Section C-C is located in the northeast of the Changjiang Estuary. As shown in Figure 1 , the summer hydrographic features on the western side of the section are strongly influenced by the northeastward extension of the CDW and northward intrusion of the TWC (Zou et al., 2001; Wei et al., 2015). The central shore is the main exchange area of the YSCC from the Yellow Sea to the ECS (Liu et al., 2021). The east side of the section is at the southern edge of the Yellow Sea Cold Water and near the ECS cold eddy (Hu, 1994; Wei et al., 2013). The bottom topography of the study region includes the shallow Changjiang Estuary (station S1), the wide-range Changjiang Bank (stations S2–S5), and the sharp ECS continental slope (stations S6–S7).

The repeated hydrographic data along Section C-C are observed by the Ningbo Marine Environment Monitoring Center, Ministry of Natural Resources (MNR) ( Figure 1 ). The hydrographic data, namely, temperature, salinity, and DO, were collected semi-annually before 2009 and seasonally after 2010. The stations span an area from 122.5°E to 125.5°E with an interval of ~55 km. The temperature and salinity are measured using a Sea Bird model 911 conductivity–temperature depth (CTD) recorder, and DO is measured on board the vessels immediately after collection using a conventional Winkler titration method (Grasshoff et al., 1999). Water samples are collected at the surface; at depths of 5, 10, 15, 20, 25, 30, 50, and 75 m; and at the bottom (3 m above the sediment) using Niskin bottles. Summer hydrographic data collected from 1997 to 2014 were used in this study. Additionally, apparent oxygen usage (AOU = [O₂]_eq − [O₂], where [O₂]_eq is the DO solubility at equilibrium with the atmosphere, and [O₂] is the in situ DO concentration) is obtained. According to the DO threshold definition in most previous studies, hypoxia status occurs when the DO value is less than 2 mg/L and low-DO status occurs when the DO value is less than 3 mg/L (Diaz, 2001; Zhu et al., 2011).

The repeated hydrographic data along Section C-C used in this study is part of the sectional hydrographic monitoring project in the Chinese marginal seas. Details about the cruise information can be found in Table S2 . The data were quality controlled by the Ningbo Marine Environment Monitoring Center and are widely used in the ECS data assimilation (Zhao et al., 2015).

To quantitatively describe the climatic and hydrographic conditions for hypoxia occurrence off the Changjiang Estuary, we use an ensemble of remote sensing and in-situ observed products, namely, SST, sea-surface wind, and land-based precipitation.

The NOAA 1/4° daily Optimum Interpolation Sea Surface Temperature (or daily OISST) is an analysis constructed by combining observations from different platforms (satellites, ships, and buoys) on a regular global grid (Huang et al., 2021). The monthly mean land-based precipitation is from the Climatic Research Unit (CRU) TS data (version 4). The spatial resolution is 0.5° by 0.5°, and the temporal range is from 1997 to 2014 (Harris et al., 2020). The cross-calibrated multiplatform (CCMP) wind product that covers the global ocean at 1/4° resolution from 1997 to 2014 (Atlas et al., 2011) is used here to illustrate the wind variation. The dataset combines all available satellite surface wind observations with conventional ship and buoy data via an enhanced variational analysis method.

These three climatic data have been widely applied to the ECS to detect seasonal and interannual signals such as marine hot waves (Yan et al., 2020), total rainfall accumulation (Zhou et al., 2008), and the summer monsoon (Sun and Yan, 2012).

The Oceanic Niño Index (ONI), which is the de-facto standard for identifying El Niño and La Niña events in the tropical Pacific, is used in this paper. It is the 3-month running mean SST anomaly for the Niño 3.4 region (i.e., 120°–170°W, 5°N–5°S). Events are defined as consecutive overlapping 3-month periods at or above the +0.5°C anomaly for warm (El Niño) events and at or below the −0.5°C anomaly for cold (La Niña) events. During the period of 1997–2014, there were five El Niño events (97/98, 02/03, 04/05, 06/07, and 09/10) and eight La Niña events (98/99, 99/00, 00/01, 05/06, 07/08, 08/09, 10/11, and 11/12).

The correlation analysis is used to evaluate the ENSO effect on coastal hypoxia and related hydrographic and climatic conditions off the Changjiang Estuary. The correlation is significant at a 95% confidence level if not specified.

The linear trend of hydrographic and chemical parameters (such as temperature, salinity, and DO) during the period is calculated based on the least squares formula (Wilks, 1995). A linear trend is significant when it passes the Student’s t-test at a 95% confidence level.

The climatological mean fields of temperature (T), salinity (S), and seawater density are shown in Figures 2A–C . The warm and low-salinity CDW (S <30, Su et al., 1996) is extended eastward from the river mouth to 125.5°E, with a mean water mass thickness of 10 m. The CDW breaks off at station S3 in association with the northeastward-southeastward diversion of the CDW offshore spreading. The thermocline beneath the CDW slopes downward to the east from 10 to 20 m along Section C-C. Two key water masses, the TWC and ECS cold eddy, occupy the bottom layer. The downwelling process can be detected near 123.5°E, which sinks the cold and DO-rich surface water into the subsurface layer. The pattern of density distribution is similar to that of salinity distribution, indicating the important role of salinity in the density pattern off the Changjiang Estuary (Zhu et al., 2016).

Figures 2D–F show the climatological mean and linear trend of DO, AOU, and DO solubility along Section C-C. The distributions of DO and AOU are extremely similar. DO concentrations are relatively high in the upper ocean, forming a relatively strong oxycline over the benthic low-DO zone. The most notable characteristic is the presence of two low-DO zones (<4 mg/L), at 122.5°–123°E and 124°–124.5°E, respectively. Generally speaking, AOU is relatively high (>3.5 mg/L) in these two low-DO zones. A notable sinking of high-DO waters associated with the downwelling process is observed near 123.5°E in the central shoal of the section, which separates the bottom low-DO water into two components: the inner area nearshore and the outer area in the offshore Changjiang Bank region. In contrast to DO and AOU, the DO solubility presents a similar pattern as that of seawater density, with an amplitude range of 6.7–8.5 mg/L.

The presence of dual-core low-DO zones is related to two hypoxia zones that frequently happen off the Changjiang Estuary. Statistically, there were three typical patterns of the hypoxic zones during 1997–2014. The nearshore core pattern, offshore core pattern, and dual-core pattern account for 33, 23, and 44%, respectively (see Table S3 ). These nearshore and offshore hypoxia zones can also be found in the statistical results of other studies (Wang, 2009; Wei et al., 2017). There are two possible dynamic mechanisms for the dual-core pattern of low-DO benthic water: First, the regional current system contains a large anticyclonic circulation and several cyclonic eddies across the Changjiang Bank. So, fresh water and nutrients are carried northeast first and southeast afterward, along with the CDW extension (Xuan et al., 2012). The terrestrial nutrient supply is less at station S3, compared with nearby stations. Second, the regional anticyclonic circulation and downwelling processes, can weaken local stratification (Zou et al., 2001; Wei et al., 2013) and are unfavorable for bottom hypoxia.

Figure 3 shows the time series of DO at two low-oxygen cores. Generally speaking, the DO is highly dynamic with an insignificant linear trend for the whole water column. Compared with the offshore region, the nearshore DO presents a larger interannual variation, indicating the influence of complex coastal processes such as terrestrial nutrient loading and tidal modulation (Zhu et al., 2017). For the bottom layer, the DO time series between nearshore and offshore cores is similar to the correlation coefficient of 0.70 (p <0.01). Five hypoxia events are found in the nearshore region, which happened in August of 1998, 2003, 2007, 2010, and 2014, respectively. One hypoxia happened offshore in August 1998.

The linear trends of bottom DO present different patterns for dual low-DO cores, with a slightly decreasing trend for nearshore cores and an insignificant trend for offshore cores ( Figures 2 , 3 ). Despite the insignificant increasing trend of the offshore surface DO, our results are in good agreement with that which occurred along the 32°N section presented by Ning et al. (2011). It means that the hypoxia events off the Changjiang Estuary do indeed became quite severe. For the nearshore core (122.5°–123.5°E), the stratification (simply represented by Δσ, bottom density minus surface density) decreased during 1997–2014 and was unfavorable for hypoxia occurrence. Meanwhile, the DO saturation increased slightly for bottom water because of bottom cooling for coastal water during 1997–2014. So, the decreasing trend of bottom DO during 1997–2014 was due to larger AOU rather than to DO saturation or stratification variation.

Figure 4 illustrates the low-oxygen and hypoxia extents observed at Section C-C based on several indices: DO minimum and longitude range of low-oxygen/hypoxia extent. Results show that DO minimum is anti-correlated with the low-oxygen/hypoxia area, with correlation coefficients of −0.64 (n = 12) and −0.67 (n = 6), respectively. Similar to DO value, these indices present no significant trends during 1997–2014. Ning et al. (2011) reported that the hypoxia area along 32°N had an eastward spreading trend during 1975–1995. We also calculate the linear trend of the eastern boundary longitude of the hypoxic area along Section C-C, but no significant trend was observed during 1997–2014 (figure not shown).

To describe the ENSO effect on DO variation, we categorize the bottom DO status by using different ENSO phases. As shown in Table 1 , the bottom hypoxia occurred in six years during 1997–2014, four of which happened in El Niño decaying summers and none happened in La Niña decaying summers. For the whole 18-year record, the low-oxygen and hypoxia probabilities are 33 and 56%, respectively. These probabilities increase to 80 and 100% for El Niño decaying summers, and decrease to 0 and 25% for La Niña decaying summers.

Figure 5 shows the correlation coefficient between DO and ONI along Section C-C. Results show that the subsurface DO value within the longitude range of 122.5°E to 124.5°E is negatively correlated with the ONI value, which is significant at a 95% confidence level. The correlation coefficient presents two local maxima (<−0.6) in deep water, which generally matches the “dual-cores” structure of low-DO areas along Section C-C.

Figure 6 shows the composite distributions of anomalies (differences to the climatological mean) of hypoxia-associated hydrographic parameters (DO, temperature, and salinity) along Section C-C during different ENSO phases. In El Niño decaying summers, a negative DO anomaly is present in the subsurface layer at stations S1-S5, with a minimum of −1.6 mg/L in the bottom layer of station S3. The DO anomaly is relatively smaller at stations S6 and S7, indicating a weaker ENSO influence in the open seas near Cheju Island. The contrary is seen in the La Niña years. A positive DO anomaly occurs in the bottom layer at stations S1–S4, meaning lower coastal hypoxia probability for La Niña years.

The temperature and salinity anomaly fields present similar patterns to those of the DO anomaly field. In El Niño decaying summers, fresher and warmer CDW water and saline and cooler subsurface TWC water are combined to produce much stronger vertical stratification, which is favorable for bottom hypoxia development. The situation in the La Niña years is the opposite. However, the anomalies for all three parameters are relatively smaller compared to those in El Niño years.

For the following summer of El Niño years, anomalously anticyclonic circulation occurs to the east of the Philippine Islands due to changes in the strength and location of the western North Pacific subtropical high. It has been found to be a key factor in generating variability in precipitation in southern China (Zhou and Yu, 2005; Xie et al., 2009). The difference in the composited wind field between the El Niño years and La Niña years confirms that the anticyclonic anomaly of moisture transport is enhanced in the El Niño decaying summer, indicating that the southwesterly moisture transport from the South China Sea increases precipitation in southern China ( Figure 7 ). As described in Park et al. (2015), the July inland precipitation in China is a primary contributor to freshening in the Yellow Sea and ECS through increasing Changjiang River discharge during El Niño decaying August. Meanwhile, the surface wind anomaly is northward-northeastward over the ECS continental shelf during July–August ( Figures 7 , 8 ) and promotes the Kuroshio intrusion across the ECS shelf via the Ekman effect (Tan and Cai, 2018), which also contributes to the enhanced oceanic nutrient and hypoxia development off the Changjiang Estuary. The contrary appears in the La Niña years.

Figure 8 shows the composite differences between the mean and standard deviation of the SST and wind fields between El Niño and La Niña decaying August. The atmospheric and hydrographic conditions are rather calmer during El Niño decaying August. The standard deviations of both SST and wind speed present negative anomalies in hypoxia areas, providing favorable conditions for maintaining the stratification and sustaining oxygen decomposition. Besides, the differences in mean of SST and wind between El Niño and La Niña decaying August is relatively small, with the mean value of 0.6°C and 0.1 m/s for hypoxia area ( Table 2 ), indicating the standard deviations for both SST and wind speed, rather than their mean values, contribute to coastal hypoxia off the Changjiang Estuary during El Niño decaying August.

In the ECS, the stratification is often disrupted by strong winds, such as during a typhoon (Chen et al., 2014; Ni et al., 2016). The modeling study confirmed that typhoon activities temporarily broke down pycnocline, and eventually terminated the hypoxia in 2006 (Zhou et al., 2017). However, according to the western Pacific typhoon records from the Joint Typhoon Warning Center (JTWC), little typhoon activity occurred in the whole ECS during El Niño decaying summers ( Figure 9 ). Li and Zhou (2012) showed that during the transition period from El Niño to La Niña, typhoon activity across the marginal seas of China is less frequently associated with changes in the relative vorticity of air and vertical wind shear. Our findings corroborate their findings. The typhoon-absent climatic condition in El Niño decaying summers helps provide the negative standard deviation (STD) anomalies of SST and wind fields in the ECS ( Figures 8C, E ), which benefit stratification maintenance and sustain oxygen decomposition.

It is difficult to directly compare the areal extent and severity of hypoxia events between previous studies and our observations due to different sampling stations and cruise periods used. The following comparison of DO minimum and hypoxia probability is based on statistics.

(a) Oxygen minimum

The time evolution of DO minimum shows consistency between our observations and previous studies to a certain extent, with the root-mean-square (RMS) residual of 1.09 mg/L ( Figure 4 ). Relatively large differences mainly occur in the following years: 1999, 2003, 2006, 2008, 2010, and 2014.

(1) No hypoxia was detected along Section C-C; however, it was reported in previous studies regarding 1999, 2006, and 2008. A possible reason is that our sampling stations and cruise period are not appropriate for the hypoxia, meaning the inter-seasonal variation of DO is also important and significant.

(2) Hypoxia was detected along Section C-C; however, no hypoxia was reported in previous studies regarding 2003, 2010, and 2014. In 2003, the sampling area focused on the outer ECS shelf but did not cover the Section C-C. In 2010, the sampling period of Liu et al. (2012) was September when hypoxia had probably weakened. In 2014, the sample stations were too sparse (Zhou et al., 2020).

(b) Hypoxia probability

The hypoxia occurrence probability is more than 67% in the literature ( Table S1 ). However, the probability is only 33% along Section C-C. Note that oceanographers usually conduct their field investigations for hypoxia detection, so the hypoxia status is probably overestimated in the literature. For example, there were no hypoxia reported in the summers of 2000 and 2001. Maybe there were field investigations but no hypoxia was found in those years, just as our results show in Table 1 . Another reason for the lower hypoxia occurrence in our study is that the hypoxia does not always extend northward enough to be detected along Section C-C. In this sense, the hypoxia/low-DO probability is probably underestimated in this study.

From the global perspective, the influence of ENSO on coastal ecological systems is significant in two different types of hydrodynamic systems: wind-induced upwelling areas (such as the eastern Indian Ocean in Gregg, 2001, and the California coast in Nam et al., 2011) and river-dominating estuarine areas (such as the Congo River and Amazon River; Signorini et al., 1999). The hypoxia area off the Changjiang Estuary belongs to the latter. As mentioned above, during most of El Niño decaying summers, the anomalous atmospheric circulation in the western Pacific transports moisture southwesterly and enhances the inland precipitation in southern China (Zhou and Yu, 2005). Park et al. (2015) reported that the Changjiang River discharge in El Niño decaying July is ~0.006 Sv greater than that for La Niña. Our result also confirms the Changjiang River discharge difference is ~0.01 Sv between different ENSO phases ( Table 2 ). The greater amount of fresh water supply in July is responsible for the greater extent of CDW spread in August. Compared with the climatological mean, the CDW range extends eastward ~90 km in El Niño decaying August and shrinks westward ~20 km in La Niña decaying August. Considering costal hypoxia is sensitive to the CDW inter-seasonal variation (Zhang et al., 2018), the larger CDW range can help sustain the vertical stratification and oxygen depletion off the Changjiang Estuary. Besides the terrestrial fresh water supply, another important ENSO effect on coastal hypoxia is providing calmer climatic and hydrographic conditions. Statistical results indicate the typhoon frequency decreases by 50% during El Niño decaying summer, and the standard deviations for wind and SST decrease by 15 and 28%, respectively ( Table 2 ). The calm wind and SST fields also help provide favorable conditions for hypoxia development.

In the past three decades, there have been two very strong El Niño events, which happened in 1997/98 and 2015/16, respectively. It is not coincidental that two severe coastal hypoxic events occurred in the summers of 1998 and 2016. Although the hypoxia in the summer of 1998 was not significant in the literature, with a smaller extent (600 km) and a relatively higher DO minimum (~1.44 mg/L) (Wang and Wang, 2007). The hypoxic event was quite severe as observed along Section C-C. Actually, hypoxia is the severest event in the repeated sectional observations during 1997–2014 ( Figure 3 ). The hypoxia water ranged from the river mouth to 124.5°E, and the DO minimum was ~0.51 mg/L. Besides, severe hypoxia (close to anoxia) occurred in summer 2016 off the Changjiang Estuary (Chen et al., 2020; Wang et al., 2021). According to the hypoxic area and DO minimum, the hypoxia is rather severer in summer 2016 than in summer 1998, although its peak ONI is smaller than that of 1998. There were two possible reasons. First, the interannual variation of coastal hypoxia is largely influenced by its inter-seasonal variation (Zhang et al., 2020). The hypoxic area calculated from DO at discrete locations often contains possible errors because hypoxia has short-term variation (Zhang et al., 2018). So the comparison between two severe hypoxia has uncertainties. Second, after the construction of the Three Gorges Dam, the coastal sediment has decreased and the light availability has increased off the Changjiang Estuary. So the current hydrographic condition is much favorable for phytoplankton bloom and subsequent benthic hypoxia. Although the river discharge in summer 1998 was much larger than that in summer 2016, both riverine nitrate and phosphate concentrations in summer 1998 were 30–40% lower than the current levels (see Figure 2 in Wang et al., 2018).

Although the composite differences of DO and associated hydrographic and climatic features are significant between El Niño and La Niña years, the linear regression between the hypoxia severity index and ONI is not strong, especially for non-El Niño years. As shown in Figure 10 , the low-oxygen/hypoxia areas tend to be much wider and the DO value tends to be much lower in El Niño decaying August. For normal and La Niña years, the scatter points are distributed rather discretely. These asymmetric ENSO effects on coastal hypoxia off the Changjiang Estuary are also reflected in the magnitude of composite anomalies of hydrographic factors along Section C-C (such as temperature and salinity; Figure 6 ).

In summary, the ENSO signal is not the prerequisite for the coastal hypoxia off the Changjiang Estuary. However, the physical and chemical parameters associated with El Niño provide favorable conditions for hypoxia occurrence and indeed increased the hypoxia probability during the past two decades. In particular, the two severe hypoxia events, with large hypoxic zones and an unprecedentedly low DO value, occurred during very strong El Niño events (1997/98 and 2015/16). It is notable that the hypoxia was controlled by regional multiple factors from anthropogenic activities and natural processes. The ENSO, the remote low-latitude impact factor, provides important regional impact for the ECS coastal ecosystem.

Regional warming along with large riverine inflow is likely to prolong stratification, reduce vertical mixing, and eventually exacerbate bottom hypoxia in many estuaries, such as the Chesapeake Bay (Boesch, 2008), the Baltic Sea (Bendtsen and Hansen, 2013), and the Gulf of Mexico (Justic et al., 2005). However, at most stations and depths, the seawater temperature in August decreased slightly along Section C-C during 1997–2014 ( Figure 2A ). This result is in agreement with previous satellite observations (Kim et al., 2018). As a result of the cooling trend, the DO saturation has increased, especially for the waters at depth.

Although the ECS warming trend has not been significant during the past two decades, synoptic extreme SST events (marine heat waves) occur frequently and strongly impact the ECS hydrology (Park et al., 2015; Cai et al., 2017). Based on satellite records, Yan et al. (2020) reported three marine heat waves occurred in the summers of 2004, 2006, and 2016. However, no hypoxia event occurred in 2004 and the hypoxia event was not severe in 2006 along Section C-C ( Figure 4 ), indicating large uncertainties in the relationship between coastal hypoxia and extreme ECS warming. The relationship between marine hot waves and 2016 hypoxia will be discussed in a separate paper.

Figure 11 shows the monthly mean and standard deviation of SST in the hypoxia area in August. The monthly mean SST was warmer in 2004, 2006, 2007, 2008, 2010, and 2013, and much cooler in 2009, 2011, and 2014. The SST interannual variation does not coincide with the occurrence of coastal hypoxia along Section C-C. However, the year-to-year variation of the standard deviation of SST is much more similar to the occurrence of hypoxia events. As mentioned in the ENSO Effect on DO Interannual Variation section, a low standard deviation value means that the SST field is relatively calm, providing a favorable condition for bottom hypoxia development. The correlation between SST standard derivation and ONI is −0.55 and significant at the 95% level.