To unravel the long-term variability of turbidity fronts and their response to local processes and ocean warming, pertinent satellite ocean color, reanalysis, and climate data were obtained and meticulously analyzed. Specifically, we utilized GOCI-derived Level-2 SSC data, released by the Korea Ocean Satellite Center, to extract turbidity fronts (http://kosc.kiost.ac.kr). Unless otherwise noted, the time coverage of the remaining datasets utilized in this study aligns with that of the SSC dataset. Wind speed, significant wave height, and net heat flux data were extracted from the monthly ECMWF ERA5 reanalysis data with a spatial resolution of 1/4°×1/4° (https://www.ecmwf.int). Daily SST data were sourced from the National Aeronautics and Space Administration Jet Propulsion Laboratory (https://podaac.jpl.nasa.gov). This dataset offers global gap-free gridded SST data with a spatial resolution of about 1 km, which is composed of multiple secondary satellite SST data collections (Chin et al., 2017). The SST data were further employed to calculate SST gradients. Monthly mean reanalysis products of mixing layer depth spanning April 2011-December 2020 were procured from the Copernicus Ocean Environmental Monitoring Service (https://data.ocean.copernicus.eu). This product has a spatial resolution of 1/12° and is vertically divided into 50 layers (Lellouche et al., 2018). Previous studies validated the applicability of this data product in the Eastern China Seas (Huang et al., 2016; Liu et al., 2023; Wu et al., 2019). Monthly water and sediment discharge data of the Changjiang River at Datong Station were derived from the China River Sediment Bulletin (http://www.mwr.gov.cn). Monthly averaged multivariate ENSO index (MEI) and Pacific decadal oscillation (PDO) index for April 2010-March 2022 were provided by the National Oceanic and Atmospheric Administration (https://psl.noaa.gov). Monthly sea surface partial pressure of CO₂ (pCO₂) from April 2011 to December 2019, with a spatial resolution of 4 km, was obtained from MODIS-Aqua satellite data. These data were derived by using a pCO₂ algorithm combining semi-mechanical and machine-learning methods (https://doi.org/10.5281/zenodo.7701112). A detailed description of the pCO₂ data and algorithm can be found in Yu et al. (2023). The bathymetry data were released by the British Oceanographic Data Center (https://www.gebco.net).

The multiscale variations of turbidity fronts in coastal and estuarine waters are highly complex, being affected to different degrees by multiple factors (such as wind, wave, river discharge, mixing layer depth, etc.) (Du et al., 2023b; Framinan and Brown, 1996; Maciel et al., 2021; Wei et al., 2022). However, the influence of these factors on the interannual variation of turbidity front in the Eastern China Seas has not been well quantified. In this study, we further identified that certain factors play a more significant role than others across different regions ( Table 1 ). Meanwhile, these factors were intercorrelated and each factor could explain the variability of frontal activity to some extent.

Specifically, the Bohai Sea is a semi-enclosed bay with relatively shallow water depths (18 m on average) compared to the other two sea areas, and it has an abundant supply of terrigenous sediments. For example, most terrestrial sediment (about 70%) from the Yellow River is deposited on the onshore delta within 15 km of the river mouth, and part of the remaining sediment is transported to the North Yellow Sea through the Bohai Strait (Bi et al., 2011; Li et al., 2022). Consequently, frontal activity in the Bohai Sea is strongest almost all year round due to shallow water and abundant sediment supply ( Figure 4B ). Meanwhile, the variability of frontal activity in the Bohai Sea is highly related to the wind-driven dynamic processes. The frontal probability anomaly is positively correlated with anomalies in wind speed, significant wave height, and mixing layer depth, with the strongest correlation observed with significant wave height anomaly (r=0.34, p<0.01). These correlations indicate that frontal dynamics are most sensitive to changes in wind wave intensity. The Bohai Sea suffers from cold waves and winter storms frequently (Mo et al., 2016; Wang et al., 2020), and the resulting strong wind waves can cause more vertical mixing and deepen the mixing layer depth. Consequently, drastic erosion and re-suspension of seabed sediment occur in most areas of the Bohai Sea, thus enhancing frontal activity (Du et al., 2023b; Wang et al., 2014).

For the East China Sea, the interannual variation of the frontal probability anomaly is mainly related to the mixing layer depth anomaly with the strongest correlation (r=0.41, p<0.01). It is also observed that the frontal probability anomaly is positively correlated with the anomalies of wind speed, significant wave height, SST gradient, and turbulent kinetic energy (p<0.05). These relationships indicate that the underlying mechanisms affecting the interannual variations of turbidity fronts in the East China Sea are more complex. First, the variation of mixing layer depth in this region is influenced not only by the wind field and sea temperature but also strongly influenced by the Changjiang diluted water (Wu et al., 2014). Their combined effect leads to the strongest response of frontal activity to the variation of mixing layer depth. Changes in the wind field will also stimulate the synchronous response of wave and turbulent kinetic energy (Liu et al., 2023). The enhanced (weakened) wind stress will increase (weaken) the turbulent kinetic energy and wave intensity (Dong et al., 2011), thus causing the fluctuation of frontal activity intensity by affecting the stability of the water column and vertical mixing. Meanwhile, the enhancement of the lateral buoyancy (temperature) gradient is conducive to triggering coastal frontal instability and mixed layer baroclinic instability (Shi et al., 2019; Yin and Huang, 2016), thereby controlling the offshore transport of coastal suspended sediment, resulting in a corresponding response in the activity intensity of offshore turbidity fronts. The coupling of shelf circulation and wind field variability also affects frontal activity. The strength and direction of coastal currents and shelf circulations in the Eastern China Seas are significantly modulated by monsoon. Under the action of winter storms, there are shelf circulations and coastal currents in Regions A, B, and C (as shown in Figure 1D ) with opposite flow directions ( Figure 1A ). The resulting shear instability of ocean currents (Liu et al., 2018) is conducive to the formation of offshore turbidity front.

In the Yellow Sea, there is a significant positive correlation between the frontal probability anomalies and SST gradient anomalies (p<0.05), but not significant with other factors. This allows us to establish a link between the interannual variability of the two. This correlation shows that the increase of the horizontal SST gradient will promote the horizontal density gradient, and the resulting strong horizontal pressure gradient is more likely to cause frontal instability (Shi et al., 2023a, 2023b), thus enhancing the frontal activity. Vice versa, a decreased SST gradient appears with weakened frontal activity. For example, large temperature (and density) differences in Regions A and B ( Figure 5A ) lead to considerable baroclinic gradient forces (Li et al., 2019), which in turn enhance frontal activity by stimulating baroclinic instability. Increased frontal activity causes more terrigenous materials to converge offshore, which are then transported to the open sea by tides, waves, and ocean currents, stimulating plankton growth. In summary, these results reveal that the variation of frontal activity at the interannual scale in the Yellow Sea is dominated by density gradient changes resulting from horizontal temperature differences.

According to the ocean environmental conditions, if the horizontal temperature difference between cold coastal waters and warm offshore waters increases in winter and spring, the horizontal density gradient of the seawater will increase accordingly. Especially for the Yellow Sea, the variations in water density are mainly modulated by the changes in water temperature (Shi et al., 2023a). This is because nearshore waters are driven by a full-depth vertical mix of intense winter cooling and winter storms, and their SST is much lower than that of offshore waters. Under the influence of ocean warming, a larger horizontal gradient will stimulate stronger frontal instability, resulting in increased frontal activity in regions with enhanced SST gradients, and more terrigenous materials will be transported offshore. Recent studies have confirmed that climate warming has increased the horizontal temperature gradient in the coastal waters of the Yellow and East China Seas (Shi et al., 2023a, 2023b). Similar climate change-driven enhancements of temperature gradients have also been identified in other regions. For example, the cooling of the Malvinas Current near the Patagonian Shelf Break Front and the simultaneous warming of adjacent shelf waters led to a significant increase in the cross-shelf SST gradient along the slope (Franco et al., 2022). Then, can the long-term patterns or trends of turbidity fronts in the Yellow Sea be attributed to long-term changes in SST gradients?

To solve this problem, we further compared the long-term trend of frontal probability ( Figure 1D ), the 10-year mean SST gradient ( Figure 5A ), and the long-term trend of SST gradient ( Figure 5B ). First, the significant enhancement of frontal activity needs to occur in regions with sufficiently large horizontal pressure gradients, as demonstrated by the close agreement between the distributions of the high SST horizontal gradients (>0.04 °C/km) and the increasing trends of the frontal probability (>0.3%/year) ( Figures 1D , 5A ) since these regions are more likely to stimulate frontal instability. Linear trend analysis shows a statistically significant increase in frontal probability from 2011 to 2021 (p<0.05) ( Figure 4C ) and this trend is followed by the long-term pattern of SST gradients ( Figure 5D ), as also shown by the good coincidence of significantly increased regions of both frontal probability and SST gradients (e.g., Regions A and B) ( Figures 1D , 5A ). This indicates that the enhancement of the horizontal temperature gradient may dominate the expansion of frontal activity in the Yellow Sea. Meanwhile, we note that the frontal probability and SST gradient in the Zhejiang-Fujian coastal waters (Region C) have a statistically significant increasing trend (p<0.05) and their spatial distributions are well consistent, indicating that the influence mechanism of SST gradient on the long-term trend of frontal activity is also applicable to the Zhejiang-Fujian coast. Interestingly, the above three areas correspond to the five major cross-shelf transport paths (Liu et al., 2023; Ren et al., 2015; Shi et al., 2019; Zhang et al., 2019), which also appears to explain the enhanced cross-shelf transport and the resulting ecological effects in the Eastern China Seas, such as more terrigenous nutrient transport conducive to algal reproduction and sustained expansion (Liu et al., 2023; Qi et al., 2022). In conclusion, the distribution pattern of the interannual enhancement of the offshore frontal activity can be explained by baroclinic instability caused by the enhancement of the density gradient due to the horizontal temperature difference, which provides further evidence for the intrinsic correlation between frontal variability and ocean warming.

When calculating the annual frontal probability anomaly and SST gradient anomaly in the Yellow Sea, we found that the frontal probability anomaly in 2013, 2016, and 2018 appeared not to be dominated by the horizontal temperature gradient anomaly ( Figure 5C ). After the three years were excluded, the correlation was significant (r=0.75, p<0.01), suggesting that factors other than SST gradients may have dominated the interannual variations in frontal activity during the three anomalous years. This could explain the significant but weak correlation between the monthly frontal probability anomaly and the SST gradient anomaly. In fact, these factors may originate from the influence of large-scale climate signals such as ENSO. For example, the locations of turbidity fronts in the Rio de la Plata estuarine system shift by a greater distance in La Niña years than in El Niño years, primarily due to abnormal river outflow during ENSO events (Nagy et al., 2008). Saldias et al. (2016) found that turbidity plume events in south-central Chile tend to occur during the warm phase of the ENSO and PDO, along with the opposite phase of the Antarctic Oscillation when the storm tracks shift northward. In addition, ENSO can lead to interannual anomalies of turbidity fronts by causing wind field and SST anomalies (Du et al., 2023a). In El Niño years, weakened winter winds and the warmer SSTs in the Yellow and East China Seas strengthen stratification, greatly inhibiting frontal activity. In contrast, frontal activity increases significantly during La Niña years. Our further analysis revealed that ENSO events had occurred exactly before these three years ( Figure 6A ), suggesting that large-scale climate signals largely determined the frontal dynamics of these years. A significant positive correlation is found between the frontal probability anomaly and the significant wave height anomaly ( Figure 7 ; r=0.52, p<0.01), revealing that the frontal activity anomaly during the anomalous year was mainly driven by the wave intensity anomaly dominated by ENSO events. The close relationship between sea surface wind speed anomalies and frontal probability anomalies further indicated that wave intensity anomaly was dominated by changes in sea surface wind speed (r=0.4, p=0.02) ( Figure 7 and Supplementary Table S3 ).

The lead-lag correlation between MEI and frontal probability anomalies was further calculated to evaluate the influence of large-scale climate events ( Figure 6B ). For the Yellow and the Bohai Seas, the maximum correlation between MEI and frontal probability anomalies was reached when the time lag was 6 months and 11 months, respectively. In the East China Sea, no significant correlation was found between MEI and frontal probability anomalies with a time lag of 0-12 months, mainly because the frontal activity was also affected by water stratification caused by river discharge (Li et al., 2021; Luo et al., 2017). This was supported by a significant negative correlation between frontal probability anomalies and river discharge anomalies over the expansion paths of Changjiang diluted water (p<0.05) ( Figure 6D ). However, no significant correlation was found between the frontal probability anomaly and the river sediment discharge anomaly ( Figure 6E ), because the river sediment usually settles rapidly in the estuary, and its direct impact is limited to waters inside the 30-m isobath (Du et al., 2023a; Li et al., 2021; Zhou et al., 2020). According to previous studies, when we only focused on this local area along the Zhejiang-Fujian coast, the interference of river discharge from the Changjiang River could be partially eliminated. At this time, a significant negative correlation is identified between the MEI and frontal probability with a time lag of 4 months showing the opposite correlation with the other two sea areas (Du et al., 2022), which was mainly related to the regional differences of wind speed anomalies driven by ENSO (Du et al., 2023b; Xu et al., 2022). Meanwhile, these correlations with different time delays were due to the delayed wind response to ENSO events, which had reached a general consensus in the East and South China Seas (Guo et al., 2023; Lu et al., 2022; Zhao et al., 2010). We also calculated the lead-lag correlation between the PDO index and frontal probability ( Figure 6C ). For the Yellow and Bohai Seas, the maximum correlation was reached when the time lag was 5 months and 11 months respectively. The length of the current time series is insufficient to explain in detail the causal relationship between the decadal variation of frontal activity and PDO. However, the in-phase combination of ENSO and PDO ( Figure 6A ) can induce more intense changes in monsoons (Kim et al., 2014), which in turn causes significant anomalies in frontal activity (Du et al., 2022).

Compared to previous studies, this study is distinguished by two key aspects. First, we expand the spatial scale of case studies to the continental shelf. This broader scope not only enhances our understanding of the long-term variability of turbidity fronts but also offers a unique perspective for observing and analyzing regional heterogeneity. By adopting this refined dataset, we conduct a comprehensive and consistent comparison, analyzing the characteristics of frontal changes across three sea areas. This approach can provide a baseline for revealing the degree and mechanisms of ecosystem responses to turbidity front changes in different regions. For example, what are the subsequent effects of abnormal frontal activity on local phytoplankton blooms, red tides, and hypoxia events? Future study is expected to address this issue by combining ecological simulation models and multi-source observation data. We further identify differences in the controlling mechanisms of turbidity front variations across various sea areas, which can provide a scientific basis for developing regional turbidity front simulation and prediction models. For example, optimal driver variable inputs can be provided for data-driven machine learning models. Secondly, this study establishes an intrinsic connection between recent rapid ocean warming and the interannual trends of turbidity fronts, detailing how frontal activity in different sea areas responds to ocean warming. This insight allows us to more effectively explain and predict how marine ecosystems respond to variations in turbidity fronts, and it is expected to provide scientific support for the formulation and implementation of relevant policies. Through such integrative and comparative research, we can better address potential environmental changes and ecological challenges that may arise in the future.

For example, quantifying the intensity of turbidity fronts and understanding their long-term variations are critical to predicting the response of the ocean carbon cycle to climate warming. Studies have shown that the fluxes of sinking organic carbon in oceanic frontal systems are roughly twice as large as those in surrounding waters or non-frontal conditions (Stukel et al., 2017). Therefore, the turbidity front has the great potential to regulate ocean carbon sequestration (Pan et al., 2020). We calculated the correlation coefficients between the monthly time series of frontal probability and pCO₂ anomalies, and found that they showed significant positive correlations in most regions (p<0.05) ( Figure 8A ). This finding reveals that there is a strong coupling between frontal activity and pCO₂ variability and that the interannual anomaly in pCO₂ is largely modulated by offshore transport of terrigenous carbon controlled by turbidity fronts. In addition, the temperature gradient of the entire Eastern China Seas has shown an increasing trend in the last 10 years ( Figure 5B ), indicating that this phenomenon is a general process of the marginal sea driven by global warming. Consequently, enhanced frontal activity in the Yellow and East China Seas may have a positive impact on offshore carbon sequestration and ocean primary productivity. We find that the regions with significantly enhanced frontal probability in the offshore waters of the Eastern China Seas (especially the five cross-shelf transport paths) coincide well with the regions with significantly enhanced seawater pCO₂ ( Figures 1D , 8B ), further supporting the large carbon sink potential of the Eastern China Seas (Yu et al., 2023). This consistency suggests that more CO₂ may be transferred from shallow coastal seas to adjacent deep oceans in the future. Meanwhile, the release of subsurface carbon back into the atmosphere is expected to decrease due to increased water depth and reduced vertical mixing, thereby enhancing ocean carbon sequestration.

Ocean warming is a universal process, and the Eastern China Seas are particularly affected by it. According to the findings presented in our study, if this increasing trend of SST in the Yellow and East China Seas continues in the future, frontal activity on the continental shelf could be further enhanced, which could provide more terrestrial carbon sources for the ocean food web (Saldias et al., 2016). Additionally, ENSO events are anticipated to become more frequent and extreme in the 21st century due to global warming (Cai et al., 2018; Geng et al., 2023). Their impacts on ocean turbidity fronts may be even more pronounced on a global scale. If these trends persist as projected, the formation, separation, location, and displacement of turbidity fronts are expected to show more frequent and intense changes. What remains to be studied is how the frequencies and amplitudes of different ENSO types affect long-term trends of turbidity fronts in global marginal seas, and how PDO affects their decadal variability. This requires extending the time series and expanding the study target area by incorporating other high-resolution satellite products such as MODIS, GOCI-II, Himawari-8, etc. In addition, future studies should study the coupling of turbidity fronts with other physical, chemical, and biological fronts (such as temperature fronts, salinity fronts, and chlorophyll fronts), which is crucial for accurately locating enhanced bioproductivity and ecotones (Chen, 2009; Jamil et al., 2023; Kahru et al., 2018; Miller, 2004; Zhang and Hu, 2021). These should be the subjects of future studies.