3.1 Sea level variability responses to ENSO
Sea surface temperatures along the coast in our study area have a close relationship with ENSO, which will then influence the sea level variability. We perform EOF analysis on all the 14 tide gauge stations corrected for VLM in Region 1. After removing VLM, the linear trend and the seasonal cycle from the SLC time series, the first three leading principal components and their eigenvectors are shown in Fig. 4. The first EOF mode accounts for 52.6% of the variance, while the second accounts for 10.0%, and the third EOF accounts for 9.0% of the total variance.
Figure 5 presents the wavelet coherence between SOI and the amplitude time series of the first EOF of Region 1, PC1. Figure 5a shows that the two series vary coherently, generally, positive SOI (i.e., La Niña events) correlating with positive amplitudes of PC1 time series while negative SOI (i.e., El Niño events) correlating with negative amplitudes of PC1 time series in Region 1. Figure 5b presents the wavelet coherence between SOI and PC1, and the results show that SOI and PC1 time series have significant (p < 0.05, against red noise) coherence in the 8-to-16-year band during 1980–2016, and the PC1 time series leads the SOI by 1/4 of the period. Therefore, it can be inferred that the impact of ENSO on sea level variations in the East China Sea, the Yellow Sea, the Bohai Sea, and adjacent waters are mainly on decadal or longer time scales.
We also perform EOF analysis on all tide gauge records in Region 2 after removing the linear trend, the VLM, and the seasonal cycle, the results are shown in Fig. 6. The variance explained by the first three principal modes are 66.5%, 20.7% and 8.5%, respectively. Figure 7 shows the wavelet coherence between the amplitude time series of the first principal mode (PC1) of Region 2 and SOI. Figure 7a also shows that the two series vary coherently, positive SOI correlating with positive amplitudes of PC1 time series and negative SOI correlating with negative amplitudes of PC1 time series, but also there are some opposite examples such as negative SOI correlating with high amplitudes (e.g., 1997). Figure 7b shows that wavelet coherence between PC1 time series and SOI has significant, in-phase relationship in the 3-to-7-year band during 1995–2015, also it is significant and in-phase in the 10-to-16-year band during 1980–2016. Thus, there is a strong coherence between SOI and PC1 on timescales from inter-annual to decadal.
As the South China Sea is a semi-closed ocean basin that exchanges water with adjacent waters, including the East China Sea, the Sulu Sea, the Java Sea and the Pacific Ocean, the adjacent seas and the Pacific Ocean must play an essential role in its sea level variability, on not only seasonal timescale, but also on inter-annual timescale and longer timescales. At the developing stage of El Niño, there is increased Luzon Strait transport flowing into the South China Sea, while higher temperature water flows out of the South China Sea through the Mindoro Strait and the Balabec Strait (Qu et al., 2000; Qu et al., 2004; Rong et al., 2007) which cools the South China Sea and even results in lower sea level, but at the developing stage of La Niña, the situation is reversed. The southwesterly monsoon also plays a role via its impact on ocean circulation. In summer, when at the developing stage of El Niño, the Western Pacific subtropical high leads to cyclonic wind anomalies, which drive water divergence in the South China Sea, cooling it and then thinning the mixed layer, thus less heat is stored lowering sea level, Still, at the developing stage of La Niña, the situation is reversed, warm water convergence deepens the mixed layer, raising sea level (Huang et al., 2004; Rong et al., 2007).
3.2 Sea level variability responses to PDO
The above studies revealed that sea level variations in Region 1 and Region 2 are affected by the changes in the Pacific Ocean on decadal scale, that is, PDO. PDO can also be regarded as a response to ENSO and atmospheric noise (Schneider and Cornuelle, 2005; Newman et al., 2003), and they are closely correlated in the low frequency band (Zhang and church, 2012). PDO can affect the intensity and frequency of ENSO. In this analysis, wavelet coherence method is adopted to analyze the correlation between PDO and ENSO. The results are shown in Fig. 8, in which Fig. 8a is the comparison between the time series of SOI and PDO, and Fig. 8b is the wavelet coherence of the two time series. It can be found in Fig. 8a that there is an obvious anti-correlation between the time series of SOI and PDO, positive SOI generally correlating with negative PDO. Figure 8b shows that the SOI time series and PDO time series have strong coherence on multiple time scales. The two time series are highly correlated from 1980 to 1990 in the two-year band with SOI leading PDO, which means, SOI may drive PDO. Also, the two series have strong anti-phase coherence in the band of 4-to-16-year from 1980 to 2016. Overall, PDO is closely related to or affected by SOI on multiple time scales.
The wavelet coherence between the PC1 time series of Region 1 and PDO is shown in Fig. 9a, and that between the PC1 time series of Region 2 and PDO is shown in Fig. 9.b. It can be seen in Fig. 9a that, the wavelet coherence between PC1 time series of Region 1 and PDO has strong anti-phase coherence in the 8-to-16-year band from 1980 to 2016. When PDO is positive, relatively low sea level is found in the East China Sea, the Yellow Sea, the Bohai Sea, and adjacent areas including Japan and Korea. When PDO is negative, high sea level is found there. During a positive PDO phase, the Aleutian low pressure becomes deeper and moves southward, while the westerlies in the North Pacific strengthen, resulting in the increasement of Ekman transport to the South, which leads to the reduction of sea surface temperature in the mid-latitude Pacific, so the steric sea level and sea surface height are relatively low and vice versa. The influence of PDO on the sea level variations in the Northwest Pacific is also considered to be related with the Kuroshio geostrophic transport (Gordon and Giulivi, 2004). The Kuroshio geostrophic transport affects the sea level difference between the Japan/East Sea and the subtropical North Pacific. Comparison between the Kuroshio and PDO reveal that when the transport is large, the sea level slope across both sides of the Kuroshio is large, which corresponds to the positive phase of PDO; When the transport is small, the sea level slope across both sides of the Kuroshio is also relatively small, which corresponds to the negative phase of PDO. In addition, PDO influencing the sea level variations in the East China Sea and Yellow Sea is also considered to be related to the runoff of the Yangtze River (Han and Huang, 2008). Since the Yangtze River is the main source of freshwater runoff to the East China Sea and Yellow Sea, the changes of runoff can significantly affect the distribution of seawater salinity in this region, and then affect the steric sea level variations.
The wavelet coherence between the PC1 time series of Region 2 and PDO shows strong anti-phase coherence in the 8-to-16-year band from 1980 to 2016, as well as strong anti-phase coherence in the 3-to-7-year band during the period of 2000–2016. When PDO is in the positive phase, sea level in the South China Sea is relatively low, and the negative phase of PDO corresponds to the high sea level. This may be due to the strengthening of easterly trade related to PDO (Merrifield et al., 2012; Cheng et al., 2016). In PDO negative-phase years, the intensified easterly trade wind and negative wind stress curl deepen the thermocline thickness of the western tropical Pacific, so the steric sea level and total sea level are at a high level, and vice versa.