofEl Niño on the variations of the East Asian monsoonis more significant than that of La Niña. The recentstudy of Wu et al. (2010)also revealed the asymmetry of the location and amplitude of the western NorthPacific low-level atmospheric circulation anomalies between the El Niño and La Niña mature winters. Wanget al.(2012)investigated changes in the relationshipbetween Meiyu rainfall over East China and La Niñaevents after the late 1970s, and found that the impact of La Niña on the Meiyu rainfall is not reversed to thatof El Niño.

According to previous studies, the EAWM wouldbe stronger than normal during La Niña winter, and the winter precipitation would be less than normal insouthern China. However, in early 2008, the freezingrain and snow disaster over southern China just occurred during the mature phase of La Niña. In thewinters of two recent La Niña events(January 2011 and January 2012), the freezing rain and snow againhappened in southern China(Song, 2012) . Therefore, this paper will examine the characteristics of winterprecipitation anomaly in southern China and exploretheir possible association with warm and cold ENSO.We aim to answer the following three questions: 1)What is the relationship between the winter precipitation anomaly in southern China and ENSO? 2)Isthe impact of La Niña on the winter precipitation insouthern China opposite to that of El Niño? 3)Whydoes the freezing rain and snow in southern China occur more often during La Niña winter in recent years? 2 Data and methods

Multiple datasets are used in this study, including monthly sea surface temperature(SST)data fromthe Hadley Center(1948-2011), monthly atmosphericreanalysis data from the NCEP/NCAR(1948-2011), monthly precipitation anomalies over global l and and oceans from the precipitation reconstruction(PREC)dataset(1948-2011; Chen et al., 2002) , and monthlystation rainfall data at more than 700 stations inChina provided by the China Meteorological Administration(1951-2011). The climatology of the stationprecipitation data is from 1951 to 2011, and that of allthe other gridded data is from 1948 to 2011. In thisstudy, the winter of 2007 refers to December 2007 and January-February 2008.

There are a number of EAWM indices proposedby using different physical variables, such as the sealevel pressure(SLP), the 500-hPa geopotential height, the zonal and meridional winds at lower and upperlevels, and so on. However, different indices may reflect different aspects of the winter monsoon features, and the specific years with strong(weak)EAWM areoften different among these indices(Guo, 1994; Zhu, 2008; Wang and Chen, 2010; Liu et al., 2012) . Therefore, we select several representative EAWM indicesfor comparison, including: 1)the Siberian high index(ISLP-Gong for short; Gong et al., 2001) , defined asthe SLP averaged in the Siberian high region(40°-60°N, 70°-120°E); 2)the east-west l and -sea pressurecontrast index(ISLP-ChL; Chan and Li, 2004), definedas the SLP difference between East Asia(30°-55°N, 100°-120°E) and the northwestern Pacific(30°-55°N, 150°-170°E); 3)the low-level meridional wind indices, including the 10-m meridional wind averaged in southeastern East Asia((10°-25°N, 110°-130°E) and (25°-40°N, 120°-140°E))(I_v10-ChW; Chen et al., 2000) , and the 850-hPa meridional wind averaged in eastern EastAsia(20°-40°N, 100°-140°E)(I_v850-Yang; Yang et al., 2002); 4)the East Asian trough index, defined as the500-hPa geopotential height averaged in(30°-45°N, 125°-145°E)(I_H500-Sun; Sun and Li, 1997) ; and 5)theEast Asian subtropical jet index, defined as the 300 hPa zonal wind shear between(27.5°-37.5°N, 110°-170°E) and (50°-60°N, 80°-140°E)(I_u300-Jhun; Jhun and Lee, 2004) . 3 Characteristics of the winter precipitation in southern China

As revealed by previous studies, the first majormode of the interannual variability of the winter precipitation over China is the consistent variation ofrainfall in the regions south of the YRV, which explainsabout 49.6% of the total variance(Wang and Feng, 2011) . Based on the station rainfall data over 700 stations in China from 1951 to 2011, we can also obtainsimilar results using the empirical orthogonal functionmethod(figure omitted). Therefore, we define a winter precipitation index(WPI)as the normalized winterprecipitation averaged over all stations in six provincessouth of the YRV, i.e., Hunan, Jiangxi, Zhejiang, Fujian, Guangdong, and Guangxi(Fig. 1). Based onone st and ard deviation of WPI(Fig. 2), there are 9yr with abnormally more winter precipitation(1958, 1968, 1982, 1984, 1989, 1991, 1994, 1997, and 2002; Group Ⅰ for short), and 10 yr with abnormally lesswinter precipitation in southern China(1950, 1959, 1962, 1964, 1983, 1985, 1986, 1995, 1998, and 2008; Group Ⅱ for short).

Fig. 1. Distribution of the observation stations in southern China.

Figure options

Fig. 2. Normalized time series of the winter precipitation index(WPI)in southern China(green bar). El/La indicates El Niñno/La Niña event in winter.

Figure options

During the winter of Group Ⅰ, winter precipitationis above normal over most parts of China except mostparts of Northeast China, northeastern Inner Mongolia, and North Xinjiang. Winter precipitation is over50% above normal in the regions south of the YRV(Fig. 3a). During the winter of Group Ⅱ, however, lessprecipitation covers most areas of China, especiallywestern and eastern Northwest China, western NorthChina, and the regions south of the YRV. Winter rainfall anomaly is over 80% below normal in southernSouth China(Fig. 3b). The composite maps basedon the global gridded precipitation data show similar results(Fig. 4). During the winter of Group Ⅰ, precipitation is above normal in the low-mid latitudes of Eurasia, with one maximum center in southern China(Fig. 4a). In the winter of Group Ⅱ, however, precipitation is mostly below normal in the low-mid latitudes of Eurasia, especially in southern China(Fig. 4b).

Fig. 3. Composite anomalous percentage(%)of winter precipitation in China for(a)positive winter precipitationanomaly years and (b)negative winter precipitation anomaly years in southern China(based on 700 station precipitationdata in China).

Figure options

Fig. 4. As in Fig. 3, but for global precipitation anomaly(mm)based on the global gridded precipitation data.

Figure options

The composite results of global precipitation alsoshow some important features in the tropics. During the winter of Group Ⅰ, precipitation is above normal in the tropical central Indian Ocean and equatorial eastern-central Pacific, and below normal in theMaritime Continent and the equatorial western Pacific(Fig. 4a). This feature is similar to the tropicalprecipitation anomaly pattern during El Niño maturewinter(Yuan and Yang, 2012) . During the winter ofGroup Ⅱ, the anomaly pattern is just the opposite, with more winter precipitation in the Maritime Continent and the equatorial western Pacific, and less winter precipitation over the tropical central Indian Ocean and equatorial eastern-central Pacific(Fig. 4b), whichis also similar to the tropical precipitation anomalypattern during La Niña mature winter. The above feature indicates a close relationship between the winterprecipitation in southern China and ENSO. However, both the station rainfall data in China and the globalgridded precipitation data show that during the winter of Group Ⅰ(Group Ⅱ), winter precipitation is above(below)normal over most areas of China. It is different from the anomalous winter precipitation patternwith less rainfall in northern China and more rainfallin southern China(more rainfall in northern China and less rainfall in southern China)during El Niño(La Niña)winter(Gong and Wang, 1998) . Therefore, we further investigate the atmospheric circulation features for Groups Ⅰ and Ⅱ, and explore their relationship with El Niño and La Niña by comparison.

During the winter of Group Ⅰ, the Siberian highis weak, and the east-west pressure contrast is smallerthan normal, with positive SLP anomaly over northwestern Pacific and negative SLP anomaly over Eurasia(Fig. 5a). In the lower troposphere, anomaloussoutherly winds control eastern East Asia and extendfrom the southeastern coast of China to NortheastAsia, indicating a weak EAWM(Fig. 5b). Sincethe anomalous southerly winds are located to thenorthwest of the low-level Philippine Sea anticyclone(PSAC), this feature denotes a close relationship between the low-level wind anomalies for Group Ⅰ and the PSAC. At 500 hPa, positive anomaly of geopotential height dominates in the north and negativeanomaly in the south of the mid-high latitudes ofEurasia. Both the Ural high ridge and the East Asiantrough(EAT)are weaker than normal, also indicatinga weak EAWM. However, positive anomaly of 500-hPageopotential height covers the tropical regions fromthe tropical Indian Ocean to Pacific, indicating an enhanced western Pacific subtropical high(WPSH), withits high ridge shifting more southward and extendingmore westward(Fig. 5c). This feature is favorable formore water vapor transport from the tropic ocean tosouthern China. Therefore, the winter precipitationtends to be above normal in southern China. At 200hPa, easterly anomalies control the regions from central China to southern Japan, while westerly anomalies cover Southeast Asia(Fig. 5d). The East Asiansubtropical jet(EASJ)is thereby weaker than normal, also reflecting a weak EAWM(Yang et al., 2002; Jhun and Lee, 2004) . The above analysis demonstrates thatduring the winter with more precipitation in southern China(Group Ⅰ), the winter monsoon circulation atdifferent levels shows a weak EAWM. Influenced bythe weak winter monsoon, anomalous southerly windscontrol southern China, and more moisture flux converges in that region. Therefore, more winter precipi-tation occurs in southern China, consistent with previous studies(Zhou and Wu, 2010; Zhou, 2011) .

Fig. 5. Composite anomalous atmospheric circulation in winters for(a-d)exceptionally more winter precipitationanomaly years in southern China and (e-h)El Niño years during 1948-2011.(a, e)Sea level pressure(gpm), (b, f)850-hPa wind(m s-1), (c, g)500-hPa geopotential height(gpm), and (d, h)200-hPa zonal wind(m s-1).

Figure options

The atmospheric circulation anomalies for GroupⅡ(years with less winter precipitation in southernChina)are approximately opposite to those of Group Ⅰ.The east-west pressure contrast is larger than normal, with a stronger Siberian high and negative anomalyof SLP over North Pacific(Fig. 6a). Anomalous lowlevel northerly winds prevail over eastern East Asia(Fig. 6b). The 500-hPa geopotential height over Eurasia displays a positive anomaly in the west and a negative anomaly in the east. The Ural high ridge is intensified and the EAT is deeper than normal, indicating astrong EAWM. Meanwhile, the 500-hPa geopotentialheight is near normal over the tropical Indian Ocean and Pacific(Fig. 6c), denoting that the WPSH is nearnormal. At 200 hPa, the EASJ is stronger than normal and moves more southward(Fig. 6d), unfavorable formore precipitation in southern China in winter(Mao et al., 2007) . Therefore, during the winter with lessprecipitation in southern China, the atmospheric circulation over Eurasia shows a strong EAWM at different levels. Influenced by the anomalous cold and dry northerly winds, winter precipitation is less thannormal in southern China.

Fig. 6. As in Fig. 5, but for(a-d)exceptionally less winter precipitation anomaly years and (e-h)La Niña years during 1948-2011.

Figure options

We also calculate the linear correlation betweenWPI and atmospheric circulation anomaly at different levels. Significant negative correlation of SLP isobserved over the Lake Balkhash(Fig. 7a), indicating that when the Siberian high is stronger(weaker)thannormal, winter precipitation tends to be less(more)than normal in southern China. Meanwhile, significant positive correlation of SLP in northwestern Pacific and negative correlation over tropical eastern Pacific(Fig. 7a)may reflect the "seesaw" relationshipof the pressure between the tropical eastern and western Pacific, i.e., the southern oscillation. This featuredenotes a possible connection between ENSO and theSLP anomaly influencing the winter precipitation insouthern China. In the lower troposphere, the meridional wind over East Asia is significantly correlatedwith WPI(Fig. 7b), suggesting that when anomalous southerly(northerly)winds control East Asia, winter precipitation would be more(less)than normal in southern China(Zhang et al., 1996; Wang and Feng, 2011) . Meanwhile, the significant correlationof zonal wind over the tropical ocean also indicatesthe possible connection of the winter precipitation insouthern China with SST over the tropical ocean. Significant negative correlation of 500-hPa geopotentialheight is observed near the Ural Mountains, and significant positive correlation is located over the regionsof the EAT(Fig. 7c). Therefore, when the Ural highridge is weaker(stronger) and the EAT is shallower(deeper), more(less)winter precipitation tends to occur in southern China. The significant correlationof the 500-hPa geopotential height over the tropicalIndian-Pacific Ocean with WPI also suggests that theintensity and position of the WPSH should have significant impacts on the winter precipitation in southern China. In the upper troposphere, the 200-hPazonal wind over the subtropical region in East Asiais significantly negatively correlated with WPI(Fig. 7d), reflecting that the winter precipitation in southern China may be more(less)than normal when thewesterly jet is weakened(intensified). The above correlation results are consistent with our previous composite analysis(Figs. 5 and 6), further confirming thata stronger(weaker)EAWM would result in less(more)winter precipitation in southern China(Zhou and Wu, 2010; Wang and Feng, 2011; Zhou, 2011) .

Fig. 7. Correlation between WPI and anomalous atmospheric circulations.(a)SLP, (b)850-hPa wind, (c)500-hPageopotential height, and (d)200-hPa zonal wind. Shading denotes areas with correlation above the 95% confidence level.

Figure options

Since different winter monsoon indices may reflect different features of the EAWM(Guo, 1994; Zhu, 2008; Liu et al., 2012) , we calculate some representative indices for the composites of Group Ⅰ and GroupⅡ. All indices show a consistent weak(strong)EAWMcirculation pattern during more(less)winter precipitation years in southern China: the Siberian highis weakened(enhanced), the east-west l and -sea pressure contrast is smaller(larger), anomalous southerly(northerly)winds prevail over East Asia, the EAT isshallower(deeper)than normal, and the East Asiansubtropical jet is weakened(strengthened)(Table 1).

Table 1. Values of various East Asian winter monsoon indices composite for years of exceptionally more(Group Ⅰ) and less(Group Ⅱ)winter precipitation anomalies in southern China

Table options

Analyses in the previous section have revealedsome ENSO signals in the atmospheric circulation anomalies associated with winter precipitationanomaly in southern China, such as SLP in the tropical oceans and low-level winds over East Asia. Inthis section, we further explore the relationship between warm/cold ENSO and the winter precipitationanomaly in southern China. Considering the decadalchange of the SST anomaly(SSTA)distribution during ENSO mature phase in recent 10 years(Ashok et al., 2007; Kao and Yu, 2009; Kug et al., 2009; Yeh et al., 2009; Lee and McPhaden, 2010) , we select all ElNiño and La Niña years from 1948 to 2011 based onone st and ard deviation of Niño3, Niño4, and Niño3.4indices in winter. There are 14 El Niño years: 1957, 1965, 1968, 1972, 1982, 1986, 1987, 1990, 1991, 1994, 1997, 2002, 2004, and 2009; and 14 La Niña years:1949, 1950, 1955, 1967, 1970, 1973, 1975, 1984, 1988, 1998, 1999, 2007, 2010, and 2011. When these ENSOyears are marked on the WPI in Fig. 2, we obtain someimplications: 1)during most El Niño winters(71.4% ofall El Niño years), winter precipitation is above normalin southern China, and during most La Niña winters(69.2% for all La Niña years), winter precipitation isbelow normal in southern China; 2)some winters withmore precipitation in southern China are not El Niñoyears, such as 1958 and 1989; similarly, some winterswith less precipitation in southern China do not occur in La Niña years, such as 1962, 1964, 1995 and 2008; 3)before 1980, the probability of El Niño winters with more precipitation in southern China is 75%, which changes a little to 70% after 1980; however, theprobability of La Niña winters with less precipitationin southern China is 100% before 1980 and largely reduces to only 42.9% after 1980.

The composite atmospheric circulation anomaliesduring El Niño and La Niña years are also shown inFigs. 5 and 6, so that we can compare their featureswith the composite results for Groups Ⅰ and Ⅱ, respectively. It is shown that the composite circulationanomalies for El Niño years are similar to those during the winters with more precipitation in southernChina(Group Ⅰ). The Aleutian low is intensified and the east-west l and -sea pressure contrast is smaller thannormal(Fig. 5e), similar to Fig. 5a. An anomalous low-level anticyclone appears around the Philippine Sea and anomalous southerly winds control theeastern coast of China. Meanwhile, anomalous easterlies are observed over the equatorial Indian Ocean, and anomalous westerlies prevail over the equatorialcentral-western Pacific(Fig. 5f), also similar to Fig. 5b. The 500-hPa geopotential height over Eurasia displays an anomalous "negative in the north and positivein the south" pattern, with a weak Ural high ridge and a shallower EAT. At the same time, significant positiveanomaly of geopotential height covers the tropical region, indicating an enhanced WPSH with its high ridgeextending more westward and southward(Fig. 5g).During El Niño winters, features of positive anomaliesof the 500-hPa geopotential height over the tropicalregion are more significant(Fig. 5g). However, during the winters with more precipitation in southern China, more significant features of the geopotential heightanomalies appear over the mid-high latitudes of Eurasia(Fig. 5c). At 200 hPa, the East Asian subtropicaljet is weaker than normal(Fig. 5h; Zhang et al., 1996) , also similar to Fig. 5d. Larger similarities between theright panel and the left panel confirm that El Niño isan important external forcing factor resulting in morewinter precipitation in southern China(Li, 1988, 1990; Tao and Zhang, 1998; Wu et al., 2003; Zhou and Wu, 2010; Zhou et al., 2010) .

However, the composite atmospheric circulationsfor La Niña winters show more differences with GroupⅡ(winters with less precipitation in southern China).During La Niña winters, the Aleutian low is weakerthan normal. However, few signals of SLP anomalycan be discerned around the Siberia and for the eastwest l and -sea pressure contrast(Fig. 6e), differentfrom the intensified Siberian high and larger east-westpressure contrast for Group Ⅱ(Fig. 6a). At 850 hPa, an anomalous cyclone is located around the Philippines, and anomalous easterlies prevail over southernChina. Meanwhile, westerly anomalies cover the equatorial Indian Ocean, and easterly anomalies dominateover the equatorial western Pacific(Fig. 6f). All thesefeatures are different from those for Group Ⅱ(Fig. 6b).The 500-hPa geopotential height over Eurasia showsan anomalous "positive in the north and negative inthe south" pattern, indicating a strong EAWM. Negative anomaly of 500-hPa geopotential height covers thetropical Indian-Pacific Ocean, denoting a weak WPSHwith its high ridge retreating more eastward(Fig. 6g).These features are also different from Fig. 6c. At 200hPa, westerly anomalies cover from northern India tosouthern Japan during La Niña winters, indicating anintensified East Asian subtropical jet(Fig. 6h). Although the subtropical jet is also enhanced for GroupⅡ, the Middle East jet is weakened(Fig. 6d), differentfrom that in Fig. 6h.

Significant positive correlations of SSTA withWPI are located over the tropical Indian Ocean and the equatorial eastern-central Pacific(Fig. 8a), denoting a close relationship between ENSO and winterprecipitation in southern China. However, the composite SSTA for Groups Ⅰ and Ⅱ shows significant differences compared with the correlation results. Duringthe winter of Group Ⅰ, an El Niño event occurs over theequatorial eastern-central Pacific. Meanwhile, SSTAis above normal in the tropical Indian Ocean and below normal over the equatorial western Pacific(Fig. 8b). These features are consistent with the SSTA distribution pattern during El Niño winter(figure omitted). However, during the winter of Group Ⅱ, exceptweak negative SSTA dominating over the equatorialeastern-central Pacific, SST over other oceans showsinsignificant anomaly features(Fig. 8c).

Fig. 8.(a)Significant correlation(above the 95% confidence level)between WPI and the global SSTA, and SSTAcomposite(℃)for(b)more winter precipitation(Group Ⅰ) and (c)less winter precipitation(Group Ⅱ)years in southern China.

Figure options

The above analysis not only confirms the close relationship between more winter precipitation in southern China and El Niño events, but also suggests that the relationship between less winter precipitation insouthern China and La Niña events is not significant.The impact of La Niña on the winter precipitation insouthern China should not be opposite to that of ElNiño. Therefore, in the following section, we furtherinvestigate the impact of La Niña on the winter precipitation in southern China and the associated physicalmechanisms. 5 Decadal variability of the La Niña impact onwinter precipitation in southern China

It is revealed in previous sections that the probability of less winter precipitation in southern Chinaduring La Niña years changes from 100% before 1980to 42.9% after 1980, suggesting that La Niña yearswith more winter precipitation in southern China takea larger probability(57.1%)after 1980. Then, doesthe impact of La Niña on the winter precipitation insouthern China show a decadal variability?

Figure 9 shows the composite anomalous percentage of winter precipitation over 700 stations in Chinaduring El Niño and La Niña years. Few differences canbe observed for El Niño years before and after 1980.They both show less precipitation over central China and more precipitation over eastern China and the regions south of the YRV(Figs. 9a and 9b). However, the composite precipitation for La Niña years exhibitssignificant differences before and after 1980. DuringLa Niña winters before 1980, less precipitation occursover most China, with minimum centers over westernChina and the regions from North China to the middle and lower reaches of the YRV(Fig. 9c). DuringLa Niña winters after 1980, however, more precipitation dominates over some regions of China, especiallyin eastern Northwest China and southern China(Fig. 9d). We have also compared the temperature anomalyover China for La Niña years before and after 1980.During La Niña winters before 1980, temperature isbelow normal over most China. After 1980, however, except Northwest China and southern China with lowtemperature, winter temperature is above normal inother regions(figure omitted). This feature probablyindicates the variation of La Niña impact on the winter temperature in China influenced by the global warming background(not discussed in this paper). Moreimportantly, since the temperature is still below normal in southern China during La Niña winter after1980 and more winter precipitation tends to occur insouthern China, this feature results in somewhat cold and wet winter in southern China during La Niña winters after 1980, but not cold and dry winter before1980. Therefore, freezing rain and snow/ice stormsbecome more likely to occur in southern China duringLa Niña winters after 1980.

Fig. 9. Composites of anomalous percentage(%)of winter precipitation in China for El Niño and La Niña years.(a)El Niño years before 1980, (b)El Niño years after 1980, (c)La Niña years before 1980, and (d)La Niña years after 1980.

Figure options

We further compare the atmospheric circulationanomalies at different levels for La Niña years before and after 1980, respectively. During La Niña wintersbefore 1980, the SLP anomaly over Eurasia shows a"positive in the north and negative in the south" pattern(Fig. 10a). After 1980, however, positive anomalyof SLP dominates over the Siberia, indicating an intensified Siberian high(Fig. 10f). Before 1980, theanomalous low-level cyclone excited by La Niña is located to the east of the Philippines, causing anomalous northerly winds over southern China(Fig. 10b).After 1980, however, the anomalous low-level cycloneinduced by La Niña moves to the west of the Philippines. Therefore, anomalous easterly winds dominateover southern China(Fig. 10g), not conducive to thestrengthening of the winter monsoon, but favorablefor more moisture flux transport from northwesternPacific to southern China.

Fig. 10. Composites of the anomalous atmospheric circulation in winter for La Niña years(a-e)before 1980 and (f-j)after 1980.(a, f)SLP(gpm), (b, g)850-hPa wind(m s-1), (c, h)500-hPa geopotential height(gpm), (d, i)200-hPazonal wind(m s-1), and (e, j)integrated moisture flux from 1000 to 300 hPa(vector; kg s-1 m-1)as well as convergence(blue shading; 10-5 kg s-1 m-2) and divergence(yellow shading; 10-5 kg s-1 m-2).

Figure options

At 500 hPa, the geopotential height in Eurasiashows an anomalous "positive in the north and negative in the south" pattern, reflecting a strong EAWMduring La Niña winter before 1980. Meanwhile, negative anomaly of 500-hPa geopotential height coversthe tropical and subtropical regions of Eurasia(Fig. 10c). This feature indicates a weakened WPSH withits high ridge retreating more eastward, unfavorablefor the water vapor transport from the tropical oceansto southern China. During La Niña winters after 1980, the 500-hPa geopotential height shows a weak "positive in the north and negative in the south" patternover Eurasia. However, positive anomaly of the geopotential height over eastern East Asia suggests a shallower EAT, which is unfavorable for the southward invasion of cold air to southern China. Meanwhile, the500-hPa geopotential height is relatively below normal over the Bay of Bengal(Fig. 10h). Therefore, the India-Burma trough would be intensified, favoringmore moisture flux transport from the Indian Oceanto southern China.

In the upper troposphere, the East Asian subtropical jet is strengthened during La Niña winter bothbefore and after 1980. However, the strengthened jetis located over southern China before 1980(Fig. 10d), unfavorable for more winter precipitation there. During La Niña winters after 1980, however, the westerlyjet moves more northward and southern China is located under the right side of the jet entrance area(Fig. 10i). Therefore, more winter precipitation would occur in southern China through the secondary verticalcirculation(Mao et al., 2007) .

During La Niña winters before 1980, anomalousdivergence of moisture flux controls southern China(Fig. 10e). After 1980, however, anomalous convergence of moisture flux dominates over southern China.Intensified moisture flux comes from the tropical Indian Ocean and western Pacific(Fig. 10j), with theformer caused by the enhanced India-Burma trough and the latter induced by the anomalous easterlies tothe north side of the anomalous low-level cyclone.

The above analyses confirm that the impact ofLa Niña on the winter precipitation in southern Chinaexhibits a significant decadal variability. During a LaNiña winter before 1980, the East Asian subtropicaljet is stronger than normal and the jet is located moresouthward. The EAT is deeper than normal and theWPSH is weakened with its high ridge retreating moreeastward. The anomalous low-level cyclone excited byLa Niña is located to the east of the Philippines, inducing anomalous northerly winds over southern China.Therefore, the intensified EAWM as well as the poormoisture flux condition would cause low temperature and less winter precipitation in southern China, i.e., a"cold and dry" winter. During a La Niña winter after1980, however, the East Asian subtropical jet movesnorthward. The EAT is shallower than normal, unfavorable for the southward invasion of the winter monsoon. On the other h and , the enhanced India-Burma trough leads to more moisture flux from the tropicalIndian Ocean. Meanwhile, the anomalous low-levelcyclone excited by La Niña is located to the west ofthe Philippines, which also favors more moisture fluxfrom the western Pacific to southern China. As aresult, unlike the "cold and dry" winter in southernChina during La Niña winter before 1980, an anomalous "cold and wet" winter occurs in southern Chinaduring La Niña winter after 1980. Therefore, the freezing rain and snow/ice storms are more likely to happenin southern China after 1980.

Considering the different external forcing patternsbetween El Niño and La Niña episodes, the physicalmechanism for more winter precipitation in southernChina during La Niña years after 1980 should not bethe same as that during El Niño years. Previous studies have clearly revealed the physical mechanism of theEl Niño impact on the increased winter precipitation insouthern China(Zhang et al., 1999; Chang et al., 2000; Wang Bin et al., 2008; Yuan and Yang, 2012) . Influenced by the warm SSTA over the equatorial easterncentral Pacific during El Niño winter, anomalous risingmotion dominates over the equatorial eastern-centralPacific and anomalous subsidence prevails over theequatorial western Pacific. This subsidence also excites anomalous upward motion over southern Chinathrough the regional Hadley circulation in East Asia(Fig. 11a; Zhang et al., 1996; Yuan and Yang, 2012) .On the other h and , El Niño also stimulates an anomalous low-level anticyclone around the Philippines(PSAC), which causes anomalous southerly winds oversouthern China. This feature not only decreases theEAWM, but also favors more moisture flux transportfrom western Pacific to southern China(Zhang et al., 1996, 1999; Wang et al., 2000; Zhou and Wu, 2010; Zhou et al., 2010) . Therefore, El Niño would result in more winter precipitation in southern Chinamainly through influencing the regional Hadley circulation over East Asia and exciting the PSAC(Fig. 12a). During La Niña winter, the cold SSTA over theequatorial central Pacific and warm SSTA over theequatorial western Pacific would intensify the Walker circulation. According to our analysis, La Niña eventsafter 1980 would increase the East Asian subtropicaljet and cause the jet moving more northward. Sincesouthern China is located under the right side of the jetentrance region, the anomalous subsidence to the leftside of the jet(North China)would induce anomalousrising motion in southern China(Fig. 11b). On theother h and , La Niña after 1980 would also increase theIndia-Burma trough and excite the low-level Philippine Sea cyclone(PSC)to the west of the Philippines.More moisture flux transports from the tropical IndianOcean and western Pacific to southern China, favoringmore winter precipitation there. Therefore, La Niñaafter 1980 may also induce more winter precipitationin southern China mainly through influencing the intensity and position of the EASJ, the India-Burmatrough, and the position of the PSC(Fig. 12b). Itcan be concluded that under the global warming background, both El Niño and La Niña would result inmore winter precipitation in southern China. However, the physical mechanisms for the anomalous rising motion over southern China as well as the moisturetransport are significantly different between El Niño and La Niña.

Fig. 11. Composite of the anomalous 500-hPa vertical velocity in winter for(a)El Niño years and (b)La Niña yearsafter 1980. The vertical velocity has been multiplied by 100. Blue(yellow)shading indicates anomalous descending(ascending)motion(0.01 Pa s-1).

Figure options

Fig. 12. Sketch map for the physical mechanism of the(a)El Niño and (b)La Niña impact on the winter precipitationin southern China after 1980.

Figure options

It is also revealed from the above analysis thatthere are three differences in the atmospheric circulation anomaly during La Niña winter before and after 1980: 1)the location of the PSC, 2)the 500hPa geopotential height anomaly in East and SouthAsia, and 3)the position of the EASJ. SSTA composite shows that La Niña after 1980(Fig. 13b)isstronger than that before 1980(Fig. 13a). Meanwhile, the negative SSTA center during La Niña after1980 moves westward to near the date line, and positive SSTA appears over southeastern and northwesternPacific(Fig. 13b), showing some central-Pacific(CP)type features after 1980(Yuan and Yan, 2013) . Somerecent research has found that the anomalous low-levelPhilippine Sea anticyclone excited by CP El Niño islocated to the west of the Philippines, unlike the anticyclone induced by eastern-Pacific(EP)El Niño tothe east of the Philippines(Feng and Li, 2011; Yuan and Yang, 2012; Yuan et al., 2012) . Similar to this feature, the response of low-level atmospheric circulationto different types of La Niña should also be different.Because of the higher-frequency occurrence of CP LaNiña after 1980(Yuan and Yan, 2013) , the anomalouslow-level Philippine Sea cyclone also moves westwardto the west of the Philippines(Fig. 10g). Therefore, the change of the cyclone's position should be relatedto the decadal variation of the SSTA distribution ofLa Niña during its mature phase.

Fig. 13. SSTA(℃)composite in winter for La Niña year(a)before 1980 and (b)after 1980.

Figure options

On the other h and , both SST and atmosphericcirculation show a decadal variability after 1980. After 1980, significant warming appears over the tropicalIndian Ocean and Northwest Pacific(Fig. 14a), whichresults in a weak warm condition in the tropical Indian Ocean during La Niña mature phase(Fig. 13b), not a cold condition as revealed by previous studies(Klein et al., 1999; Lau and Nath, 2003; Yang et al., 2007) . Coupled with the warming conditions over thetropical Indian-western Pacific Ocean, westerly windsincrease over the equatorial Indian Ocean and easterlies also strengthen over the western Pacific, whichlead to enhanced convergence and a cyclone aroundsouthern South China Sea(Fig. 14b). Therefore, decadal warming over the tropical Indian-western Pacific Ocean not only favors the westward movement ofthe PSC excited by La Niña, but also increases themoisture condition for more precipitation in southernChina. At 500 hPa, the geopotential height has significantly intensified over the tropical and subtropicalregions of Eurasia(including the region of the EAT)after 1980. However, the geopotential height over theUral decreases to some extent(Fig. 14c). These features may reflect the decadal decrease of the EAWMafter 1980(Zhu, 2008; Wang and Chen, 2010) . At200 hPa, westerly winds cover the regions from central Asia to southern Japan in the difference map(Fig. 14d), indicating a significant decadal increase of theEASJ. Therefore, besides the decadal variation of theSSTA distribution of La Niña, the decadal variationof the atmospheric circulation in the Northern Hemisphere should also be an important reason for morewinter precipitation in southern China during La Niñawinter after 1980.

Fig. 14. Decadal variability(average of 1980-2011 minus average of 1948-1979)of(a)SST(℃), (b)850-hPa wind(ms-1), (c)500-hPa geopotential height(gpm), and (d)200-hPa wind(m s-1)in winter.

Figure options

Using atmospheric circulation data from NCEP/NCAR, SST data from the Hadley Center, stationrainfall data over 700 stations in China, and globalgridded precipitation data, this paper has investigatedthe EAWM and the atmospheric circulation anomaliesduring exceptionally more and less winter precipitation years in southern China, and further exploredtheir asymmetric relations with El Niño and La Niña.We have basically answered the three questions proposed in the introduction section: 1)What is therelationship between the winter precipitation anomalyin southern China and ENSO events? During morewinter precipitation years in southern China, thereis usually a significant El Niño event maturing inthe equatorial eastern-central Pacific. During the ElNiño mature phase, more winter precipitation tendsto occur in southern China. The composite atmospheric circulation patterns show much similarity formore winter precipitation years in southern China and for El Niño winters, which confirms that El Niño isan important external forcing factor for more winter precipitation in southern China. However, duringless winter precipitation years in southern China, fewLa Niña signals can be discerned over the equatorialPacific. 2)Is the impact of La Niña on the winter precipitation in southern China opposite to thatof El Niño? The answer is no. El Niño would result in more winter precipitation in southern China.However, the impact of La Niña on the winter precipitation in southern China shows a decadal variability.During La Niña winter before 1980, the EAWM issignificantly stronger than normal, and the WPSH isweakened with its high ridge retreating more eastward.Thereby, anomalous northerly winds control southernChina and cause a "cold and dry" winter there. However, during La Niña winter after 1980, the intensityof EAWM is largely reduced as compared with thecondition before 1980, whereas the tropical moistureflux increases significantly. As a result, more winterprecipitation occurs in southern China, which inducesa "cold and wet" winter there. 3)Why does the freezing rain and snow in southern China occur more oftenduring La Niña winter in recent years? During LaNiña winter after 1980, the EASJ is intensified and moves northward. Since southern China is located tothe right side of the jet entrance, anomalous risingmotion dominates over southern China through thesecondary vertical circulation. Meanwhile, the EastAsian trough is weaker than normal, unfavorable forthe southward invasion of the cold air. The intensified India-Burma trough favors more moisture fluxtransport from the tropical Indian Ocean to southernChina. The anomalous low-level cyclone excited by LaNiña is located to the west of the Philippines, causinganomalous easterlies over southern China and favoringmore moisture from northwestern Pacific to southernChina. Therefore, during La Niña winter after 1980, the freezing and snow storm disasters are more likelyto occur in southern China. We have also analyzedsome La Niña cases with more winter precipitation insouthern China after 1980(such as 1984, 1988, 2007, and 2011), and found that the winter monsoon circulation features in each case are similar to our compositeresults. Our further analyses have pointed out thatthe high-frequency occurrence of CP La Niña after1980 causes the low-level anomalous cyclone to moveto the west of the Philippines.

On the other h and , the atmospheric circulationsover the tropical and subtropical Eurasia show significant decadal variability after 1980. We have alsochecked the possible impact of the decadal variabilityof the atmospheric circulation background throughchanging the climatology. Based on the climatology of 1980-2011, the anomaly features of both thestation precipitation in China and the atmosphericcirculation over East Asia show evident differences ascompared with those based on the climatology from1948 to 2011(figures omitted). It is therefore inferredthat the decadal variability of the climate backgroundshould have some effects on the decadal variation ofLa Niña impact on the winter precipitation in southern China. To conclude, the change of the SSTAdistribution type of La Niña during its mature phase, as well as the decadal variability of the atmosphericcirculation in the Northern Hemisphere are the important reasons for the decadal variation of the La Niñaimpact on the East Asian climate after 1980.

It should be noted that the decadal change of theLa Niña impact on the tropical atmosphere, especiallyon the anomalous low-level cyclone around the Philippines and the tropical moisture transport, may bemore direct and significant. The atmospheric circulation anomaly over the mid-high latitudes of Eurasia inwinter may be influenced by some other factors, suchas the Arctic Oscillation(Gong et al., 2001) , the NorthAtlantic Oscillation(Wu and Huang, 1999), the Arcticsea ice(Wu et al., 2011), the snow cover over Eurasia(Zuo et al., 2011) , and so on. Further investigationsare still needed to explore the circulation anomaliesover the mid-high latitudes of Eurasia associated withwinter precipitation anomalies in southern China.

Our analysis reveals that winter precipitation isabove normal over most parts of China during morewinter precipitation years in southern China. However, more precipitation over southern China mainlyoccurs in the next January and February. It is alsorevealed that the "cold and wet" winter in southernChina is only significant in January during La Niñawinter after 1980, and the atmospheric circulation inthe decaying January of La Niña is most similar tothat in the whole winter of La Niña. In recent years, the freezing rain and snow storm disasters also happenmore in the decaying January of La Niña, like January2008, January 2011, and January 2012. Therefore, theabove features are probably related to the subseasonalvariation of the EAWM and the low-frequency activities of atmospheric circulations(Bueh et al., 2008; Wang Yun et al., 2008; Shao et al., 2011) .

Some previous studies have proposed the asymmetry of El Niño and La Niña impact on the climate.For example, Zhang et al. (1996)found that the EastAsian monsoon shows significant anomaly during ElNiño mature phase, while exhibits negligible anomalyduring La Niña mature phase. Recently, Wu et al. (2010)revealed that the response of the atmosphericcirculation anomaly over the northwestern Pacific toLa Niña events is asymmetric to that of El Niño. Neither the intensity nor the position of the anomalouslow-level Philippine Sea cyclone excited by La Niñais consistent with those of the anticyclone excited byEl Niño. Wang et al. (2012)also suggested thatLa Niña events after 1970 may induce more Meiyuin the YRV, not opposite to the impact of El Niño.It is also revealed from our analysis that the impactof La Niña on the winter precipitation over southern China shows evident asymmetric features withthat of El Niño. Considering the complication of LaNiña impact on the climate under the global warmingbackground, the impact deserves more studies in nearfuture.

Acknowledgments: We thank the two anonymous reviewers. Their valuable comments and suggestions have greatly improved the quality of thismanuscript.