Relationship over southern China between the summer rainfall induced by tropical cyclones and that by monsoon

Abstract This study investigates the relationship between tropical cyclone–induced precipitation (P TC) and summer monsoon–induced precipitation (P SM) in southern China (SC) during June–August. The spatial patterns of the first interannual mode are uniform in sign over SC, with positive anomalies for P TC and negative anomalies for P SM. The background of an increase in cyclonic vorticity, an increase in RH, and a decrease in vertical wind shear over the South China Sea (SCS)–western north Pacific (WNP) provides favorable conditions for more TC genesis. The positive equatorial central Pacific SST anomaly and negative North Indian Ocean SST anomaly contribute to the anomalous cyclone over the SCS–WNP, which causes decreasing P SM in SC together with an anomalous anticyclone over eastern China–Japan. By contrast, whilst the spatial patterns of the first interdecadal mode are also uniform in sign over SC, there are positive anomalies for both P TC and P SM. The first interdecadal principal component features significant positive correlation with the number of TCs forming in the SCS. There is a significant increase in P TC and P SM after early 1990s. A positive tropical Indian Ocean (TIO) SST anomaly persists from the preceding winter to summer. During the preceding winter and spring after the early 1990s, a positive western Pacific SST anomaly can result in TIO SST warming through vertical circulation. Then, the positive TIO SST anomaly triggers an anomalous WNP anticyclone and contributes to the interdecadal increase in SC P SM in the succeeding summer. The persistent heating source over SC from May to summer related to an earlier onset of the SCS monsoon may strengthen the East Asian subtropical summer monsoon.


Introduction
Southern China (SC) is a subtropical coastal area in the Asia-Pacific region and, according to the division of the summer monsoon domain put forward by Wang and Ho (2002), belongs to the East Asian summer monsoon region. The Asian summer monsoon prevails in May-September every year and brings warm and moist air from the oceans to eastern China (Zhang, Liao, and Li 2010). Climatological rainfall over SC has two peaks, one appearing in April-June and the other in August-September. These peaks are mainly associated with the summer monsoon and the passage of tropical cyclones (TCs), respectively (Chan and Shi 1999). Chan and Shi (1999) defined the total rainfall between April and June over SC as summer monsoon precipitation (P SM ). Subsequently, this abbreviation has similarly been adopted when referring to the total data are used to explore the differences in circulation and SST between the 1993SST between the -2002SST between the and 1983SST between the -1992 means in a parallel analysis. TC track data are derived from the JTWC western North Pacific (WNP) best-track data for the period 1945-2010, with an interval of 6 h. For consistency with the daily rainfall, we convert the best-track data to daily means.
TCs form in the SCS (0°-23°N, 100-120°E) or WNP (north of equator, 120-180°E). Also, TCs may form in the WNP and then enter into the SCS. The definition of summer P TC follows that of Chen, Wu, and Wen (2012). When a station falls within an effective radius of 550 km from the center of a TC, the daily station rainfall is defined as P TC . Then, the residual parts of total summer rainfall minus P TC constitute P SM . Boreal summer refers to the months from June to August in this study. The interdecadal component is extracted with a nine-year Gaussian filter. The interannual component is obtained by subtracting the interdecadal component from the original data. Correlation and regression analyses are conducted, and the significance is based on the Student's t-test. In the significance test for interdecadal components, the effective degrees of freedom is determined based on the method of Chen and Chen (2011). To investigate the atmospheric response to the SST boundary condition, the atmosphere component (CAM4) of CESM1.0.4 is employed. The control run is forced by the climatological mean seasonal cycle of global SST with a 26-year integration, and the outputs over the last 20 years are used for analysis. Sensitivity experiments forced by positive SST anomalies over the western Pacific imposed on the climatological mean seasonal cycle of global SST in winter and spring, are performed to demonstrate the role of western Pacific SST anomalies.

Relationship between interannual P TC and P SM over SC
From the spatial distributions of climatological mean P TC and P SM in summer, it can be seen that the centers of P TC and P SM are located over SC (Supplementary Figure S1). Therefore, [105][106][107][108][109][110][111][112][113][114][115][116][117][118][119][120] is selected as the key region for this research. We investigate the joint variability features between P TC and P SM with EOF analysis, which can delineate the spatiotemporal characteristics of rainfall variability. Both P TC and P SM from 71 stations in SC are jointly subjected to EOF analysis for their interdecadal and interannual components, which is regarded as a combined EOF analysis. In this study, we only discuss the major eigenmodes accounting for 41.9% and 25.5% of the interdecadal and interannual rainfall variance from the selected station, respectively.
An opposite sign distribution is observed between P TC and P SM for their interannual components (Figure 1(a) and (b)). Positive anomalies of P TC and negative anomalies of precipitation over South China during May to June (Chan and Zhou 2005;Zhou, Li, and Chan 2006).
The severe wind and torrential rain associated with landfalling TCs and prevailing TC tracks cause considerable economic losses and casualties in affected areas. Zhang, Liu, and Wu (2009) indicated that, on average, seven TCs made landfall each year during the period 1983-2006 over the Chinese mainland and Hainan Island, leading to average annual direct economic losses of 28.7 billion RMB and 472 deaths. Typhoons usually induce precipitation over China during April to December, with the maximum amount of precipitation falling in August (Wang et al. 2008). Looking specifically at the SC coast, they tend to land mainly during June to November, with the heaviest rainfall observed in August-September (Tian, Ma, and Lin 1999). Previous studies have pointed out that the ratio of annual TC-induced precipitation (P TC ) in Southeast China to total annual precipitation is more than 10% (Ren et al. 2006;Rodgers, Adler, and Pierce 2000;Xu 2007). Taken together, therefore, TCs and the summer monsoon induce a substantial quantity of rainfall in summer over SC.
TCs forming in the South China Sea (SCS) contributed to an increase in SC summer rainfall around 1993 (Chen, Wu, and Wen 2012). But is there a relationship between P TC and P SM in SC? In Taiwan, P TC and P SM tend to vary inversely on both interdecadal and interannual time scales Chen, Li, and Shih 2010). So, given that SC and Taiwan are situated at almost the same latitude, what are the interdecadal and interannual variabilities of summer P TC and P SM over SC?
The organization of this study is as follows: The datasets and method are described in Section 2. Sections 3 and 4 delineate the interannual and interdecadal relationships between P TC and P SM over SC and their regulating processes. A summary is provided in Section 5.

Datasets, methods, and model
Daily rainfall data from 756 stations for the period 1951-2008, provided by the Chinese Meteorological Data Service Network, are used in this study. We only select stations with missing data for no more than one month in an individual year, and without site changes from 1959 to 2008. In all, 525 stations are included for the Chinese mainland. It has been reported that unrealistic interdecadal variability may exist in NCEP-NCAR reanalysis data (Greatbatch and Rong 2006;Inoue and Matsumoto 2004;Wu, Kinter, and Kirtman 2005;Yang, Lau, and Kim 2002); therefore, we adopt ERA-40 (Uppala et al. 2005) to analyze the circulation for the period 1959-2002. The monthly SST is obtained from ERSST.v3 (Smith et al. 2008), which is available from 1854 to the present day at a resolution of 2.0° × 2.0°. ERA-Interim (Dee et al. 2011) and HadISST (Rayner et al. 2003) P SM exist in the first spatial mode of interannual variability. From the corresponding principal component (PC), the year 1990 was the typical year for increasing P TC and decreasing P SM , and the year 1993 was opposite. The correlation coefficient between the interannual time series of P TC and P SM averaged over SC is −0.3, significant at the 95% confidence level.
So, what are the regulating processes from the largescale anomalies of the oceans and atmosphere? Anomalies of SST, 500-hPa p-vertical velocity and 850-hPa winds regressed on the normalized first interannual PC of P TC and P SM are shown in Figure 2(a) and (b). Significant positive SST anomalies are observed in the tropics from 160°E to the date line. The positive anomalies of SST in the equatorial central Pacific are associated with ascent over the SCS-WNP, which triggers an anomalous cyclone over the WNP as a response to the Rossby wave. This result is consistent with He and Wu (2014). Meanwhile, significant negative SST anomalies appear in the northern Indian Ocean (NIO). The NIO SST anomalies contribute to an anomalous cyclone over the WNP through an east-west vertical circulation between the NIO and WNP, which was previously demonstrated by a sensitivity experiment using CAM4 in He and Wu (2014). Divergent winds dominate SC, which is affected by a southern cyclone in the SCS and northern anticyclone in eastern China-Japan. As such, P SM is suppressed over SC following weakened southwesterly flows.
Important factors contributing to TC activity include 850-hPa vorticity, 600-hPa RH, and vertical wind shear between 200 and 850 hPa (Camargo, Emanuel, and Sobel 2007;Gray 1979). In Figure 2(a) and (c), we can see an increase in cyclonic vorticity at 850 hPa, an increase in RH at 600 hPa, and a decrease in vertical wind shear over most of the SCS-WNP region. Thus, the increase in SC P TC can be explained by the TCs forming in the SCS and WNP. The correlation coefficient between the first interannual PC and the number of TCs forming in the SCS and WNP that bring rainfall to SC is 0.251, significant at the 90% confidence level.

Relationship between interdecadal P TC and P SM over SC
For the first modes of interdecadal components in Figure  1(c) and (d), their spatial variabilities feature a better coherence, with positive P TC and P SM anomalies over SC. Obvious positive anomaly centers of P TC are observed over coastal areas, whereas the distribution of P SM is highly uniform. There are noticeable interdecadal oscillations in the corresponding temporal variability. It can be seen that a significant decrease occurred in the early 1960s, and an increase in the 1990s. These results indicate that P TC and P SM decreased during 1975-1990 and increased before the early 1960s and after the early 1990s. The correlation coefficient between the interdecadal time series of P TC and P SM averaged over SC is 0.64, significant at the 99% confidence level.
But why is there a different relationship between P TC and P SM for the interdecadal and interannual components? To answer this, the regulating processes for the first interdecadal mode of P TC and P SM are examined in this section. The anomalies of SST, 500-hPa vertical velocity, and are responsible for the local ascending motion, while the descending motion over the WNP contributes to the locally warmer SST. Kajikawa and Wang (2012) indicated that the onset date of the SCS summer monsoon (SCSSM) advanced significantly around 1993/1994. The correlation coefficient between the first interdecadal PC and the SCSSM onset date (Wang et al. 2004) is −0.63, significant at the 99% confidence level. After the onset of the SCSSM, atmospheric heating is enhanced and moves northwards (Zhu et al. 2011). The earlier SCSSM onset after the early 1990s may strengthen the atmospheric heating in SC. Meanwhile, the East Asian subtropical summer monsoon (EASSM) and SC P SM become stronger. From the anomalies of vertically integrated atmospheric heating from SLP to 300 hPa regressed on the normalized first interdecadal PC, the positive anomalies of atmospheric heating are over the SCS and SC in May. The atmospheric heating in SC persists from May to 850-hPa winds in summer are obtained by regression onto the first interdecadal PC of P TC and P SM (second column of Figure 2). Significantly warm SST anomalies occur in the tropical Indian Ocean (TIO) and WNP, and there is no notable SST change in the equatorial central-eastern Pacific. Corresponding to the SST anomalies, the p-vertical velocity at 500 hPa displays anomalous ascent over the TIO and anomalous descent over the WNP. Anomalous ascending motion is seen over SC. At 850 hPa, strong anticyclonic winds cover the WNP and the regions from Lake Baikal to Northeast China. Southwesterly anomalies blow from the SCS to SC and converge with northerly anomalies from the northern anticyclone. Wu et al. (2010) indicated that the development of the northern anticyclone was related to an increase in Tibetan Plateau snow cover during the preceding winter-spring, and the southern anticyclone was associated with an increase in the equatorial Indian Ocean SST. The positive anomalies of SST over the TIO related to an increase in local SST, with a contrast between the 1983-1992 mean and the 1993-2002 mean. The differences in the 1993-2002 mean and 1983-1992 mean of SST in the preceding winter and spring, based on ERSST data, are shown in Figure 3(a) and (e), respectively. Significant positive SST anomalies over the TIO are observed in the preceding winter and spring, which is similar to that based on HadISST data. The warm SST anomalies over the WNP induce ascending motion and lower-level convergence from the preceding winter to spring, which is concurrent with the descent and lower-level divergence over the TIO (Figure 3(b)-(h)). The differences in lower-level and higher-level circulation between 1993-2002 and 1983-1992, derived from ERA-40 (Supplementary Figure  S3), exhibit a distribution that is slightly weaker than that derived from ERA-Interim (Figure 3(b)-(h)). This implies summer and becomes more significant (Supplementary Figure S2). This is conducive to the enhancement of the EASSM and increasing SC P SM . The EASSM (definition in Liu et al. (2012)) became much stronger after the early 1990s, while the SCSSM (defined in Wang et al. (2009)) became weaker. This implies that the advanced SCSSM onset contributes to an enhanced EASSM through atmospheric heating, which results in the increase of SC P SM .
The correlation between the first interdecadal PC and the number of TCs forming in the WNP or entering into the SCS that bring rainfall to SC are not significant, whereas there is close correlation between the first interdecadal PC and the number of TCs forming in the SCS that bring rainfall to SC (0.644, significant at the 99% confidence level, according to the Student's t-test). Chen, Wu, and Wen (2012) suggested that the increase of TCs in the SCS is notes: dots in (a, b) and (e, f) and shading in (c, d) and (g, h) denote the differences in sst, vertical velocity and divergent winds significant at the 90% confidence level, according to the student's t-test. contour intervals are 0.1 × 10 6 s −1 at 850 hpa in (c, g) and 0.3 × 10 6 s −1 at 200 hpa in (d, h). the sst data-set is obtained from ersst.v3. the winds and vertical velocity datasets are derived from erA-interim. accounts for 41.9% (25.5%) of the total variance. The interdecadal PC features an interdecadal increase in the early 1990s, and its eigenvector exhibits uniform patterns with positive P TC and P SM anomalies over SC. In contrast, the interannual eigenvectors feature uniform patterns with positive P TC anomalies and negative P SM anomalies.
For the interannual components, regression analysis reveals that anomalous ascent and cyclonic circulation over the SCS-WNP feature a response to significantly positive SST anomalies in the equatorial central Pacific and negative SST anomalies in the NIO. An anomalous cyclone in the SCS and anomalous anticyclone in eastern China-Japan cause divergent winds over SC and negative anomalies of SC P SM . For P TC anomalies, more TCs forming in the SCS and WNP cause an increase in SC P TC . An increase in cyclonic vorticity, an increase in RH, and a decrease in vertical wind shear over the SCS-WNP, provide favorable conditions for TC genesis.
For the interdecadal components, regression analysis reveals significant positive anomalies of SST over the WNP and TIO. Anomalous ascent and anomalous descent are observed over the TIO and WNP, respectively, which can be connected by vertical circulation. An increase in P SM is attributed to common effects of the WNP anticyclone and Northeast China anticyclone. The anomalously warmer SST that there is northeast-southwest oriented vertical circulation between the WNP and TIO, which is demonstrated in Figure 4. To highlight the role of western Pacific SST anomalies, we conducted an experiment with positive SST anomalies added in the tropical western Pacific (TWP) in winter and spring (Figure 4(a) and (d)). From the differences in 850-hPa winds, 500-hPa p-vertical velocity, and vertical circulation between the experiments of positive TWP SST and the CAM4 control in winter and spring, the circulation anomalies feature a response to the positive SST anomalies over the TWP. The anomalous ascent over the TWP can cause descent over the TIO through vertical circulation. The anomalously warmer SST over the TIO is affected by atmospheric circulation during the preceding winter and spring, and turns to force the atmosphere in summer.

Summary
Summer rainfall over SC is evidently affected by both the monsoon system and TC activity. The main purpose of the present analysis was to reveal the variability of P TC and P SM over SC and their regulating processes. The combined EOF analysis for both P TC and P SM anomalies from 71 stations in SC discloses that the first interdecadal (interannual) mode over the TIO can persist from the preceding winter to summer. During the preceding winter and spring, warmer SST in the TWP can cause descent and anomalously warmer SST over the TIO through northeast-southwest titled vertical circulation, as demonstrated by numerical experiments with CAM4. The air-sea interaction is opposite to that in summer. Due to the earlier onset of the SCSSM after the early 1990s, the persistent positive heating source over SC strengthens the EASSM and contributes to an increase in P SM . The increase in P TC over SC is related to more TCs forming in the SCS.
The above results help us to gain greater insight into the variability of SC summer rainfall through separating total rainfall into P TC and P SM . If SC summer rainfall is analyzed using the traditional approach of total rainfall, the interesting phenomena uncovered in this study will be ignored.