Sea level is an integrated manifestation of the ocean’s response to all dynamic and thermodynamic processes of oceanic, atmospheric, cryospheric, and terrestrial origin. Seasonal, interannual and decadal variability of sea surface height(SSH)in China coastal waters is expected to have signifi cant largerscale coupling effect in the ocean and climatic system. The SSH variability in China coastal waters and western North Pacific Ocean is under the infl uence of the monsoon, the Kuroshio, the El Niño/Southern Oscillation(ENSO) and other factors. Affected by various factors, the SSH variability has signifi cant multi-scale features and patterns, and the mechanisms of that are spatially variable. In the tropical Pacific and South China Sea(SCS), previous studies have stressed the role of ENSO and its teleconnection with extratropical atmospheric and oceanic circulation(Graham, 1994; Trenberth and Hurrel, 1994; Wu and Chang, 2005 ; Fang et al., 2006). On the interannual scale, the SSH in the tropical Pacific and SCS are signifi cantly correlated with the ENSO(Wu and Chang, 2005 ; Fang et al., 2006; Li et al., 2013; Wang and Cheng, 2013). Other large-scale atmospheric and oceanic internal variability such as the North Pacific Gyre Oscillation(NPGO) and Pacific Decadal Oscillation(PDO), may also affect the regional SSH from interannual to inter-decadal time scales, in areas such as the Japan/East Sea(JES) and western Pacific(Gordon and Giulivi, 2004; Zhang and Church, 2012).

The NPGO, defi ned as the second dominant mode of sea surface height anomaly(SSHa)in the central and eastern North Pacific, closely tracks the second dominant mode of North Pacific sea surface temperature(SST)anomalies, and is associated with decadal-scale variations in ocean gyre circulation(Nathan and Hare, 2002; Lornzo et al., 2010). The NPGO is driven by the atmosphere through the North Pacific Oscillation(NPO)(Di Lorenzo et al., 2008, 2010). Rossby wave dynamics excited by the NPO could propagate the NPGO signature from the central North Pacific into the Kuroshio-Oyashio Extension(KOE)(Ceballos, 2009), which suggests that the NPGO can be used to track changes in the entire northern branch of the North Pacific sub-tropical gyre and can provide a link between the eastern North Pacific and western boundaries. The PDO closely tracks the fi rst mode of North Pacific SST variability, and is connected to atmospheric circulation anomalies associated with the ENSO(Mantua et al., 1997). Low-frequency variability of SSH in the JES was in phase with the PDO index(Gordon and Giulivi, 2004). Modulation in mean Kuroshio Extension jet was remotely forced by wind stress curl anomalies in the eastern North Pacific Ocean related to PDO(Qiu, 2003). Han and Huang(2008, 2009)discussed the potential role of the PDO in the inter-anual and longer-term sea-level variability, in terms of regional manifestations such as ocean temperature and salinity and Kuroshio transport in China coastal waters. Their study revealed that sea-level variabilities in Bohai, Yellow and East China seas were correlated with the PDO. Besides, the rapid rates of sea-level rise in the western Pacific are partially due to decadal climate variability of PDO(Zhang and Church, 2012). Recently, it was found that sea surface wind anomalies induced by PDO strongly contributed to the mass induced SSH variability in the western North Pacific(Cheng et al., 2013).

Collectively, the factors contributing SSH variations are complex and vary over timescales and regions. Thus, interannual and decadal variations in SSH need to be systematically investigated for the North Pacific, which affects Chinese climate and coastal environment greatly. In this study, interannual and decadal variations of SSH in China seas were investigated using EOF, correlation analysis and regression analysis to examine the China Ocean Reanalysis dataset(CORA). The rest of this paper is organized as follows. Section 2 introduces the data used. SSH data from CORA is verifi ed by using tide gauge stations and satellite altimeter data in section 3. Section 4 examines the spatial-temporal patterns of CORA SSH from 1958–2008. Section 5 investigates interannual and decadal variability of SSH and their relationship with climate factors such as ENSO, PDO and NPGO. A brief summary is given in Section 6.

In this study, the CORA reanalysis dataset is used to examine the climatic modulation of ENSO/NPGO/ PDO on SSH variations in China seas. The CORA is an ocean reanalysis product of China seas that contains reanalysis of SSH, 3D temperature, salinity, and currents, covering a period of 51 years ranging from January 1958 to December 2008. The domain ranges from 99°E to 150°E and from 10°S to 52°N, covering the Bohai Sea, the Yellow Sea, the East China Sea(ECS), the South China Sea(SCS), and adjacent seas. The product contains monthly mean fi elds with horizontal resolution of 0.5° and 35 vertical levels. Multigrid 3D-var data assimilation is adopted to assimilate historical observation data, including temperature/salinity profi les from Nansen bottle, CTD, various bathythermograph and Array for Real-time Geostrophic Oceanography(Argo)fl oats, sea surface temperature(SST)from satellite remote sensing and a merged and gridded MSLA(Maps of Sea Level Anomaly)product(AVISO), which was produced based on TOPEX/Poseidon, Jason 1, and ERS-1 and ERS-2 data. The SSH product provides sea-level anomalies with spatial resolution of 1/4°×1/4° in Cartesian grid and daily temporal resolution. The time period from January 1994 to December 2008 was also used to examine the CORA SSH in Section 4.

Daily water-level data from tide gauges were sourced from the National Marine Data and Information Service(NMDIS) and The University of Hawaii Sea Level Center(http://uhslc.soest.hawaii.edu/uhslc/jasl. html), which is used to validate the CORA SSH. The monthly mean wind stress curl and net heat flux are calculated by the daily wind speed and latent heat flux, sensible heat flux, net long wave radiation and net shortwave radiation, which are provided by NCEP/NCAR reanalysis dataset. The period of 51 years ranging from January 1958 to December 2008 are used in this study.

In this study, the effective degrees of freedom in the correlation signifi cance test are estimated from the formula(Quenouille, 1952; Medhaug and Furevik, 2011; Wang et al., 2012):

The SSH data from 87 tide gauge stations and satellite altimeter in China seas are used to examinant the CORA SSH in this section. Prior to the verifi cations, the linear trend of SSH is removed. The daily SSH of the CORA are interpolated onto the positions of the tide gauge stations, and then compared with tide gauge stations data. The spatial distribution of correlation coeffi cient between the SSHa of CORA and tide gauge station is shown in Fig. 1, 70 of the 87 stations’ correlation coeffi cients exceed 0.5, which are statistically signifi cant at the 95% confi dence level. The correlation for only 3 of the 87 stations are below 0.2.

Fig. 1 Correlation coeffi cients between the SSHa of CORA and 87 tide gauge stations In the fi gure, red, green, blue point and blue fork denote correlation coeffi cient of 0.5–1, 0.2–0.5, 0–0.2 and -1–0, respectively.

Figure Options

Figure 2 shows the temporal series of SSH in station Nagasaki, Naha, Laohutan-A and Kanmean-A. The positions of the four stations are lined out in Fig. 1. The result indicates that the SSH of CORA can refl ect the temporal variations consistent with tide gauge stations.

Fig. 2 Time series of SSHa in tide gauge station Nagasaki, Naha, Laohutan-A and Kanmean-A The positions of the four stations are represented in Fig.1.

Figure Options

To better describe the spatial-temporal characteristics of the SSH, the empirical orthogonal function(EOF)method is used to analyze the gridded monthly SSHa from the CORA and altimetry observations from 1994–2008. Figure 3 shows the fi rst EOF spatial mode(EOF1)of the SSHa from the Altimetry observation and CORA, which account for 27.2% and 17.3% of the total variance, respectively.As shown in Fig. 3, the EOF1 of the CORA agrees well with that of Altimeter observation, which featured three gyre circulations with the anticyclonic gyres in the SCS and subtropical North Pacific, and the cyclonic gyre in the tropical North Pacific. The fi rst principal component(PC1)of the CORA and the altimeter are represented in Fig. 4. The correlation coeffi cient of the two series is 0.93, which is statistically signifi cant at the 99% confi dence level. In summary, these comparisons confirm that the SSH of the CORA successfully capture the temporal and spatial variability of SSH in China seas. In the next section, the CORA SSH is used to investigate SSH variations between 1958–2008.

Fig. 3 First EOF spatial mode of the SSHa from Altimetry observation (a) and CORA (b) The two modes account for 27.2% and 17.3% of the total variance, respectively.

Figure Options

Fig. 4 First EOF principal component of SSHa from CORA (blue dotted line) and Altimetry observation (black dotted line) Both time series are normalized, and the linear trend is removed.

Figure Options

To examine the spatial-temporal features of the SSH, the EOF is used to analyze the monthly SSHa of the CORA. Figure 5 shows the fi rst three leading modes of the CORA SSHa decomposition including the spatial distribution and the principal components(PCs)which account for 17.3%, 14.7% and 5.8% of total variance, respectively. The PCs are smoothed by a 24-month low-pass fi lter.

Fig. 5 First three EOF spatial modes (a, c, e) and principal components (b, d, f) of SSHa (CORA) The principal components are low-pass fi ltered (>24 months) and normalized before fi ltering. The three modes account for 17.3%, 14.7% and 5.8% of total variance, respectively.

Figure Options

As shown in Fig. 4, the EOF1 reveals remarkable zonal belt-like feature in the SSHa(Fig. 4a). In the mid-high latitudes(north of 20°N), there is a dominant anticyclonic gyre pattern embedded with many eddies. In the tropic(0°–20°N), a closed negative center locates around 10°N, suggesting that a cyclonic gyre covers the whole tropical area. The SCS shows positive center and negative loading along the coast, which indicates a basin-scale anticyclonic gyre. The PC1mainly refl ects the interannual variation(Fig. 4b).

The EOF2 represents the zonal-alternate characteristics of the SSHa(Fig. 4c). The negative loadings are found in China coastal waters including the Bohai Sea, Yellow Sea, ECS and SCS with a basin-scale cyclonic gyre. Positive loadings dominate regions in the east of Kuroshio and its extension. There is an anticyclonic gyre situated around 145°E, 25°N. The PC2 shows obvious interannual and decadal variations(Fig. 4d). The time series of PC2 reachs a minimum in 1974 and peaks at 2004, and shift from a negative to a positive phase in 1986, which denotes remarkable multidecadal variability over a 40-years period.

The EOF3 exhibits meridional-alternate feature of the SSHa(Fig. 4e). Negative loadings overlay the area from mid to high latitudes(35°–52°N) and in the tropic(0°–15°N). In contrast, the latitude belts of 15°–35°N and the SCS are occupied by positive loading. It is noticeable that the SCS exhibits the features with low loading in the center and high loading in the circumambience. The time series of PC3 refl ects obvious decadal variations, usually with negative values between 1958 and 1972 and positive values between 1973 and 1998.

Large-scale climate factors such as ENSO, NPGO and PDO are often invoked to explain physical and biological fl uctuations in the North Pacific Ocean(Lynn et al., 1998; McGowan et al., 1998 ; Lavaniegos and Ohman, 2003, 2007). In this section, we investigate the interannual and decadal variabilities of SSH, and their association with ENSO/NPGO/PDO. The method of EOF analysis is used to pick up the main information of the SSHa variations in Chinese seas. Although the EOF analysis cannot distinguish the effects of each climate signal separately, but the correlation in different time scale can reveal the relationship of the SSHa variations and different climate signals.

In the tropical Pacific and SCS, previous studies have stressed the role of ENSO and its teleconnection with the extratropical atmospheric and oceanic circulation(Graham, 1994; Trenberth and Hurrel, 1994; Fang et al., 2006). On the interannual and interdecadal scale, the SSHa in the SCS has signifi cant correlation with the Niño3.4 index(Wang and Cheng, 2013). Furthermore, the SSHa in the tropical Pacific has a close relationship with the ENSO events in the autumn, winter and spring(Li et al., 2013). China coastal waters and adjacent seas are under the infl uence of ENSO, and thus we now discuss the modulation of ENSO on the SSH in this area. The interannual variability of SSH was examined, with the time series of PC1/PC2/PC3 and ENSO index shown in Fig. 4. The PC1/PC2/PC3 and Niño3.4 index are smoothed by 24-month low-pass fi lter. As shown in Fig. 6, PC1 appears to fl uctuate signifi cantly on an interannual scale, particularly in the periods 1972/1973, 1987/1988, and 1997/1998(El Niño events) and 1971/1972, 1988/1989, and 1998/2000(La Niña events). The Niño3.4 index is in phase with PC1, which means that the positive(negative)phase of ENSO corresponds to the positive(negative)SSHa. The simultaneous correlation coeffi cient between PC1 and Niño3.4 index is 0.79, which is statistically signifi cant at 95% confi dence level. This indicates that the interannual variations of SSH in China seas are strongly modulated by the ENSO, but there are weaker correlations between the PC2/PC3 and Niño3.4 index.

Fig. 6 Comparison between the fi rst three EOF principal components of SSHa (black, blue and green line) and the climatic index of ENSO (red line) Each time series of PC1, PC2, PC3 and Niño3.4 index is normalized and low-pass fi ltered (>24 months). The correlation coeffi cients between PC1/ PC2/PC3 and Niño3.4 index are R1 (0.79)/R2 (0.14)/R3 (0.33), respectively. All of R1, R2 and R3 are statistically signifi cant at 95% confi dence level.

Figure Options

The Rossby waves dynamics excited by the NPO can take the decadal signals(NPGO and PDO)across the eastern Pacific to the western boundary. Therefore, the decadal signals in the eastern Pacific can also affect the middle and western Pacific. The result of Ceballos et al.(2009)revealed the mechanism of the NPGO and PDO affecting the SSH variability in central and western Pacific based on the signature patterns of the sea level pressure and wind. They found that the signifi cant fraction of the SSHa in the western Pacific was explained by the NPO related the wind stress curl anomalies, whereas the related Aleutian Low anomalies did not compare well with the SSHa. According to the conclusion, the effect of the sea level pressure that related to the NPGO and PDO will not be discussed in this paper.

NPGO has been identified as a decadal mode of climate variability that is linked to previously unexplained fl uctuations of salinity, nutrient, and chlorophyll in the Northeast Pacific(Di Lorenzo et al., 2008, 2009). The PDO is the other important source of multidecadal climate variability in the North Pacific, which has an ENSO-like spatial signature in the SST fi eld(Mantua et al., 1997). To investigate the decadal variability of SSH in China seas, the PC1/ PC2/PC3 of SSHa, NPGO and PDO index are smoothed by a 120-month low-pass fi lter. The time series of PC1/PC2/PC3 appear to exhibit signifi cant decadal fl uctuations.

NPGO shows an interdecadal oscillation with an approximate 15-year cycle, with the positive phases during the periods 1958–1963, 1972–1980, 1986– 1990, and 1998–2005 and the negative phases during the periods of 1963–1972, 1980–1986, and 1990– 1998(Fig. 7). The NPGO is signifi cantly negatively correlated with PC1. The correlation coeffi cient is -0.62, which is statistically signifi cant at 95% confi dence level. However, the correlations between the PC2/PC3 and NPGO index are not statistically signifi cant.

Fig. 7 Comparison between the fi rst three EOF principal components of SSHa (black, blue and green line) and climatic indices of NPGO (red line) and PDO (orange line) Each time series of PC1, PC2, PC3, NPGO and PDO index are normalized and low-pass fi ltered (>120 months). The correlation coeffi cients between PC1/PC2/PC3 and NPGO/PDO index are R11 (-0.62)/R12 (0.11)/R13 (-0.23) and R21 (-0.03)/R22 (0.48)/R23 (0.79), respectively. R11, R13, R22 and R23 are statistically signifi cant at 95% confi dence level.

Figure Options

The PDO is generally in its negative phase between 1958 and 1974 and switches to a positive phase after 1974(Fig. 7). However, it switches back to a negative phase after 1997. The PDO index is positively correlated with PC1/PC2, but the correlation coeffi cient is lower and not statistically signifi cant. PDO index is in phase with PC3 and the correlation coeffi cient is 0.79, which is statistically signifi cant at the 95% confi dence level.

Figure 8 shows the regressions of SSH onto the NPGO and PDO index. The spatial patterns in Fig. 7a highly resemble the opposite effect of EOF1 on SSHa(Fig. 4a), which further confirms the out-of-phase relationship between NPGO and PC1. Figure 8a also indicates that the positive(negative)NPGO phase tends to weaken(strengthen)the anticyclonic gyres in the SCS and subtropical North Pacific, and the cyclonic gyre in the tropical North Pacific. The spatial pattern in Fig. 7b shows consistency with the EOF3 of SSHa(Fig. 4e). This indicates that the positive PDO phase can induce the negative SSH anomaly in the east coast of China including the Bohai and Yellow sea, and the positive SSH anomaly in the subtropical western North Pacific. The reverse is true during the negative PDO phase.

Fig. 8 Regressions of SSH onto the climatic indices of NPGO (a) and PDO (b) Both climatic indices of NPGO and PDO are low-pass fi ltered (>120 months) before regressing. In the both fi gures, the black dots denote the regressions that are statistically signifi cant at 95% confi dence level.

Figure Options

The above analysis using the correlations and regressions suggests that the NPGO and PDO are clearly associationed with the decadal variability of SSH in the China seas. It is important to underst and the mechanism by which the NPGO and PDO could infl uence the SSH in the area. Recently, it is found that in the middle latitudes, the SSH is largely determined by wind stress curl variations associated with the forcing of NPGO and PDO(Ceballos et al., 2009; Zhang et al., 2010, 2011). Therefore, the climate thermodynamic and dynamic mechanisms of SSH are now discussed by analyzing the wind stress curl and net heat flux associated with the PDO and NPGO.

Figure 9 shows the correlations between PC1/PC2/ PC3 of SSHa and wind stress curl/net heat flux. The positive correlations between PC1 and wind stress curl extend from the subtropical western North Pacific to the whole SCS basin. Negative correlations are found in the east coast of China and tropical North Pacific. This suggests that the positive(negative)PC1 phase is relevant to the positive(negative)wind stress curl anomaly covering from subtropical western North Pacific to total SCS basin, and negative(positive)wind stress curl anomaly in the east coast of China and tropical North Pacific. The correlations between PC2 and wind stress curl represent the meridional-alternate, the southeast-northwest belt pattern. The signifi cant negative correlations between PC3 and wind stress curl stretch from the south of the SCS to northwest of Luzon Isl and whereas the positive correlations are less signifi cant in most of the region. The correlations between PC1/PC2/PC3 and net heat flux are less signifi cant.

Fig. 9 Correlations between PC1, PC2, PC3 of SSHa and wind stress curl ((a), (b), (c)) and net heat fl ux ((d), (e), (f)) In these fi gures, the black dots denote the correlations that are statistically signifi cant at 95% confi dence level.

Figure Options

Figure 10 shows the regression of the wind stress curl and net heat flux onto the NPGO and PDO index. The regression of wind stress curl onto the NPGO(Fig. 9a)exhibits the opposite pattern to Fig. 8a, especially in the area of SCS and tropic Pacific, which is concurrent with the negative correlation between the PC1 of SSHa and NPGO index(Fig. 6). These results indicate that the NPGO actually change the wind stress curl, and therefore induces the decadal variability of SSH. The regression between net heat flux and NPGO is less signifi cant in most portions of China seas, except for the Luzon Strait and south of the SCS(Fig. 9c). This suggests that the net heat flux anomaly related to the NPGO has less effect on the decadal variability of SSH.

Fig. 10 Regressions of wind stress curl (a, b) and net heat fl ux (c, d) onto climatic indices of NPGO (a, c) and PDO (b, d) In these fi gures, the black dots denote the correlations that are statistically signifi cant at 95% confi dence level.

Figure Options

Compared with the pattern in Fig. 9c, f, it is obvious that the wind stress curl and net heat flux connected to the PDO strongly modulate the EOF3 of SSHa(Fig. 8c, d). The spatial patterns in Fig. 9a are consistent with that of Fig. 8c. The two patterns are in phase, with the positive(negative)correlation between the PC3 and wind stress curl corresponding to the positive(negative)regressions between wind stress curl and PDO. These indicate that the PDO can cause the wind stress curl anomaly, and in turn modulate the SSH of the third mode. Furthermore, the spatial patterns of Fig. 9d are consistent with that of Fig. 8f in the tropics, including the positive patterns in the SCS and at 135°E, 16°N and 138°E, 3°N, and the negative center at 148°E, 8°N. This suggests that the variability of net heat flux in the tropic associated with PDO can contribute to the decadal variations in the third mode SSHa. However, it is should be noted that the third mode of the SSHa is largely determined by the wind stress curl anomaly connected to the forcing of PDO, rather than the net heat flux associated with the PDO.

In this paper, the temporal-spatial characteristics of SSH and the climate modulation of ENSO/NPGO/ PDO on SSH in China seas are investigated using the CORA dataset released by National Marine Data and Information Service.

The data from satellite altimeter and 87 tide gauge stations are used to verify the reanalysis SSH from CORA. The analysis shows that the temporal and spatial variations of CORA SSH are well consistent with the observations in China seas. Furthermore, based on correlation and regression analysis, the effects of ENSO/NPGO/PDO on the SSH in China seas are explored. It is found that theinterannual variations of SSH in China seas are strongly modulated by the ENSO, and that wind stress curl variability connected to the NPGO and PDO is the dominant contributor to the decadal variability of the SSH. The NPGO/PDO can actually change the wind stress curl, and therefore induces the decadal variability of SSH. In addition, the effect of net heat flux associated with the NPGO and PDO on decadal variability of SSH is also investigated. Results demonstrate that the net heat flux variability in the Luzon strait and tropic Pacific may infl uence the decadal variability of SSH, but has a reduced effect on the decadal variation of SSH in most regions of China seas.

In this study, our results confirm that decadal variations of SSH are linked to the strong dynamic and thermodynamic variability induced by NPGO and PDO. However, we only investigated the effect of the local wind stress curl and heat flux associated with NPGO/PDO on the decadal variability of SSH. It should be notet that the open boundary effect on the SSH variability has not been discussed in this paper. The eastern boundary in the regional model has strong volume transport and heat exchange that may be strongly impacted by the ENSO/NPGO/PDO. How ENSO/NPGO/PDO infl uence the volume and heat transport in the eastern boundary and thus modulate the SSH variability in the China seas warrants further studies.