In September 2019, Lingling and Tapah successively passed northward over the eastern ECS ( Figure 1A ), with similar wind directions and comparable paths, affecting the pollutant transport processes off the CJE. The tropical cyclone best track data, including the center positions, 1-min maximum sustained winds, and the radius of maximum wind, were provided at 6-hourly resolution by the U.S. Joint Typhoon Warning Center (JTWC; https://www.metoc.navy.mil/jtwc/jtwc.html).

Lingling formed east of the Philippine Sea and developed into a tropical storm on September 2. Early on September 5, it underwent rapid intensification and soon became a Category 4 (on the Saffir-Simpson Scale) typhoon east of Taiwan, with a maximum wind speed of approximately 61 m/s and a minimum central air pressure of about 928 hPa. Lingling weakened gradually as it traveled through the ECS, eventually making landfall in western North Korea.

Tapah originated east of Taiwan on September 18 and moved northwestward. On September 21, it intensified into a Category 1 typhoon after entering the ECS, reaching a maximum wind speed of approximately 33 m/s and a minimum central air pressure of around 967 hPa. Tapah then weakened rapidly, moving northeastward and fully dissipating in the East Sea (Japan Sea) by September 23.

Data on annual Changjiang discharge from 2000 to 2022 were collected at Datong hydrological station from the River Sediment Bulletin of China, compiled by the Chinese Ministry of Water Resources. Annual reports regarding the water quality of the Zhejiang coastal waters for the same period were extracted from the Ecological and Environmental Status Bulletin and the Marine Environment Bulletin of Zhejiang Province, published by the Ministry of Ecological Environment and the Zhejiang Provincial Bureau of Ocean and Fisheries, respectively.

Two ship-borne surveys were conducted off the CJE and in the Zhejiang coastal waters during August 16–21 (before the typhoon passages) and September 24–27 (after the typhoon passages), 2019, from which salinity and nitrate data were collected for model validation. The August and September surveys took in situ measurements at 33 and 37 stations, respectively ( Figure 1B ). Salinity was measured throughout most of the water column with a bin size of 1 m using an SBE911 plus CTD. Seawater samples were collected from the surface, middle, and bottom layers, and were immediately filtered through 47 mm GF/F glass fiber filters. The nitrate concentrations of the filtered samples were measured using a continuous flow analyzer (San++, Skalar, Breda, Netherlands) with colorimetric methods following Grasshoff et al. (2009).

The coupled physical-biogeochemical model utilized in this study was configured specifically for the CJE and its adjacent seas. The physical module was based on the Regional Ocean Modeling System (ROMS; Shchepetkin and McWilliams, 2005; Haidvogel et al., 2008), which has been broadly employed to investigate estuaries and open seas (Zhou et al. 2009b; Zhou et al., 2015, Zhou et al., 2017a; Xie and Li, 2018; Tian et al., 2022). The biogeochemical module was based on the Carbon, Silicate, and Nitrogen Ecosystem (CoSiNE) model (Chai et al., 2002; Xiu and Chai, 2014). The coupled ROMS-CoSiNE model domain covered the entire Bohai Sea, Yellow Sea, and ECS, spanning from 117.5°E to 132.0°E and from 23.5°N to 41.0°N, with a fine horizontal resolution of 1/24°×1/24° and 30 vertical S-coordinate layers (Song and Haidvogel, 1994). This model was well-calibrated and able to reproduce the seasonal and synoptic variations of the CDW extension and associated nutrient transport (Zhou et al., 2015, 2020; Meng et al., 2022).

The coupled ROMS-CoSiNE model was initialized using climatological mean January values based on multi-year observations (Chen, 1992), and spun up for ten years driven by climatological forcings from the Comprehensive Ocean-Atmosphere Data Set (COADS) monthly fields (Da Silva et al., 1994), including surface wind stress, shortwave radiation, heat, and freshwater fluxes. The climatological run outputs during December served as the initial field for a realistic run from 2018 to 2019. To establish suitable boundary conditions for the realistic run, the daily average outputs from the Hybrid Coordinate Ocean Model (HYCOM) simulations were utilized to specify temperature, salinity, velocity, and sea surface elevation. Fifteen tidal components derived from the TPXO9 global tide model (Egbert and Erofeeva, 2002) were considered at the offshore open boundary, accounting for over 95% of the total tidal variability in the study region. As for surface forcings, the surface heat and freshwater fluxes were obtained from the European Centre for Medium-Range Weather Forecasts latest reanalysis (ERA5) product (Hersbach et al., 2020). The 10-m wind field was extracted from the Cross-Calibrated Multi-Platform (CCMP) wind vector analysis product, at resolutions of 0.25°×0.25° and 6-hourly (Atlas et al., 2011). Considering the underestimation of high typhoon wind speeds in the CCMP dataset (Uhlhorn and Black, 2003; Uhlhorn et al., 2007; Klotz and Uhlhorn, 2014), a blended wind field was constructed for Lingling (due to its strong intensity), which was more representative than that from the CCMP data (see the Supplementary Material for further information on wind field reconstruction). Daily runoff and monthly nutrient concentrations of the Changjiang were specified by measurements at Datong hydrological station. The modeled phytoplankton concentration (mmol N/m³) was converted to chlorophyll-a (Chl-a) concentration (mg/m³) using a conversion factor of 1.58 g Chl-a per mole nitrogen (Cloern et al., 1995). This ratio is frequently employed in modeling studies of the ECS (Liu et al., 2010; Zhou et al., 2017b; Tian et al., 2022). More details on the model setup were elaborated by Zhou et al. (2015), and additional information regarding parameter configuration and model validation in the ECS was provided by Zhou et al. (2017). The realistic run of 2019 was referred to as Exp.R, and the 6-hourly average model outputs were analyzed in this study.

Four sensitivity experiments with different wind forcings were designed to assess the separate and combined impacts of Lingling and Tapah, as well as monsoon winds, on pollutant transport off the CJE ( Table 1 ). Exp.C was the climatological experiment using 6-hourly climatological wind field (2001–2020). In Exp.L, Exp.T and Exp.L+T, the realistic wind fields were replaced with the climatological wind field for the periods of Lingling (September 4–7), Tapah (September 17–22) and these two typhoons, respectively.

Given that this study primarily aimed to investigate the impacts of wind fields on pollutant transport off the CJE, other factors remained consistent across all experiments. Note that the influence of typhoon-induced variations in river runoff was not considered, due to the challenge of differentiating whether alterations in pollutant transport resulted from changes in river runoff or wind field.

To illustrate the differences in wind forcings across these experiments, wind vectors averaged over the region covering the Changjiang outlets and adjacent waters (dashed red box in Figure 1B , hereinafter referred to as region A) were calculated ( Figure 2 ). During the passage of Lingling, the wind speed off the estuary mouth increased from 6.3 m/s to a maximum of 24.7 m/s (northerly) at 12:00 on September 6, and then rapidly decreased to 1.3 m/s, while the wind direction underwent a significant counterclockwise shift within 48 hours. During the passage of Tapah, the wind speed increased from 7.2 m/s to a peak of 16.6 m/s at 12:00 on September 21, with an average speed of 10.6 m/s. The wind speed was relatively low during the monsoon transition in early autumn 2019, with variable wind directions. In Exp.C, the climatological winds showed limited variability, with a monthly average speed of 3.9 m/s and a prevailing direction of northeasterly.

Ocean pollutants (e.g., nitrogen and phosphorus) undergo both physical processes (i.e., advection and diffusion) and biogeochemical processes involving diverse biological variables and environmental components. An Eulerian tracer was utilized to specifically investigate the effects of physical processes on pollutant transport. This tracer, representing conservative pollutants, changes its concentration solely through advection and dispersion (Yang et al., 2012; Du and Shen, 2017; Zhang et al., 2019), in contrast to nitrate, which represents non-conservative pollutants. Given that the Changjiang is the predominant source of terrestrial pollutants for the ECS (Beardsley et al., 1985; Zhang, 1996; Liu et al., 2009), we introduced conservative pollutants via a continuous tracer injection into the downstream area (marked by a red pentagram in Figure 1B ) in every experiment. The initial tracer concentration was set to 1.0 (arbitrary unit) for all vertical layers, while the initial concentration within the model domain was zero. The tracer concentrations were recorded at a 6-hourly interval.

According to the national seawater quality standards of China (GB 3097–1997), seawater quality is classified into five categories, namely clean water (Category I), relatively clean water (Category II), slightly polluted water (Category III), moderately polluted water (Category IV) and heavily polluted water (inferior to Category IV). Among monitored pollutants, DIN stands out as a significant threat to water quality of the Zhejiang coastal waters. The maximum DIN concentration recorded in these waters during 2019 has exceeded the Category IV standard by nearly sixfold, as reported by the Ecological and Environmental Status Bulletin. The permissible limits of DIN for Category I to IV are 0.2, 0.3, 0.4, and 0.5 mg/L, respectively.

Nitrate is generally the predominant form of DIN in estuaries and adjacent coastal waters (Statham, 2012), constituting over 84% in regular seasons and over 95% in flood season along the Changjiang (Wu et al., 2023). Since only two forms of DIN, nitrate (NO₃-N) and ammonium (NH₄-N), were simulated in the model, this study assumed a constant fraction of 90% for nitrate in DIN to assess seawater quality by comparing the modeled nitrate concentrations against the water quality standards.

The comparison between modeled and observed salinity in the Zhejiang coastal waters revealed that, the model could accurately capture the spatial characteristics and effectively reproduce the spreading pattern of the CDW ( Figures 3A–D ). Both modeled and observed salinity demonstrated the presence of low-salinity water (< 28) in the Zhejiang coastal region following Tapah. The root-mean-square error (RMSE) and mean absolute error (MAE) for surface (bottom) salinity were 1.10 and 0.85 (1.13 and 0.84), respectively, with a correlation coefficient of 0.92 (0.66) (p<0.01). Discrepancies mostly occurred in coastal regions with small islands, mainly due to inadequate spatial resolution. For vertical distributions along transects a–c, as indicated in Figure 1B , the RMSE and MAE were 0.84 and 0.65, respectively, with a correlation coefficient of 0.88 (p<0.01) ( Figure 4 ). Evaluations of salinity suggested that our model has a reliable performance in reproducing the behavior of the CDW extension under typhoon conditions.

The modeled surface nitrate concentration agreed reasonably well with field observations off the CJE and in the Zhejiang coastal waters ( Figures 3E–H ). The RMSE and MAE for surface nitrate concentrations during the survey in August (September) were 7.07 and 8.81 (6.25 and 7.12) μmol/L, respectively, with a correlation coefficient of 0.81 (0.92) (p<0.01). Both modeled and observed results exhibited similar spatial patterns, suggesting a satisfactory skill of the model in reproducing the spatial and temporal variations in nitrate concentrations.

Figure 5 shows the spatiotemporal variations in nitrate and tracer concentrations before and after the typhoon passage. Prior to Lingling’s arrival (September 1–3), due to weak seaward momentum, high-concentration nitrate (marked by a 60 μmol/L contour) accumulated near the estuary mouth, forming a bulge shape. This structure was completely disrupted by Lingling, with the majority of nitrate rapidly moving southward, and a southwestward extension of lower concentrations (marked by a 20 μmol/L contour) along the Zhejiang coast (September 8–10).

The nitrate distribution remained largely unchanged until Tapah passed through, where high concentrations reached the Ningbo coastal region and lower concentrations extended throughout the Zhejiang coastal waters (September 24–26). Notably, the average nitrate concentration in the Zhejiang coastal waters (indicated by the dashed green line in Figure 1B ) increased from 18.5 to 26.5 μmol/L during Lingling, and subsequently from 26.3 to 32.7 μmol/L during Tapah. The spatiotemporal variations in the tracer field were broadly consistent with those in the nitrate field ( Figures 5E–H ). The increase in average tracer concentration during Lingling (0.08) was nearly 1.3 times that during Tapah (0.06), paralleling the ratio (1.25) of changes in average nitrate concentration.

If Lingling or Tapah had not passed through this region, where would pollutants off the CJE and adjacent waters have been transported? To what extent would these pollutants have affected the Zhejiang coastal waters? To address these questions, we carried out Exp.L, Exp.T and Exp.L+T to examine the separate and combined impacts of Lingling and Tapah, respectively. The results revealed that, Lingling and Tapah significantly enhanced the southward transport of pollutants off the CJE, adversely affecting the water quality in the Zhejiang coastal region ( Figure 6 ). In the absence of individual typhoon, the average nitrate concentration increased by 2.3 and 3.2 μmol/L during the periods of Lingling and Tapah, respectively. As expected, the sum of these two values closely matched the combined increase of 5.7 μmol/L during the period of the two typhoons in Exp.L+T. Note that the spatial distribution differences in nitrate and tracer between experiments with and without the typhoon exhibited similar patterns in coastal regions west of the 30-m isobath, with correlation coefficients of 0.88 (p<0.01) during Lingling ( Figures 6B, C ) and 0.90 (p<0.01) during Tapah ( Figures 6E, F ). This finding suggested that physical processes could exert a remarkable influence on pollutant transport during short-term extreme weather events like typhoons.

To quantitatively assess the impacts of Lingling and Tapah on pollutant transport off the estuary mouth, we calculated the masses ( M ) of nitrate and tracer across each bordering section of region A during either typhoon event. The masses were determined by integrating their respective fluxes ( F ), the product of concentration and flow velocity, over a specified period:

Here, C represents the concentration of nitrate or tracer, u → is the flow velocity at the cross-section, and n → is a unit normal vector heading inward (positive entering the region). H is the total water depth, L is the section length, and τ is time. The units for nitrate and tracer masses are mol and m³, respectively.

The budget analyses indicated that, Lingling and Tapah significantly affected the north-south pollutant transport off the estuary mouth ( Figure 7 ). Influenced by the typhoons, the net nitrate mass across the southern section during Lingling and Tapah were 5.62 and 3.47 Gmol, respectively, accounting for nearly 6% and 4% of the total nitrate flux from the Changjiang in 2019. The net southward transports during Lingling and Tapah were respectively 4.1 and 7.4 times larger than those in experiments without the corresponding typhoon. These ratios were higher than those obtained from tracer experiments (3.6 and 5.4; Figures 7C, D ), primarily due to greater inputs from the northern section, as nitrate was non-conservative and had multiple sources. Note that the difference in net southward transports between Exp.R and Exp.L+T closely matched the total of those differences derived from Exp.R−Exp.L and Exp.R−Exp.T, suggesting that the combined impact of the two typhoons was approximately equal to the sum of their individual impacts (see Supplementary Figure 2 ). Based on a momentum balance analysis, this significant southward transport was driven predominantly by strong local southward wind stress in the coastal region and a substantial cross-shore sea level gradient due to continuous seawater accumulation in nearshore areas ( Figure 8 ).

To quantitatively describe the pollutant intrusion into the Zhejiang coastal waters, the mean concentrations of nitrate and tracer were estimated, and significant increases were observed during the typhoons ( Figure 9 ). In Exp.R, the average nitrate concentration peaked at 26.6 and 32.8 μmol/L on September 9 (post Lingling) and 24 (post Tapah), respectively. As these two typhoons passed through, the average nitrate concentration gradually decreased. By October 7 (October 9), the mean nitrate concentrations in Exp.R and Exp.L (Exp.T) became identical, suggesting a lasting effect of Lingling (Tapah) on nitrate levels for approximately 34 (23) days. Note that the tracer concentrations took about 10 days longer than nitrate to return from typhoon-affected levels to normal conditions, since nitrate decreased partly due to phytoplankton uptake ( Figure 9B ).

Due to the influence of Lingling, both modeled and GOCI-observed surface Chl-a concentrations off the CJE and adjacent waters showed significant increases ( Figures 10A, B ). The Chl-a concentration during September 11–14 was much higher than that during August 21–24, with the maximum change rate exceeding 3. The temporal evolution of the rate of changes in modeled surface Chl-a concentration showed a consistent trend with that from GOCI-observed results (with spatial coverage of >75%), suggesting that the Chl-a concentration in the Zhejiang coastal waters peaked ~1–2 weeks after the passage of Lingling ( Figure 10C ). A similar time lag between the typhoon event and peak biomass was also revealed in other modeling studies of this region (Zhou et al., 2020). Tapah also induced an increase in Chl-a concentration, though to a lesser extent, probably because phosphate (PO₄), the primary limiting nutrient in the CJE and adjacent waters, was insufficient to support enhanced phytoplankton growth and reproduction (Fan and Song, 2014; Liu et al., 2016). The nitrate/phosphate ratio in the Zhejiang coastal waters during Tapah (82.2) was nearly 2 times larger than that during Lingling (40.6).

In terms of net transport and lasting effects, Lingling exerted a more significant influence on southward pollutant transport compared to Tapah, partly because Lingling approached the CJE in closer proximity with greater intensity. Additionally, Lingling forced pollutants near the estuary mouth southward ahead of Tapah, thus reducing Tapah’s impact to a certain extent. These findings underscore the crucial role of typhoons in facilitating pollutant migration from the Changjiang to the Zhejiang coastal waters.

The water quality off the CJE and its adjacent waters was classified based on nitrate levels, and the results showed degradation in the Zhejiang coastal region during the typhoon passages ( Figure 11 ). The spatial pattern of water quality closely aligned with water quality mapping in autumn based on remote sensing and a deep learning model (Wu et al., 2022). In Exp.R, the percentage of heavily polluted water in the Zhejiang coastal waters increased from 12% to 26% during Lingling and subsequently from 28% to 40% during Tapah ( Table 2 ). In the absence of typhoons, the increases in heavily polluted water were only 6% and 4% during the periods of Lingling and Tapah, respectively. As the typhoons passed through, seawater exceeding the Category II standard (at least slightly polluted) rapidly spread over the Zhejiang coastal region, increasing from 29% during September 1–3 to 61% during September 24–26.

We conducted a climatological experiment (Exp.C) to investigate the impact of monsoon winds on pollutant transport off the CJE and its adjacent waters. The seasonal mean wind fields for Exp.R and Exp.C are depicted in Figure 12 . During spring (March to May) and autumn (September to November), the study region was largely controlled by northerly winds, with wind speeds in Exp.R approximately 15% and 9% higher, respectively, than those in Exp.C. In contrast, during summer (June to August), wind speeds in Exp.R were about 17% lower than those in Exp.C, and the wind directions were slightly more easterly.

The similar spatial distributions of nitrate and tracer ( Figures 13A, B, D, E ) indicated a significant influence of the autumn monsoon transition on the southward transport of pollutants discharged off the CJE. This spatial pattern was consistent with the previous study, which found heightened susceptibility of the Zhejiang coastal waters to these pollutants in autumn, mainly due to the climatological wind field (Cao et al., 2024). Additionally, the typhoon-induced strong northerly winds further enhanced the southward movement of pollutants, resulting in slightly higher concentrations of nitrate and tracer along the Zhejiang coast in Exp.R than Exp.C ( Figure 13C, F ).

The temporal evolutions of mean pollutant concentrations in the Zhejiang coastal waters showed that during September 1–3, the concentrations in Exp.R were lower than those in Exp.C ( Figure 9 ), indicating reduced southward transport, primarily due to a delayed monsoon transition relative to the climatological condition. However, the average pollutant concentrations in Exp.R increased significantly and exceeded those in Exp.C during the typhoon events, suggesting that Lingling and Tapah were the predominant drivers of the water quality degradation in September 2019, besides the overall impact of the monsoon transition throughout the autumn.

Riverine nutrient input is also considered a major factor impacting seawater quality in coastal regions (Liu et al., 2017; Tong et al., 2017; Liu et al., 2021). According to the River Sediment Bulletin of China, the total discharge of the Changjiang in 2019 was 9.33×10¹¹ m³, closely matching the decadal average of 2013–2022 (9.17×10¹¹ m³). The annual mean nitrate concentration measured at the Datong hydrological station in 2019 was 104.18 μmol/L, slightly below the average of 1999–2017 (114.92 μmol/L; Zhou et al., 2008). Based on monthly river runoff and nitrate concentration recorded at the Datong station, the total nitrate flux in 2019 was 94.81 Gmol, closely aligned with the average of 2013–2019 (95.34 Gmol; Wu et al., 2023). Therefore, the river discharge and DIN flux of the Changjiang in 2019 were generally consistent with the multi-year average, suggesting that the degradation of water quality in the Zhejiang coastal region was not due to increases in riverine nutrient input.

Note that of all eight watersheds located in Zhejiang Province (Qiantang River, Canal, Yongjiang, Oujiang, Jiaojiang, Tiaoxi, Feiyun River, and Aojiang), only nutrient input from Qiantang River was considered in this study. The other seven rivers were excluded based on the fact that the annual total DIN flux from these rivers was 10.8 Gmol, accounting for only 8.4% of the combined flux from the Changjiang and Qiantang River (128.5 Gmol; Chen et al., 2017; Wu et al., 2023).