The world’s oceans are one of the most importantfactors in determining atmospheric anomalies.The ocean varies much more slowly than the atmosphere, with a strong “memory” that results in a sustainedeffect on the atmosphere. Many previous studieshave examined the influences of sea surface temperature(SST)in the eastern equatorial Pacific, thewestern Pacific warm pool, and the Indian Ocean(IO)on summer precipitation in China. As a strong signalon the interannual timescale in the tropical Pacificair-sea system, the El Niño-Southern Oscillation(ENSO)is a strong source of interannual variabilityin the tropical Pacific air-sea system that significantlyinfluences weather and climate anomalies both globally(Webster et al., 1998) and within China(Yu and Jiang, 1994). The impact of ENSO on precipitation inChina has been one of the main concerns of Chinesescientists. Huang and Wu(1989)reported that summermonsoon precipitation increased in the Yangtze-Huaihe River valley but decreased in northern China and south of the Yangtze River during the El Niñodeveloping stage. By contrast, precipitation anomalieswere opposite during the El Niño decaying stage.Jin and Tao(1999)also found that different phasesof ENSO had different impacts on summer precipitationin eastern China. Zhang et al.(1996)showed thatthe impact of El Niño on precipitation in southernChina resulted from the development of an abnormalatmospheric circulation over East Asia during the maturestage of El Niño, with different impacts by season(Zhang et al., 1999, Zhang and Sumi, 2002). Wu et al.(2003)identified two main seasonal rainfall anomaliesin East Asia induced by the anomalous circulation systemsassociated with different phases of ENSO. Long and Li(1999)simulated below-average precipitation ineastern China during the summer following an El Niñoevent and below-average precipitation in the Yangtze-Huaihe River valley during the summer following a LaNiña event. ENSO is therefore often used as an importantpredictor of summer precipitation in China. However, Gao and Wang(2007)pointed out that the significanceof ENSO as a predictor of summer precipitationin China has declined since the 1970s. Some preliminarystudies have also shown that ENSO-relatedprecipitation anomaly patterns in China were oppositebefore and after the mid 1970s(Gao, 2006; Zhang et al., 2008). Ashok et al.(2007)identified a decreasein the frequency of ENSO events in recent years thathas resulted in changes in the ENSO influence on traditionalregions. These results suggest that more attentionshould be paid to how SST changes in otherocean basins influence summer precipitation in China.

Saji et al.(1999)introduced the concept of theIndian Ocean Dipole(IOD). Subsequent studies haveraised questions regarding whether the IOD is a localair-sea coupling phenomenon independent of thePacific(Behera et al., 1999; Webster et al., 1999; Yamagata et al., 2002; Yu and Lau, 2005), or whetherit has a strong relationship with ENSO(Murtugudde et al., 2000; Xie et al., 2002; Baquero-Bernal et al., 2002; Krishnamurthy and Kirtman, 2003). Studies ofthe influence of IO SST on precipitation have focusedon the relationship between the IOD and precipitation(Xiao et al., 2002); however, Xie et al.(2009), Yoo etal.(2006), and Yang et al.(2010)have found that themain mode of the IO SST anomaly(SSTA)was uniformtemperature increase or decrease throughout theentire basin, and these uniform temperature changesare significantly correlated with ENSO and have a substantialinfluence on Asian monsoon climate. Wu et al.(2009)demonstrated the importance of basin-wide IndianOcean warming to the East Asian-western Pacificsummer monsoon, and showed that it plays an activerole in modifying the northwestern Pacific anomalousanticyclone during ENSO decay through atmosphericKelvin waves and the Hadley circulation.

Recent studies have indicated that the IndianOcean has a significant impact on climate variabilityduring summer in the northwestern Pacific and EastAsia(Yoo et al., 2006). Observations and model simulationsshow that the Indian Ocean plays a particularlyimportant role in modifying the wind field overthe western Pacific(Zhang and Yang, 2007; Zhang et al., 2009; Yoo et al., 2010; Kug and Kang, 2006).Xie et al.(2009)found that a warm IO SSTA couldinduce eastward-propagating warm equatorial Kelvinwaves in the troposphere, while cold IO SSTA mightinduce cold equatorial Kelvin waves. These waves significantlyaffect the atmospheric circulation and summerclimate anomalies over the northwestern Pacific and East Asia. Zhu and Houghton(1996)found thatvariability in the Asian summer monsoon is very sensitiveto SST conditions in the tropical South IndianOcean. Yu et al.(2003)showed that the Indian Oceanwas more influential than the Pacific in determiningthe variability of tropical and monsoon climates duringspring in an air-sea coupled model. The impact ofIO SST on summer precipitation in China is thereforedeserving a further study. Unlike the IOD, few studieshave examined the influence of basin-wide changesin IO SST on summer precipitation in China. Likewise, although many previous studies have focused onENSO, few have discussed the relative roles of the Pacific and Indian oceans. Wu et al.(2010)used numericalexperiments to demonstrate that the relativecontributions of SSTA in the tropical Indian Ocean and western Pacific to maintaining the northwesternPacific anomalous anticyclone are seasonally dependentduring El Niño decay. The contribution of localSSTA in the western Pacific occurred in early summerin their simulations, while that in the tropical IndianOcean occurred in late summer. However, theseresults may be model-dependent. The present studytherefore aims to determine a means of effectively separating the role of SSTA in these two ocean basinsusing observational data. These results will then beused to determine the relative influences of IO SSTA and ENSO on summer precipitation in China.2. Data and methods

The data used in the present study include:(1)monthly mean precipitation data from 160 stationsover China compiled by the National Climate Center;(2)gridded(2°×2°)monthly SST data obtainedfrom the National Oceanic and Atmospheric Administration(NOAA)Extended Reconstructured SST V3(ERSST)dataset(Smith et al., 2008; Xue et al., 2003)(ftp://ftp.ncdc.noaa.gov/pub/data/cmb/ersst/v3b); and (3)gridded(2.5°×2.5°)atmospheric reanalysesof 500-hPa geopotential height and 850-hPa winds(u, v)obtained from NCEP/NCAR(Kalnay et al., 1996)(http://www.esrl.noaa.gov/psd/data/gridded/data.ncep.reanalysis.html). All three datasets are analyzedfor the same period(1979–2010). Two analysis domainsrepresenting the Pacific(30°S–30°N, 110°E–80°W) and the Indian Ocean(30°S–30°N, 40°–115°E)are selected for this study.

The methods applied in this study include empiricalorthogonal function(EOF)analysis, singularvalue decomposition(SVD), partial correlation analysis, and linear regression. EOF analysis is usedto investigate the temporal and spatial variations ofSSTA in the Pacific and Indian oceans during borealspring. SVD is used to study the distributions ofSSTA and summer precipitation in China, as well asthe relationship between them. The partial correlationmethod is used to determine the differing impactsof IO SST and Pacific SST(PSST)anomalies on summerprecipitation in China. Linear regression is usedto remove ENSO signals from SST and atmosphericfields to more clearly identify the influence of SSTAon precipitation anomalies in China.3. Dominant modes of variability in spring PSST and IO SST

Previous studies have commonly performed independentEOF analysis over either the Pacific or theIndian Ocean. In the present study, the EOF analysisis performed over both oceans as a combination(Fig. 1). The first EOF mode(Fig. 1a)contributes 35.55%of the total variance. Positive SSTA is apparent in theeastern equatorial Pacific, similar to the characteristicEl Niño pattern. El Niño events are typically decayingduring boreal spring, so the amplitude of SSTAin spring is less significant than in autumn or winter.The SSTA in the Indian Ocean is also positive and consistent throughout the entire IO analysis domain.The second EOF mode(Fig. 1b)contributes 18.8% ofthe total variance. This mode features negative SSTAin the eastern tropical Pacific, northwestern Pacific, and South Pacific, with positive SSTA in the easternNorth Pacific. The IO SSTA is negative and consistentthroughout the entire basin. The time coefficientof the first mode(Fig. 1c)varies mainly on interannualtimescales, while the time coefficient of the secondmode(Fig. 1d)varies on both interannual and interdecadaltimescales. Both EOF patterns include consistentSST anomalies throughout the tropical IndianOcean. The dominant modes of variability in the IndianOcean must therefore be investigated separately.

Fig. 1.(a, b)Spatial distributions and (c, d)time series of(a, c)EOF1 and (b, d)EOF2 of boreal spring SST anomaliesin the Pacific–Indian Ocean.

Figure options

Figure 2 shows the EOF analysis over the IndianOcean. The first two EOF modes explain 43.59% and 14.05% of the total variance, respectively. The spatialdistribution of the first mode(Fig. 2a)is very similarto the SSTA distribution in the Indian Ocean regionshown in Fig. 1a. Variability in the correspondingtime coefficient in Fig. 2c is again predominantly oninterannual timescales. The correlation coefficient betweenthe two time series in Figs. 1c and 2c is 0.43.The spatial distribution of the second mode(Fig. 2b)contains a clear east-west dipole, with positive SSTAin the east and negative SSTA in the west. The correspondingtime coefficient(Fig. 2d)varies predominantlyon interannual timescales, with no clear interdecadalvariability. Both the spatial mode and thetime coefficient differ substantially from those shownin Figs. 1b and 1d. The correlation coefficient betweenthe two time series(Figs. 1d and 2d)is –0.19.An EOF analysis has also been performed for SSTAin the Pacific region(30°S–30°N, 110°E–80°W). Thespatial distributions of the first two modes in the EOFanalysis over the Pacific domain(figure omitted)aresimilar to those of the first two modes of the full domain, with correlation coefficients of 0.99 between thetwo time series in Figs. 1c and 2c and 0.96 between thetwo time series in Figs. 1d and 2d. Considering thetwo oceans as a unit primarily captures the variabilityin the Pacific while concealing the variability in theIndian Ocean. It is therefore necessary to investigatethe Indian Ocean and its influence on precipitation inChina independently.

Fig. 2.(a, b)Spatial distributions and (c, d)time series of(a, c)EOF1 and (b, d)EOF2 of boreal spring SST anomaliesin the Indian Ocean(30°S–30°N, 40°–115°E).

Figure options

The leading mode of PSST variability is consistentwith that reported by other studies, while thesecond mode of IO SST variability is consistent withsome but not all previous results. Many previous studieshave identified close links between variations in IOSST and PSST(Yoo et al., 2010; Kug and Kang, 2006;Zhu and Houghton, 1996), but the correlation coefficientbetween the time series shown in Figs. 1c and 2cis only 0.43. This relatively low correlation suggeststhat IO SST also varies independently of PSST. Thefirst EOF mode of the IO SSTA contains an approximatelyuniform basin-wide change in temperature regardlessof whether the analysis is performed for theIO domain or the full domain. The atmospheric impactof this mode of variability in IO SST has not beenstudied as rigorously as the impact of the IOD.4. Effects of boreal spring SSTAs in the Indian and Pacific oceans on summer precipitation in China

Interactions between the different ocean basinsmay cause anomalies in PSST to be implicitly containedin studies regarding relationships between IOSST and precipitation in China. Similarly, anomaliesin IO SST may appear in studies regarding relationshipsbetween PSST and precipitation in China.Here, the independent influence of each ocean basinon precipitation in China is investigated by removingthe component of SST variability that is related toanomalies in the other basin.

ENSO is the most significant mode of interannualvariability in PSST anomalies. The Niño3.4 index istherefore used to characterize SSTA in the Pacific. Asthis study focuses on the first EOF mode of the IOSSTA, the time series shown in Fig. 2c(abbreviatedas IOidx)is used to characterize SSTA in the IndianOcean. Figure 3 shows the Niño3.4 index(solid line) and the IOidx(dashed line)after removing the lineartrends. The correlation coefficient between thetwo indices is 0.59. It should be emphasized that inthis study PSSTA targets ENSO variability, while IOSSTA targets variations in the first EOF mode.

Fig. 3. Normalized and detrended Niño3.4(solid line) and IOidx(dashed line)indices during boreal spring.

Figure options

The influences of each ocean basin on precipitationin China are evaluated by applying the partialcorrelation method(Behera and Yamagata, 2003)tothe indices shown in Fig. 3. Any SSTAs in the IndianOcean that were linearly related to Niño3.4 arededucted from the IO SST. Likewise, any SSTAs inthe Pacific Ocean that were linearly related to IOidxare deducted from PSST. This procedure enables aninvestigation of how variability in PSST and IO SSTindependently influences precipitation in China. Thepartial correlation equation is defined as

Click Browse formula

Figure 4 shows the results of the partial correlationanalysis between the boreal spring IOidx/Niño3.4 and summer precipitation in China. A positive IOidxanomaly may reduce precipitation in the area southof the Yangtze River, most of Inner Mongolia, and parts of Liaoning Province(Fig. 4a). Conversely, apositive IOidx anomaly may enhance precipitation inthe Yangtze River valley, southwestern and northernChina, northeastern Inner Mongolia, and parts of HeilongjiangProvince. A positive Niño3.4 index appearsto have the opposite effect on precipitation in China(Fig. 4b). These results indicate that changes inducedby spring IO SSTA may counteract changes inducedby PSST anomalies(labeled as “+” and “–” in Fig. 4).Although the correlation coefficient between the IOidx and Niño3.4 index is 0.59, the ways in which these twoocean basins influence precipitation in China may differ.

Fig. 4. Partial correlation maps between anomalous precipitation during late summer in China and (a)spring IOidx and (b)Niño3.4 index. Blue contours represent precipitation anomalies that are significant at the 95% confidence level.

Figure options

The impact of ENSO on precipitation in Chinahas been studied widely. It is therefore valuable toconsider the influence of SSTAs with the ENSO signalremoved on precipitation in China. The ENSO signalis removed from the SST and precipitation fields usingthe linear regression method outlined by An(2003).The specific method is as follows.

The variable ξ^*(x, y, t) is defined as a function ofspace(x, y) and time(t), while the time series Z(t)(i.e., Niño3.4 index)is defined as a function of timet. The modified variable ξ(x, y, t) is calculated by removingthe signal of Z(t)from ξ*(x, y, t):

Click Browse formula

An SVD is performed on boreal spring SSTA and summer precipitation anomalies in China after removingthe ENSO signal. The first two modes are shownin Fig. 5. The first and second modes contribute34.6% and 15.4% of the total covariance, respectively, with corresponding hetero-correlation coefficients of0.85 and 0.77. The main modes of SSTA variability inthe Indian Ocean remain uniform basin-wide increasesor decreases in temperature even after removal of theENSO signal; however, the ENSO pattern is removedfrom the leading modes of SSTA variability in theeastern tropical Pacific(Figs. 5c and 5d). The correspondingspatial distributions of precipitation aresignificantly different from the first and second modesshown in Fig. 5. The first mode(Fig. 5a)is characterizedby increases in summer precipitation overMongolia, northeastern and northwestern China, and decreases over most other parts of China. The secondmode(Fig. 5b)is characterized by decreases in precipitationover most regions south of the Yangtze River and increases in the Yangtze River valley, northern and northeastern China, and parts of Mongolia. Thetime series corresponding to the first mode(Fig. 5e)is similar to that over the full analysis domain(Fig. 1d), with clear interdecadal variations and an abruptchange around 1997. By contrast, the time series associatedwith the second mode(Fig. 5f)varies mainlyon interannual timescales. The time series based onSSTA(black solid lines in Fig. 5f)are similar to IOidx, with statistically significant(> 99%)correlation coefficientsas high as 0.6. This study primarily concernsvariability on interannual timescales, so the followinganalysis will focus on the relationship between borealspring SSTAs and anomalous summer precipitation inChina shown in Figs. 5b and 5d. The relationshipshown in Figs. 5a and 5c, which varies predominantlyon interdecadal timescales, will be largely ignored, althoughit is worth noting that Fig. 5c suggests thatthe recent weakening of the East Asian summer monsoonmay be attributable to warming in the tropicalIndian and western Pacific oceans. This result is consistentwith the conclusions of previous studies(Zhou et al., 2008, 2009).

Fig. 5. Hetero-correlation fields for(a, b)precipitation, (c, d)SST, and (e, f)time series of(a, c, e)SVD1 and (b, d, f)SVD2 for Pacific-Indian Ocean SST anomalies during boreal spring and summer precipitation anomalies in Chinaafter removing the ENSO signal. The solid and dashed lines in(e, f)denote the SST and precipitation time series, respectively.

Figure options

In Figs. 5b and 5d, it is seen that uniform basinwidewarming in the Indian Ocean is associated withdecreases in precipitation over most parts of the regionsouth of the Yangtze River and increases in precipitationover the Yangtze River valley, northern and northeastern China, and parts of Mongolia. Figure 6shows regression maps of anomalous summer precipitationrelative to the detrended time series of borealspring SSTA and vice versa. The regressed distributionsof anomalous precipitation(Fig. 6a) and SSTA(Fig. 6b)are both similar to the anomalous distributionsof precipitation and SSTA shown in Figs. 5b and 5d. For precipitation(Fig. 6a), this similarityextends not only to the spatial distribution but alsoto the center of the anomalies. The regressed SSTAis significant at the 95% confidence level for most ofthe Indian Ocean, but for only a small portion of thePacific Ocean between 10°S and 10°N. These resultsindicate that IO SSTAs have a more substantial influenceon precipitation over China than Pacific SSTAswhen the ENSO signal is removed. The distributionof anomalous precipitation when ENSO signals areremoved from SSTAs is consistent with the distributionof anomalous precipitation(Fig. 4a)associatedwith IO SSTAs. This further indicates that IO SSTAsdominate the non-ENSO relationship between anomalousSST and anomalous precipitation over China.

Fig. 6.(a)Regression map of summer precipitationanomalies(mm)against detrended SSTA(black solid linein Fig. 5f). Blue contours represent precipitation anomaliessignificant at the 95% confidence level.(b)Regressionmap of boreal spring SSTA(℃)against detrended precipitationanomalies(red dashed line in Fig. 5f). Shadedareas represent SST anomalies(℃)significant at the 95%confidence level.

Figure options

This section explores possible mechanisms behindthe impacts of boreal spring SSTA without ENSOsignals(Fig. 5d)on anomalous summer precipitationover China. Figure 7 shows anomalous summer500-hPa geopotential height and 850-hPa wind fieldsregressed against the detrended time series of SSTAwithout ENSO signals(solid line in Fig. 5f, and notrend). The regressed summer anomalous 500-hPageopotential height field in the Northern Hemisphere(Fig. 7a)contains a statistically significant positiveanomaly in the region from the Philippine Isl and s tosubtropical Indochina, including the region south ofthe Yangtze River. Moreover, an east-west dipole ofgeopotential height anomalies occurs over midlatitudeAsia, with negative anomalies in the east and positiveanomalies in the west, and a summer blockinganticyclone is apparent over the Sea of Okhotsk(locatedto the north of Japan). The regressed anomalous850-hPa wind field(Fig. 7b)contains a significanteasterly wind anomaly along the equator in the tropicalwestern Pacific region and a prominent low-levelanticyclone over the South China Sea and northwesternPacific Ocean. This anomalous anticyclone shiftswestward relative to the anticyclone reported byWanget al.(2003) and extends partially over mainl and China. A convergence belt in the 850-hPa wind fieldappears around 30°N, between the cyclonic anomaliesnorth of 30°N and the anticyclonic anomaly south of30°N. This convergence belt favors the formation ofprecipitation. A positive geopotential height anomalyprevails to the south of 30°N in the regressed 500-hPageopotential height field, while a negative anomalyprevails to the north(Fig. 7a). This is consistent withthe positive precipitation anomaly near 30°N shownin Figs. 5b and 6a. Conversely, the positive 500-hPageopotential height anomaly and anticyclonic circulationin the subtropics south of 30°N correspond to anegative precipitation anomaly(Figs. 5b and 6a).

Fig. 7. Regression maps of summer(a)500-hPa height anomalies(gpm) and (b)850-hPa wind anomalies(m s−1)against detrended SSTA(black solid line in Fig. 5f). Shaded areas in(a)represent values significance at the 95%confidence level.

Figure options

The regressed SSTA field shown in Fig. 6b indicatesthat precipitation over China is strongly affectedby SSTAs mainly in the Indian Ocean, particularly inthe southern Bay of Bengal just north of the equator.Easterly anomalies in low-level wind near the equatorare a part of the atmospheric response to positiveSSTA in the tropical IO and the accompanyingKelvin wave(Gill, 1980). The most significant 500-hPa geopotential height response to heating of thetropical IO is located over the extratropical and midlatituderegions of continental Asia, consistent withthe conclusions of Lin(2009). The summer blockinghigh near the Sea of Okhotsk may also influencesummer precipitation in China(Wu and Zhang, 2011). The negative SSTA in Northwest Pacific(Fig. 6b)contributes to the formation of the anticyclonicanomaly over the South China Sea(Fig. 7b). Theeasterly wind anomalies in the tropics enhance thisanticyclone. Positive tropical IO SSTA and negativenorthwestern PSST anomalies combine to induce astrong anticyclonic anomaly over the South China Sea-Northwest Pacific region. This anticyclonic anomalyexerts substantial influences on the climate over EastAsia(Zhang et al., 1996). Lau and Nath(2000)showed that the Northwest Pacific anticyclone playsan important role in the East Asian summer monsoonsystem. Positive IO SSTAs related to ENSO promotethe development and maintenance of this anticycloneduring boreal summer through a mechanism similarto that of a thermal capacitor(Behera et al., 1999;Murtugudde et al., 2000). This study has shown thatthe development of an anticyclonic anomaly over theSouth China Sea-Northwest Pacific area is not dependenton ENSO; rather, this anomaly can arise independentlyfrom local negative SSTAs in Northwest Pacific and positive SSTAs in the tropical Indian Ocean. Wuet al.(2010)have previously demonstrated this resultusing idealized SSTA-driven AGCM experiments.They showed that the combined effects of local SSTAsin the tropical Indian Ocean and northwestern Pacificcould create an anticyclonic anomaly in the westernPacific even in the absence of SSTAs in the equatorialcentral-eastern Pacific. Further study regarding theimpact and extent of ENSO-related warming in thetropical IO and the formation of anticyclones in thewestern Pacific is still required.7. Conclusions

The dominant modes of boreal spring variabilityin Pacific and Indian Ocean SST are derived, and thepatterns of variability in IO SST are demonstrated tobe more robust. Spring IO SST(with the portion relatedto the Niño3.4 index removed) and PSST(withthe portion related to IO SST variability removed)are investigated and shown to have substantial influenceson summer precipitation over China. Anomalouswarming of the IO results in a negative-positivenegative-positive pattern of precipitation anomaliesin eastern China from south to north. Anomalouswarming in the Pacific region results in a qualitativelyopposite distribution of precipitation anomalies. Afterremoving ENSO signals, summer precipitationanomalies in China are influenced primarily by thestrength and position of the lower tropospheric anticycloneover the South China Sea–Northwest Pacific and related geopotential height anomalies in the middletroposphere. Positive SSTA in the IO inducesremote atmospheric responses in the extratropical and midlatitude regions of Asia. Low latitude easterlywind anomalies triggered by IO SSTA enhancethe anomalous anticyclonic circulation in the SouthChina Sea–Northwest Pacific. Recent observationshave indicated a weakening of the ENSO influence onsummer precipitation over China. If this trend continues, relationships with spring SSTA in the tropical IOmay become increasingly useful for forecasting summerprecipitation in China.

Acknowledgments. The language editor forthis manuscript is Dr. Jonathon S. Wright.