Large-scale Circulation Anomalies Associated with Interannual Variation in Monthly Rainfall over South China from May to August

One of the most notable features of South China (SC) rainfall during the rainy season is the seasonal evolution, and the stepwise evolution of SC rainfall in the rainy season is closely related to the seasonal evolution of large-scale circulations over East Asia and the western North Pacific (WNP). From April to mid-May, SC rainfall is mainly characterized by frontal rainfall, affected by the convergence of cold air from the midlatitudes and southwest air along the west flank of the western North Pacific subtropical high (WNPSH) (e.g., Qin et al., 1994). From late May to June, after the summer monsoon onset over the South China Sea (SCS), the monsoonal rainband advances up to SC, and SC rainfall mainly results from southwesterlies, which transport water vapor into SC (Chen et al., 2000; Zheng et al., 2006). From July to September, the monsoonal rainfall becomes relatively weakened, and tropical cyclones contribute appreciably to the rainfall over SC, with the ratio being around 10% of tropical cyclone-induced rainfall to total rainfall (Ren et al., 2002; Lee et al., 2010).

The atmospheric circulation anomalies associated with SC rainfall variability have been revealed in previous studies (e.g., Yang and Sun, 2005; Yang, 2011; Chen et al., 2012). It was found that the zonal displacement of the WNPSH is closely related to a SC rainfall anomaly in early summer. During the period of the so-called pre-rainy season, which persists from late April to mid-June, SC rainfall is above-normal (below-normal) when the WNPSH shifts westwards (eastwards) (Yang and Sun, 2005). The southwesterlies over the coastal region of SC, whose intensity is determined by the location and intensity of the WNPSH, have been emphasized to be a crucial factor affecting SC rainfall in June. When the southwesterlies are strengthened, SC rainfall is reinforced (Yang, 2011; Chen et al., 2012). (Mao et al., 2011) examined the general circulation anomalies associated with the interannual variations in bimonthly (May and June) SC rainfall, and found that there is a lower-tropospheric anticyclonic anomaly over the SCS and the Philippine Sea. They indicated that this anticyclonic anomaly is stronger before the end of the 1970s but greatly weaker afterward.

The atmospheric circulation in East Asia and the WNP exhibits remarkable seasonal evolution in summer (e.g., Wu and Wang, 2001; Zhao et al., 2007; Su and Xue, 2010). In mid-May, the WNPSH advances northeastward and the monsoon trough establishes over the eastern SCS (e.g., Wu and Wang, 2001).The SCS monsoon commences and a planetary-scale monsoon rainband is established (e.g., Qian and Lee, 2000). In mid-June, the WNPSH shifts northward and the monsoon trough migrates northeastward. Meanwhile, closely associated with the change of circulation, the intensive convection over the SCS extends into the southwestern Philippine Sea (Wu and Wang, 2001) and the planetary-scale monsoon rainband advances northward into the mei-yu/baiu region. In late July, the WNPSH retreats eastward and marches further northward (e.g., Su and Xue, 2010), and the monsoon trough is strengthened and extends northeastward (Wu and Wang, 2001). At the same time, the East Asian upper-tropospheric westerly jet advances northward (Lin and Lu, 2008), and its core shifts westward (Zhang et al., 2006).

It may be inferred from these seasonal evolutions of rainfall and circulation that it is necessary to investigate the relationship between interannnual variations of circulation and SC rainfall on the subseasonal timescale. (Ye and Lu, 2011) have found that since the late 1970s, there has been a subseasonal difference in ENSO-related East Asian rainfall anomalies between early summer and late summer, and a decaying El Niño corresponding to more SC rainfall in early summer but less rainfall in late summer. Therefore, summer mean (June, July, and August in their study) rainfall in SC exhibits a very weak anomaly in the ENSO decaying phase.

However, previous studies on interannual variability in SC rainfall have seldom taken subseasonal variation into consideration. Most previous studies have focused on the early summer rainfall (e.g., Chen et al., 2004; Ma et al., 2009, among many others), but late summer rainfall has been ignored. What is the relationship between early and late summer SC rainfall? What kinds of circulation anomalies are associated with interannual variation in late summer SC rainfall? Are these circulation anomalies similar to those associated with early summer SC rainfall variability? To answer these questions, the present study examines interannual variations of monthly SC rainfall and the circulation anomalies corresponding to monthly SC rainfall variability. The rest of this paper is organized as follows. In section 2, the datasets used in the present study are presented. In section 3, the main features of interannual variation in monthly SC rainfall during summer are shown. In section 4, the general circulation anomalies associated with SC rainfall anomalies are revealed. And finally, section 5 provides a summary.

In the present study, two precipitation datasets are used: one is monthly data from 160 stations provided by the Chinese Meteorological Data Center, and the other is monthly precipitation from the Global Precipitation Climatology Project (GPCP). SC is specified by the region east of 105^°E on mainland China between the latitudes 21^°N and 28^°N. There are 29 stations distributed evenly over this region (Fig.1), at which precipitation is averaged to represent SC rainfall. This region includes Guangdong Province, Guangxi Province, Fujian Province, and southern parts of Jiangxi Province, Hunan Province, and Guizhou Province.

The 40-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis (ERA-40) dataset (Uppala et al., 2005) is also used; specifically, the physical variables of horizontal winds and geopotential heights. In addition, the improved extended reconstructed global SST (ERSST) dataset (Smith and Reynolds, 2003) is used. The time periods of data used in this study are 1954-2006, 1979-2002, 1958-2002 and 1957-2003 for station rainfall data, GPCP data, ERA-40 data and SST data, respectively.

Figure 1.  Distribution of the stations used in this study.

Figure 2 shows monthly SC precipitation and its standard deviation. There is an obvious annual cycle, characterized by a peak in June, increasing rapidly from January to June, and decreasing quickly from August to December. The standard deviation shares a similar annual cycle, with the maximum in June also. Interestingly, the precipitation in February is small, but with great variability. Our result is consistent with previous results, despite slight differences in the specified regions used in these studies (e.g., Ma et al., 2009; Chen et al., 2012).

Figure 2.  The precipitation amount (solid line) and interannual standard deviation (dashed line) of monthly rainfall over SC. Units: mm month^-1.

In the present study, we investigate monthly SC rainfall in May, June, July and August, and the associated circulation anomalies. Monthly SC rainfall is richest in these four months (Fig. 2), with the amounts of rainfall being 219 mm (May), 248 mm (June), 178 mm (July) and 184 mm (August). The total amount of these four months accounts for more than half (or more exactly, 55.7%) of annual precipitation, indicating the clear existence of a rainy season in SC. The standard deviations are also the greatest in these four months, being about 60 mm. These standard deviations are 1/4-1/3 of monthly precipitation: 24.6% in May, 25.1% in June, 30.3% in July, and 28.2% in August.

Figure 3 shows the time series of SC rainfall in these four months. The interdecadal variation is appreciable in each set of monthly rainfall results. SC rainfall in May experienced a dry phase in the 1960s, and a wet phase in the 1970s, while SC rainfall in June experienced a wet phase in the 1960s, a dry phase in the 1980s, and a wet phase in the 1990s. Interdecadal variations of SC rainfall in July and August are roughly in phase. The rainfalls in both months experienced a dry phase prior to the early 1990s, and turned to a wet phase afterward. It has been revealed that SC summer (June-July-August, JJA) rainfall experienced a decadal change in the early 1990s (Kwon et al., 2007; Wu et al., 2010). Figure 3 indicates that such a decadal change can also be detected in the rainfall variations in each month of JJA.

Figure 3.  Time series of rainfall anomalies over SC: (a) May; (b) June; (c) July; (d) August. The bars represent original rainfall anomalies, and the curves represent the decadal change obtained by the 9-yr running mean. Units: mm month^-1.

We focus on the interannual variations of SC rainfall in the present study, which are obtained by removing the interdecadal variation (9-yr running average) from the original rainfall time series. The ratios of interannual variance to total variance are 83.3% (May), 77.6% (June), 90.9% (July) and 95.9% (August), indicating that the interannual components are much stronger than the interdecadal components. Figure 4 shows the standardized interannual variation of monthly SC rainfall. There are many years of monthly rainfall below minus-one standard deviation or above one standard deviation, ranging from 17 to 21 years and accounting for almost half of the total number of years analyzed. This result is consistent with the strong interannual standard deviation of SC rainfall in each month from May to August.

Figure 4.  Standardized interannual variations of rainfall over SC: (a) May; (b) June; (c) July; (d) August.

Figure 5 shows the correlation coefficients of SC rainfall with rainfall at each station. There is a significant positive correlation over SC. Among 29 stations in SC (Fig.1), the numbers of significantly correlated stations are 25 (May), 24 (June), 26 (July) and 23 (August). In addition, SC rainfall is negatively related to the rainfall north of the Yangtze River, but this relationship is not clearly significant and there are only a few stations with significant correlation coefficients.

Figure 5.  Correlation of monthly SC rainfall with rainfall at 160 stations: (a) May; (b) June; (c) July; (d) August. The contour interval is 0.1 and the contour line of zero is omitted. The shading and marked stations show correlations significant at the 95% confidence level by the Student's t-test. Solid and hollow circles represent stations positively and negatively correlated to SC rainfall, respectively.

SC rainfall is basically independent among months. The absolute values of correlation coefficients are almost less than 0.2, except for the one between rainfall in May and July, which is -0.41 (Table 1). This significant relationship between May and July is likely to be related to the ENSO cycle. Figure 6 shows the lead-lag correlation coefficients between monthly SC rainfall from May to August and the Niño3.4 index. The Niño3.4 index is defined as the SST anomalies averaged over the region (5^°S-5^°N, 170^°-120^°W). It indicates that SC rainfall in May and July is related to the developing phase of ENSO events: a positive May (July) rainfall anomaly tends to occur during the developing phase of La Niña (El Niño) events. The correlation coefficients between SC rainfall and the following winter (November-December-January) mean Niño3.4 index are -0.34 for May and 0.38 for July. Therefore, May and July rainfall anomalies tend to show different signs during developing-ENSO years, contributing to the negative correlation coefficient between rainfall in these two months. Though this May-July relationship is statistically significant, there is only a small portion (16.8%) of variance in common between May and July rainfall, implying that monthly SC rainfall is roughly independent.

Figure 6.  Lead-lag correlation of monthly Niño3.4 index with SC rainfalls. The vertical bars show the simultaneous correlation. The solid horizontal lines indicate the 95% confidence level by the Student's t-test.

Figure 7 shows the regressed rainfall anomalies of May, June, July and August in the GPCP dataset. Roughly speaking, the SC rainfall anomalies are local phenomena, i.e., rainfall anomalies are basically confined to SC, although there is a significant positive anomaly in the subtropical WNP in June. There is a negative anomaly north of the Yangtze River in July and August, consistent with the results from station rainfall data (Fig. 5). This negative anomaly, however, exhibits an apparent difference in significance area between the station and GPCP data. For instance, the significant negative anomaly in July occupies a larger area than that in August when using the GPCP data (Figs. 5c and d), but it is in contrast when using the station data (Figs. 5c and d). This difference is likely to be caused by the differences in periods between these datasets. The period for the station data is 1958-2002, while that for the GPCP data is 1979-2002. We examined the correlation coefficients using the station data during the period of 1979-2002, identical to that for the GPCP data, and found that the number of stations in central China significantly negatively related to SC rainfall was eight in July and five in August, while during the period 1958-2002 it was five in July and 12 in August. Thus, this difference in the analysis period leads to the different seesaw patterns in rainfall anomalies between the SC and central China shown in Figs. 5 and 7.

Figure 7.  Precipitation anomalies regressed upon SC rainfall derived from the GPCP dataset from 1979 to 2002: (a) May; (b) June; (c) July; (d) August. The shading indicates significance at the 95% confidence level by the Student's t-test. The contour interval is 0.5 mm d^-1 and the contour line of zero is omitted.

SC rainfall anomalies in May and July are related to significant precipitation anomalies over the tropics (Figs. 7a and c). The May SC anomaly is associated with a significant negative anomaly in the tropical North Pacific and a positive anomaly in the tropical South Pacific, and the July SC anomaly is associated with a significant positive anomaly in the Philippine Sea and a negative anomaly in the maritime continent. The relationship between the developing phase of ENSO events and SC rainfall anomalies in May and July may be responsible for these rainfall anomalies over the tropics.

The large-scale circulations exhibit a clear seasonal evolution in the lower troposphere over East Asia and the WNP (Fig. 8). The level of 700 hPa is used to represent the lower troposphere in the present study. There are some mountains with an elevation of 1000-2500 m in SC, and thus the data are reliable for analysis at this pressure level. In addition, there is abundant moisture at this level during summer over the northern Indian Ocean and SCS. In May and June, there are strong southwesterlies over SC (Figs. 8a and b). In July and August, by contrast, southerlies dominate over SC and the SCS, corresponding to the appearance of the monsoon trough over the SCS and WNP in these two months (Figs. 8c and d). The wind intensity decreases with time from May to August.

Figure 8.  Climatological horizontal winds at 700 hPa: (a) May; (b) June; (c) July; (d) August. Units: m s^-1.

Figure 9 shows the horizontal wind anomalies at 700 hPa associated with monthly SC rainfall variations. The circulation anomalies in May and June are characterized by the anticyclonic circulation anomaly over the SCS and the Philippine Sea. Associated with this anticyclonic anomaly, there is a southwesterly anomaly over SC and the subtropical WNP, and an easterly anomaly in the tropical WNP and southern part of the SCS. In July and August, by contrast, the circulation anomalies are characterized by a cyclonic circulation anomaly, which is centered over SC. Both the significant southwesterly anomaly over SC and the subtropical WNP and the easterly anomaly in the tropical WNP and southern part of the SCS shown in May and June do not appear in the results for July and August. Furthermore, a westerly anomaly appears in the tropical WNP and southern part of the SCS in July, in contrast to the easterly anomaly over these regions in May and June.

Figure 9.  Horizontal wind anomalies at 700 hPa regressed upon SC rainfall in (a) May; (b) June; (c) July; (d) August; (e) May-August; and (f) June-August. Units: m s^-1. The contour lines show the climatological geopotential height. Units: gpm. The shading indicates significance at the 95% confidence level by the Student's t-test.

In addition, there is a northeasterly anomaly north of SC in May and it becomes stronger in June. In July and August, on the other hand, a northeasterly anomaly dominates in the north and northwest of SC, illustrating a weakening of southwesterlies over there. These northeasterly anomalies, together with southwesterly or southerly anomalies, are favorable for the lower-tropospheric convergence over SC. Actually, without these northeasterly anomalies, the stronger southwesterly over the north or east of SC would transport more water vapor out of SC, eastward or northward (Fig. 10).

Figure 10.  Composite horizontal winds at 700 hPa for wet years (a-d) and dry years (e-h). Units: m s^-1. The contour lines indicate the wind speed. The contour interval is 2 m s^-1 in May (a, e) and June (b, f), and 1 m s^-1 in July (c, g) and August (d, h). The shading indicates the areas of wind speed greater than 7 m s^-1 in May (a, e) and June (b, f), 4 m s^-1 in July (c, g), and 3 m s^-1 in August (d, h).

These circulation anomalies correspond well to the climatological winds in each month. The geopotential heights are overlapped with the circulation anomalies in Figs. 9a-d to describe the climatological winds and facilitate a comparison between the climatology and anomalies. The regions of the southwesterly anomalies in May and June (Figs. 9a and b), i.e., SC and the subtropical WNP, are the same as the northern flank of the WNPSH, indicating an enhanced southwesterly over these regions. In addition, the easterly anomaly in the tropical WNP and southern SCS, and the southerly anomaly over the western SCS, are located at the southern and western flanks of the WNPSH, corresponding to enhanced winds over these regions. In July and August, on the other hand, the significant westerly anomaly over the Bay of Bengal and Indochina Peninsula, and the southwesterly anomaly over the northern SCS, also correspond to enhanced winds over these regions.

The above-mentioned results indicate that the enhanced winds in each month correspond to distinct wind anomalies shown in Figs. 9a-d, due to the remarkable change in climatological winds between May-June and July-August. These remarkable differences in the circulation anomalies among the months lead to very weak circulation anomalies associated with seasonal mean SC rainfall anomalies for either May-June-July-August (MJJA) or JJA, with a weak cyclonic anomaly over SC and a westerly anomaly over the northern Indian Ocean (Figs. 9e and f). Figure 10 shows the composite horizontal winds at 700 hPa. The wet years are defined when monthly SC rainfall is greater than one standard deviation and the dry years are defined when the rainfall is less than minus-one standard deviation during 1958-2002 (refer to Fig. 4). It should be mentioned that the wet and dry years are quite different among these four months, being consistent with the basic independence of interannual rainfall variation among the months.

In May and June, the southwesterlies are strengthened in the wet years (Figs. 10a and b) and weakened in the dry years (Figs. 10e and f). The composite southwesterlies are greater than 7.0 m s^-1 for the wet years, but less than this value for the dry years over the greater part of SC in both May and June. The wind speed for the wet years is also greater than the dry years over the northern SCS, with the former being around 5 m s^-1 and the latter being around 3 m s^-1. These results are highly consistent with the enhancement of winds indicated by Fig. 9.

In July and August, the southerlies are also reinforced in the wet years (Figs. 10c and d) and weakened in the dry years over the east part of SC and northern SCS (Figs. 10g and h). In July, the composite wind speed is greater than 4 m s^-1 in wet years over the Indochina Peninsula, northern SCS and SC (Fig. 10c), suggesting a channel of strong water vapor transportation. The composite wind speed for the dry years, by contrast, is weaker over these regions (Fig. 10g). In August, similar differences in wind speeds appear between the wet and dry years (Figs. 10d and h). In both July and August, there are strengthened southwesterlies in the northwest of SC in the dry years, which are related to a westward extended WNPSH. This westward extended WNPSH occupies SC and thus results in dry conditions in the region.

In the present study, interannual variations in monthly rainfall over SC from May to August during 1958-2002, and associated general circulation anomalies, have been examined. Both rainfall amounts and interannual standard deviations are the largest in these four months through the year. The corresponding standard deviation is around 60 mm month^-1, and accounts for 1/4-1/3 of monthly precipitation. Interannual variations in SC monthly rainfall are basically independent among these four months. Only May and July SC rainfall is significantly negatively correlated, partially due to the developing ENSO. During the developing phase of El Niño (La Niña) events, SC rainfall tends to be suppressed (enhanced) in May but enhanced (suppressed) in July.

The interannual variations in SC monthly rainfall are significantly associated with lower-tropospheric wind anomalies. Horizontal winds are stronger (weaker) in the lower troposphere over SC and the upstream parts in wet (dry) years for each of the four months. Corresponding to this monthly consistent enhancement of horizontal winds, however, the wind anomalies exhibit distinct features between May-June and July-August, due to the remarkable change in climatological winds. Climatologically, over SC and the northern SCS, southwesterlies dominate in May and June, as the northwest flank of the WNPSH, and southerlies dominate in July and August, resulting from the intensified monsoon trough. Therefore, a positive SC rainfall anomaly is associated with an anticyclonic circulation anomaly over the SCS and the Philippine Sea in May and June, but a cyclonic circulation anomaly centered over SC in July and August, and vice versa.