Chlorophyll-a (Chl-a) is a key pigment in phytoplankton and serves as a proxy for changes in the marine ecosystem. Satellite observations have enabled long-term monitoring of Chl-a, offering insights into ocean productivity and biomass. Phytoplankton influence carbon dioxide levels in both the ocean and the atmosphere. Chl-a concentrations are shaped by environmental factors such as sea surface temperature, sunlight, and surface winds, which affect stratification and nutrient availability1,2,3. Satellite ocean color observations have significantly contributed to our understanding of Chl-a changes in marine ecosystems over the past few decades. SeaWiFS and MODIS data, in particular, have provided valuable opportunities to investigate long-term trends in satellite-observed concentrations across many local oceans and the global ocean4,5,6,7,8,9. In the seas around the Korean Peninsula, long-term trends and the phenology of Chl-a concentrations have been examined and discussed, focusing on potential causes of their variability. These studies explore relationships with sea surface temperature, wind speed, mixed layer depth, stability, nutrient supply, and wind forcings10,11,12,13. Changes in the physical environment are reported to play a key role in the long-term variations of Chl-a in the seas around Korea in a warming world14. In the coastal regions, river discharge is one of the most important factors influencing the physical environment as well, affecting salinity distributions due to freshwater input, nutrient supply from land, stratification caused by fresh water, and changes in surface currents15.

Excluding the Amazon River, one of the most notable regions influenced by river discharge is the area near the Yangtze River (YZR) (also called the Changjiang River), which flows into the Yellow Sea (YS) and the East China Sea (ECS). The YZR is the third-longest river in the world, stretching 6,300 km, and ranks fourth in terms of discharge volume. Due to the large volume of freshwater it releases, it has a significant impact not only on coastal ecosystems near its estuary mouth but also on far-offshore regions along the path of its discharge. Not a few studies have also noted its contribution to ocean color variability in Chinese coastal regions, suggesting its potential impact on marine environments16,17,18. Seasonal and interannual variations in Chl-a concentrations in the YS and ECS, as well as the effects of YZRD on these variations, have been studied using SeaWiFS data. To gain a detailed understanding of Chl-a variability, extensive oceanic and atmospheric measurements are essential. The Ieodo Ocean Research Station (I-ORS), located at 125° 10′ 56.81″ E, 32° 07′ 22.63″ N(indicated by an asterisk in Fig. 1, has been operational since June 2003, measuring a variety of atmospheric and oceanic variables19. The I-ORS is situated in the middle of the ECS and the eastern part of the potential river discharge passage, making it an ideal location to capture data on the physical environment that influences ocean color variability induced by river discharge. Many previous studies have examined the impact of the YZRD near the Chinese coastal region; however, the relationship between long-term Chl-a concentration and the physical environment remains unexplored in the region around I-ORS, a key ocean measurement site in the ECS.

Fig. 1
figure 1

(a) Comparison of Soil Moisture Active Passive (SMAP) sea surface salinity (SSS) and in-situ salinity data of I-ORS for the period 2015–2020, (b) relationship between Yangtze River discharge and surface salinity observed at I-ORS. The red dashed line represents the least-squares linear regression (surface salinity = − 7.3656 × 10⁻⁵ river discharge + 32.9391, p-value = 2.3165 × 10⁻⁸), and (c) spatial distribution of monthly mean of SMAP SSS (psu) for April 2015 to March 2020; asterisk indicates location of I-ORS.

Full size image

In this study, we analyze the spatial and temporal variations in optical water types from 2003 to 2022 and investigate the relationship between Chl-a concentration near I-ORS and YZR. To achieve this, we validate the Soil Moisture Active Passive (SMAP) sea surface salinity (SSS) with in-situ measurements from I-ORS. This study aims to understand the impact of riverine discharge on the marine environment of the ECS and the importance of physical and biological interactions in shaping Chl-a distributions in summer over two decades.

Validation of satellite sea surface salinity

To evaluate the extent of dispersion and advection of low-salinity water from the YZRD within the study area, the accuracy of satellite-derived salinity data was validated. The I-ORS, located east of the Yangtze River, provides real-time surface salinity measurements at 10-minute intervals, making it an ideal reference for satellite salinity validation. SMAP SSS was validated against in-situ salinity measurements from I-ORS using a spatial limit of 5 km and a temporal limit of 10 min, based on a total of 929 matchups collected from 2015 to 2020. These criteria are significantly more stringent than those used in previous studies20,21, allowing for a more precise evaluation of satellite-derived salinity accuracy.

Monthly comparison between satellite-derived salinity and in-situ measurements at I-ORS were compared using a scatter plot color-coded by month from January to December (Fig. 1a). The matchups yielded an RMSE of approximately 1.84 psu, a standard deviation of 1.52 psu, and a mean bias error of 1.03 psu (N = 929, r = 0.72). During the peak discharge period of the YZRD in summer (June and July), in-situ salinity at I-ORS ranged from 25.2 to 25.4 psu, while satellite-derived salinity was approximately 26 psu, resulting in a satellite-in-situ discrepancy of around 0.6–0.8 psu. This reduced mismatch indicates improved agreement between satellite and in-situ salinity measurements in the presence of low-salinity water22,23. In contrast, the winter and early spring months (January to March) exhibited significant overestimation errors and greater scatter in the data. These observations align with reported uncertainties in Aquarius and SMAP salinity datasets21. Although satellite-derived salinity generally tended to overestimate salinity outside the summer months, an overall linear correlation was observed. The notable reduction in satellite salinity errors during the summer underscores the suitability of satellite-derived salinity data for monitoring the dispersion of low-salinity water from the Yangtze River during this season compared to other times of the year (Fig. 1a).

Relation of river discharge and I-ORS in-situ salinity

Low-salinity water from the Yangtze River can have two potential impacts on the physical environment in terms of marine ecosystem responses. One is the supply of nutrients to nutrient-deficient seawater during the summer, and the other is the strengthening of stratification due to the lower density of river water compared to the relatively denser seawater. When diluted low-salinity river water flows into I-ORS, the sea surface salinity will vary annually. To test this hypothesis, we investigated the relationship between sea surface salinity measured at I-ORS and river discharge. The river discharge observed at Datong peaks in July, and this low-salinity water spreads eastward and northeastward across the ECS throughout the summer.

To investigate overall relationship between YZRD and SSS at I-ORS, we analyzed all available salinity observations at I-ORS from 2002 to 2020, irrespective of season (Fig. 1b). A least-squares linear regression analysis revealed a statistically significant inverse correlation between SSS and YZRD, with a slope of − 7.3656 × 10⁻⁵ psu s m⁻³ and a p-value of 2.3165 × 10⁻⁸, indicating significance at the 95% confidence level. The regression line, shown as a red dashed line in Fig. 1b, suggests that lower discharge is generally associated with higher salinity at I-ORS, whereas higher discharge tends to result in decreased salinity. These findings support the hypothesized inverse relationship between YZRD and SSS at I-ORS. However, despite statistical significance, the correlation coefficient was relatively weak (r = − 0.45). This may be partly attributed to the temporal phase lag in comparing variables from the same month, without accounting for the 1–2 month delay typically required for low-salinity water to be advected from the river estuary to I-ORS. Moreover, the transport of freshwater is governed by complex physical processes—including tidal forcing, frontal dynamics, ocean currents, and variable air-sea interactions24—that may introduce nonlinearity and variability in both spatial pathways and arrival times. Previous studies have emphasized the influence of interannual variations in wind patterns on the dispersion of the river plume; in particular, changes in wind direction can significantly alter the trajectory of low-salinity water25. Despite these complexities, our results consistently demonstrate that increased YZRD is associated with lower SSS at I-ORS.

Dispersion of satellite-observed sea surface salinity

The seasonal variation in the amount of YZRD is hypothesized to influence the SMAP satellite-derived salinity at I-ORS. As YZRD delivers freshwater to the coastal region, it decreases the salinity near the river estuary and presumably affects salinity values at I-ORS, depending on the discharge volume and coastal currents. To investigate the dispersion characteristics of the freshwater plume, we analyzed SMAP SSS data for the period from April 2015 to March 2020. The five-year averaged SSS data reveal that in January, low-salinity water (< 15 psu) was confined to the YZR bank. From February to March, this freshwater plume began to disperse offshore, with salinity ranging between 20 and 25 psu (Fig. 1c). The iso-haline of 30 psu, denoted by a solid black contour, illustrates the gradual eastward spreading of diluted seawater between April and June. During summer, the low-salinity water extended farther eastward, reaching I-ORS in July and subsequently spreading to Jeju Island, south of the Korean Peninsula. However, this eastward extension rapidly retreated westward in September and October, becoming constrained along the Chinese coastal zone (Fig. 1c). Despite the inherent uncertainties of satellite-derived SSS, the analysis confirms that low-salinity water from the YZRD exerts its greatest influence on I-ORS during July and August each year. This finding highlights the significance of YZRD’s freshwater discharge in shaping the salinity distribution in the ECS.

Long-term optical water types from satellite data

Seasonal water types from GOCI data. Despite the low-salinity outflow from the YZR, the seawater near I-ORS frequently appears optically clear, contrasting with the turbid waters near the river estuary (Fig. 1a). Seasonally classified water types were derived from GOCI data at a spatial resolution of 500 × 500 m, based on the highest frequency of estimated types at each pixel (Fig. 2a). Previous studies classified water types 1–8 as relatively clear, 9–14 as moderately turbid, and 15 or higher as highly turbid26.

Fig. 2
figure 2

(a) Seasonal distribution of most frequently occurring optical water types derived from GOCI L2 remote sensing reflectance (Rrs) spectra in the study area for 2015; the asterisk indicates the location of I-ORS. Optical water type was classified according to Wei et al.(2016) and Cui et al.(2020). (b) Monthly distribution of optical water types derived from MODIS data over two decades (2003–2022).

Full size image

Most of the study area exhibited relatively clear waters, with types 3–8 observed in regions such as the Northwest Pacific (types 3–4), the eastern ECS (types 4–6), and the central YS (types 6–8) (Fig. 2a). Seasonal water type classifications reveal that highly turbid waters (types 13–15) were present near the YZR in winter (December–February) and transitioned to less turbid waters (types 9–12) as they dispersed southeastward in spring (March–May). During spring and winter, water types at I-ORS were classified as moderately turbid (type 11), likely due to vigorous mixing driven by tidal currents and strong northerly and northwesterly winds. In contrast, summer water types at I-ORS were classified as 6–7, indicating relatively clear conditions, which persisted into fall. This finding contrasts with the hypothesis of spatial dispersion of highly turbid, low-salinity water toward I-ORS during summer, as inferred from SSS distribution.

Water types from MODIS data. The right panel of Fig. 2b presents the monthly distribution of classified water types derived from two decades of MODIS Rrs data, which aligns closely with the findings from GOCI data. Near the I-ORS, water types were predominantly classified as type 10 in January, transitioning to type 8 by April, and between types 6 and 7 near the frontal zone from May to June. Interestingly, from July to September, the waters were classified as relatively clear (types 5–6), contradicting the expectation that low-salinity water originating from the YZRD would exhibit higher turbidity. Typically, river-derived low-salinity water is expected to possess elevated turbidity and thus fall into turbid optical water type categories. However, the present results reveal the opposite. This finding appears inconsistent with the earlier observation that low-salinity water from the YZRD extends to the I-ORS during summer27. The classification at I-ORS, situated east of the frontal zone where abrupt changes in water types occur in summer and fall, suggests that the surrounding waters are relatively clear. This observation implies that while the YZRD spreads as low-salinity water, materials that significantly contribute to turbidity—such as suspended sediments and nutrients—may be substantially utilized for Chl-a blooms, settle out of the water column, or fail to reach the I-ORS. The persistently low optical water type values observed in summer suggest that the I-ORS area maintains relatively clear surface waters during this season. This clarity is likely not due to the absence of riverine or current-driven influence, but rather reflects a combination of strong summer stratification, weak wind-induced mixing, and dilution of turbid constituents before they reach the station27. Therefore, summer optical clarity at I-ORS may be more strongly controlled by local biogeochemical and optical conditions than by physical transport mechanisms. This dynamic underscores the complex interplay between physical dispersion and biological processes in determining optical water types in the ECS.

Long-term monthly Chl-a variations and coherence with river flow

The long-term annual mean Chl-a concentrations from 2003 to 2022 reveal a pronounced spatial contrast in the ECS, with generally higher concentrations in the western regions and significantly lower concentrations in the eastern ECS (Fig. 3a). During spring, a Chl-a bloom exceeding 1.5 mg m⁻³ extends broadly to the eastern part of the I-ORS in April, reflecting a widespread increase in primary productivity. However, the Chl-a front, characterized by a sharp spatial gradient, begins to retreat westward from May to June, establishing a persistent pattern throughout the summer months as observed by Kim et al.(2009)27. This westward retreat and the resulting spatial gradient suggest that the I-ORS is situated within a transitional zone for Chl-a distribution in the ECS. The observed patterns underscore the influence of both physical (e.g., riverine inputs, stratification) and biological processes (e.g., nutrient availability, phytoplankton dynamics) on the spatial variability of Chl-a. These dynamics highlight the importance of understanding transitional zones like the I-ORS for interpreting the interactions between river discharge and marine ecosystems in the ECS.

Fig. 3
figure 3

(a) Monthly distribution of optimally interpolated MODIS Chl-a concentrations over two decades (2003–2022); the asterisk indicates the location of I-ORS. (b) Coherence wavelet transform (CWT) scalogram between the Chl-a time series at I-ORS and river discharge, where the dashed line denotes the 95% confidence level based on a chi-square distribution with 2 degrees of freedom. Regions outside this cone are not statistically significant and were thus excluded from the interpretation.

Full size image

As the low-salinity water discharged from the YZRD propagates eastward and reaches the I-ORS, it is expected that changes in the physical environment would influence the temporal variability of Chl-a at the station. However, the coherence wavelet analysis revealed a weak and sporadic relationship between the temporal variability of river discharge and Chl-a concentrations at the I-ORS. The wavelet power spectra and coherence were tested for statistical significance using a chi-square distribution with two degrees of freedom (DOF = 2), as is standard when applying complex Morlet wavelets (Fig. 3b). In 2017, the dominant coherence between the two variables appeared at higher frequencies corresponding to intraseasonal variations. However, this may reflect a localized event or unexplained noise, rather than a consistent intraseasonal pattern (Fig. 3b). Contrary to the initial hypothesis, there was no consistent or significant correlation between YZRD and Chl-a in the I-ORS, except for these specific frequencies. This unexpected result suggests that additional factors, such as local oceanic processes, stratification, nutrient utilization, or biological responses, may play a more significant role in governing Chl-a variability at the I-ORS than previously anticipated. Further investigation is needed to elucidate these interactions and the potential influence of other environmental drivers.

Spatial response of Chl-a to river discharge

Previous studies have reported that phytoplankton Chl-a concentrations from September to November exhibited a clear response to YZRD from July to September, with a time lag of 1–2 months17. The eastward advection of high Chl-a concentrations observed via satellite corresponded with the spread of diluted river water, which was estimated to take approximately 1–2 months to reach Jeju Island28. In this study, it is hypothesized that, once the Yangtze River discharge (YZRD) enters the sea and is advected and dispersed by surrounding currents, it influences the spatial distribution of Chl-a in the YS and the ECS. To explore this relationship, we assumed that the Chl-a concentration at each pixel is temporally correlated with the summer YZRD, with a potential phase lag. An experiment was conducted to assess how the correlation between YZRD and Chl-a evolves as the phase lag increases. Since Datong station is located approximately 624 km inland from the river estuary, we utilized river discharge data from July and August to account for the time required for freshwater to reach the sea. Phase lags ranging from 0 to 3 months were applied to examine the relationship between summer YZRD and Chl-a at each pixel. This approach allowed for an analysis of the temporal dynamics and spatial extent of the riverine influence on Chl-a distribution across the region.

When there is no phase lag between the YZRD and Chl-a, a high positive response is observed over a wide area near the river mouth, extending along the coastal regions (Fig. 4a). Figure 4a illustrates the Chl-a response when there is no phase lag between the YZRD and Chl-a. A high positive response is observed over a wide area near the river mouth, extending along the coastal regions. Interestingly, while a northeastward flow of the YZR low-salinity water was anticipated to elicit a strong response in the northern waters near the I-ORS, the extent of this response was more limited than expected. Conversely, a broad counterclockwise (cyclonic) deflection emerged south of I-ORS, suggesting notable dispersion patterns in that region. With a 1-month phase lag (Fig. 4b), the Chl-a response extended further northward of I-ORS, while the influence southward expanded more broadly compared to the no-lag condition. This indicates a temporal delay in the advection and diffusion of riverine water and its associated nutrients. For a 2-month phase lag (Fig. 4c), the response north of I-ORS seemed to extend to Jeju Island and the Korea Strait, highlighting a progressive northeastward spread of the riverine influence similar to the schematic pathways of Liu et al.(2021)15. Simultaneously, the counterclockwise pattern around I-ORS persisted and exhibited an even more pronounced southward extension. At a 3-month phase lag (Fig. 4d), the positive response aligned along the southern Chinese coast, with the cyclonic pattern around I-ORS remaining weakly visible. These results highlight the gradual northeastward dispersion of YZR low-salinity water and its associated nutrients while emphasizing the residual influence of riverine water south of I-ORS. The cyclonic pattern around I-ORS remains evident even after three months, though its intensity weakens over time.

Fig. 4
figure 4

Spatial distribution of Chl-a concentration response (mg m−3) to YZRD (m3 s−1) in summer for phase lags of (a) zero month, (b) one month, (c) two months, and (d) three months from 2003 to 2022; the asterisk indicates the location of I-ORS. Statistically significant areas at the 95% confidence level are marked with black dots.

Full size image

Similarity between Chl-a response patterns and surface current pathways

The YZRD is advected northeastward, reaching the vicinity of I-ORS in July and August as a low-salinity water mass (< 30.5 psu). However, despite the presence of low-salinity water, optical water type classification indicated that the seawater around I-ORS remained relatively clear during this period. This discrepancy suggests that while the low-salinity water forms a stratified layer at the surface, its turbidity remains low. A possible explanation is that nutrients transported with the YZRD may have been largely consumed by phytoplankton during advection, or, in the case of weak surface currents, suspended sediments may have settled below the surface27. Satellite-derived Chl-a concentrations around I-ORS were at their lowest during summer. Phase-lagged Chl-a responses indicate that the northeastward influence of YZRD was weak north of I-ORS, while a pronounced counterclockwise pattern emerged, propagating from the south. These findings suggest that the physical transport and biogeochemical dynamics of YZRD play a key role in shaping spatial and temporal phytoplankton distributions around I-ORS.

The observed pattern of Chl-a in response to YZRD variations could be attributed to advection coupled with turbulent diffusion, following the regional surface currents. To investigate this, we analyzed the overall direction of the surface current field using 5-day interval Ocean Surface Current Analysis Real-time (OSCAR) data29,30. Drifter-driven surface current vectors exhibit a distinct cyclonic curvature south of I-ORS (Fig. 5a), suggesting that freshwater-induced biological responses may propagate along these circulation pathways. To further examine the transport mechanism of riverine waters, water particles were released at four locations along 123°E (31.3°N (T4), 32°N (T3), 32.6°N (T2), and 33°N (T1) in Fig. 5a), marked by red, green, yellow, and pink squares, respectively, and tracked using surface current vectors. The trajectory of the particle at 33°N, 123°E (pink) demonstrated a southward movement in July (asterisk), an eastward advection in August (diamond), a cyclonic rotation in September (triangle), and a northward shift in October (hexagram), ultimately positioned west of I-ORS. Similarly, the yellow particle initially moved northward but later shifted southward, remaining west of I-ORS without progressing eastward, partly due to strong horizontal density gradient across strong tidal fronts31. In contrast, the green-marked particle, originating near the YZRD, exhibited a distinct eastward advection, moving significantly south of I-ORS before following a large curvature and shifting north-northeastward. The particle south of the YZRD (red) was advected further south, following a strong counterclockwise circulation toward the Korea Strait. This counterclockwise turning pattern was consistent with previous analyses, with the particle ultimately reaching the Korea Strait by November. Overall, the trajectory of water particles demonstrated strong agreement with the observed Chl-a response patterns to river discharge south of I-ORS, reinforcing the hypothesis that physical transport mechanisms, particularly advection by surface currents, play a dominant role in shaping phytoplankton variability in this region.

Fig. 5
figure 5

(a) Tracks of water particles deployed at four sites, marked in pink (T1), yellow (T2), green (T3), and red (T4) rectangles, along 123°E near the Yangtze River mouth from July to November by using OSCAR current product, where the background vectors represent the spatial distribution of drifter-derived currents (m s–1), provided by Atlantic Oceanographic and Meteorological Laboratory (AOML), in the study area. Different colors indicate the trajectories of individual particles. (b) Comparison of surface (black) and 0–30 m averaged (white) current vectors from the KHOA numerical circulation model. (c) Spatial distribution of nitrate (NO3−) in August 2022, optimally interpolated using KODC station data (black circles) along observation lines 315, 316, and 317 in the ECS.

Full size image

During summer, the intensification of stratification can reduce topographic constraints, facilitating the cross-isobath transport of riverine water between July and August32. As stratification weakens in the fall, the deepening of the mixed layer enhances seawater homogeneity. At this stage, a portion of the riverine water may be advected from the estuary into deeper offshore regions under the influence of bathymetric features. This process can establish a quasi-geostrophic balance, characterized by positive relative vorticity, ultimately contributing to the development of a Chl-a response branch south of I-ORS during summer, potentially forming a cyclonic curvature15. The OSCAR surface currents, derived from a quasi-steady geostrophic model, incorporate wind-driven ageostrophic components and thermal adjustments to represent large-scale upper-ocean circulation. However, as these currents are depth-averaged over the upper 30 m, discrepancies may arise between OSCAR-derived velocities and actual near-surface currents, particularly during summer when stratification is most pronounced. To evaluate this potential bias, we utilized daily reanalysis output from the Korea Hydrographic and Oceanographic Agency’s coastal ocean circulation model for July 2020. Monthly mean current fields were constructed using both surface (1 m depth) and depth-averaged (1–30 m) velocity vectors. Figure 5b presents a comparison of these vectors, where black arrows represent 1 m currents and white arrows indicate the depth-averaged flow. The results show that, except in the vicinity of the Yangtze River mouth, the current directions across the study region were nearly identical, with speed differences generally within 0.05 m s−1. Despite the vertical averaging inherent in OSCAR, this consistency supports the use of OSCAR currents as a reasonable proxy for upper-layer advection. Moreover, the agreement between the spatial structure of the phase-lagged Chl-a response and particle-tracking simulations implies a robust transport mechanism. These findings reinforce the role of surface advection in modulating phytoplankton variability in the ECS under the influence of riverine discharge.

Similarity between in-situ nutrient distribution and surface current path

Phytoplankton blooms in seawater are influenced by multiple environmental factors, including nutrient availability, solar radiation, and stratification. Previous studies based on in-situ measurements of nutrients, salinity, and Chl-a concentrations have highlighted the role of riverine freshwater discharge in shaping marine ecosystems27. Field observations west of I-ORS indicated that phosphate and nitrate from the Yangtze River estuary disperse eastward during summer, with higher concentrations in the western ECS and lower levels in the eastern region27.

If phytoplankton blooms are driven by nutrients transported from the Yangtze River, the spatial distribution of nutrients should exhibit a corresponding pattern. Analysis of objectively interpolated KODC (Korea Ocean Data Center) nitrate (NO₃⁻) data in August 2022 at stations, marked in circles along observation lines, reveals high nitrate concentrations (> 10 µmol L−1) near the YZR estuary, extending northeastward toward Jeju Island (Fig. 5c). Notably, nitrate concentrations exceeding 7 µmol L-1 display a counterclockwise curvature south of I-ORS, aligning with the inferred water movement in this region. Conversely, the waters surrounding I-ORS exhibit relatively low nitrate levels (1–4 µmol L−1). This spatial distribution closely corresponds to the Chl-a response observed in this study, indicating the critical role of nutrient availability in sustaining phytoplankton growth within the oligotrophic oceanic environment during summer. Additionally, the counterclockwise circulation south of I-ORS, identified through two decades of satellite Chl-a observations, represents a previously unreported feature in earlier studies.

Discussion and conclusion

The influx of low-salinity water from the YZR plays a crucial role in shaping the physical environment and, consequently, the marine ecosystem. In addition to the aforementioned factors, the vertical stratification effect of seawater may also be important. During summer, the YZRD introduces essential nutrients into the oligotrophic waters, while its lower density causes it to spread across the surface layer, reinforcing vertical stratification. The interannual variability of summer YZRD exhibited a strong inverse correlation with surface salinity fluctuations at I-ORS, indicating that YZRD significantly influences stratification by reducing surface salinity. Enhanced stratification has profound implications for nutrient dynamics and biological productivity. By suppressing vertical mixing, it can prevent the downward transport of nutrients, effectively maintaining nutrient availability in the surface layer and fostering phytoplankton growth in typically nutrient-depleted waters. Additionally, stratification alters the vertical distribution of light by trapping solar radiation in the surface layer, which may further enhance phytoplankton productivity33. However, given the high solar insolation during summer, light availability is unlikely to be a limiting factor in this region.

This study examined the long-term influence of YZRD on phytoplankton Chl-a variability in the waters surrounding I-ORS from 2003 to 2022. YZRD contributed to a persistent low-salinity environment at I-ORS during summer, showing a strong inverse correlation between river discharge and sea surface salinity. While satellite-derived SSS is generally subject to greater biases and uncertainties during winter due to the reduced performance of passive microwave radiometers20, the summer ocean–atmosphere conditions in the YZRD region—characterized by low salinity, high sea surface temperature, and weak winds—are favorable for improving SMAP SSS accuracy13. Under these conditions, satellite-based salinity data are considered sufficiently reliable for evaluating the dispersion patterns of the YZRD plume. Despite this freshwater influence, satellite-derived optical water type classifications consistently indicated the clearest water conditions at I-ORS during summer. This suggests that turbidity-related materials were either depleted, settled, or did not reach the I-ORS region in significant quantities. Suspended sediments originating from the Yangtze River likely underwent substantial deposition on the inner shelf due to gravitational settling and distance from the river mouth. Given the transport distance of several hundred kilometers, it is plausible that turbid components were largely diluted or removed before reaching I-ORS. Furthermore, rapid uptake of riverine nutrients by phytoplankton may have contributed to the classification of clear optical water types, even under nutrient-rich conditions27. Additionally, one possible explanation involves frontal dynamics. Our findings are consistent with model-based studies32, which demonstrated that coastal frontal structures act as transport barriers that restrict cross-shelf dispersion of plume waters. These physical barriers can increase residence time within the near-field zone, facilitating nutrient consumption and enhancing phytoplankton growth near the front. The observed phase-lagged Chl-a response near I-ORS supports this mechanism, indicating that frontal trapping and delayed advection may play an important role in modulating transport timescale depending on various conditions, such as winds, tidal current, fronts, ambient currents, and inherent Chl-a variability32.

Although this study employed the YOC algorithm, which applies a regionally optimized selection between the OC3 and Tassan-like algorithms, the accurate estimation of Chl-a concentrations may still be limited in optically complex Case-2 waters. Nevertheless, the phase-lag analysis, consistent with our hypothesis, revealed that nutrient-rich, low-salinity water from the YZR predominantly followed a cyclonic trajectory south of I-ORS before dispersing northeastward. This pattern was consistent with particle-tracking simulations, which demonstrated that YZRD-advected waters followed a quasi-cyclonic geostrophic current structure. Furthermore, the modeling study offered valuable insights into the potential transport pathways of YZRD15, showing general agreement with our conceptual trajectories and indicating that the northeastward advection of YZRD during summer is primarily driven by prevailing southerly winds. However, it should be noted that the OSCAR surface currents used in this study represent 5-day mean velocity fields averaged over the upper 30 m. During summer, when stratification strengthens, this depth-averaged representation may not fully capture near-surface (0–10 m) dynamics, potentially leading to underestimation of surface advection strength and trajectory accuracy. To address this limitation, future studies incorporating high-resolution numerical modeling and direct in-situ current measurements are required. Nevertheless, our findings provide critical insight into the spatial and temporal dynamics of riverine influence on satellite-observed Chl-a variability in complex coastal environments. Similar river discharge effects on marine ecosystems are likely in other coastal regions worldwide. Continued research integrating high-resolution oceanographic observations and modeling will be essential to further elucidate these interactions and their implications for biogeochemical cycles in river-influenced marginal seas.

Data and methods

Data set. a. Chrorophyll-a data. We used MODIS-Aqua level-2 (L2) swath data with a 1 km resolution (version 2022, https://oceancolor.gsfc.nasa.gov/) to produce Chl-a concentration data and optical water types from 2003 to 2022.

b. Salinity data. We collected NASA’s SMAP SSS products using the Combined Active-Passive (CAP) algorithm V4.3, developed by the Jet Propulsion Laboratory (JPL) (http://podaac.jpl.nasa.gov/smap/)34. Time series of sea surface salinity data were used to explore the relationship between river discharge and salinity changes at I-ORS.

c. River discharge data. Time series data on YZRD rates at Datong Station from January 2003 to December 2022 were obtained from the Changjiang Water Resources Commission, China (www.cjh.com.cn)35,36.

d. KODC nitrate data. KODC nitrate (NO₃⁻) observations from stations along zonal transects 315, 316, and 317 were collected for August 2022 (https://www.nifs.go.kr/kodc/index.kodc).

Calculation of Chl-a concentration and monthly composites. Yellow Sea Large Marine Ecosystem Ocean Color Project (YOC) algorithm was applied to calculate Chl-a by using the NASA algorithm for clear waters, the Tassan-like algorithm for turbid waters, and linear interpolation for intermediate water conditions37,38. To generate monthly composite fields from 2003 to 2022, Chl-a data were processed for monthly data using speckle removal, weighted average, and optimum interpolation39.

Classification of seawater optical types. By using the representative spectra of these 23 optical water types proposed by Wei et al. (2016)40 and Cui et al. (2020)26, we calculated the spectral similarity using the spectral angle mapper between satellite data and reference reflectance data41. To verify the reliability of the water type results from the daily MODIS data with lots of cloud cover, a total of 2,920 Geostationary Ocean Color Imager (GOCI) images of hourly Rrs data across 8 bands in 2015 were also used to investigate monthly variations in water types.

Phase lag experiment. To understand the dispersion of the impact of river discharge, we conducted experiments on the river discharge and Chl-a at each grid, introducing phase lags ranging from zero to three months. It was assumed that the two variables had a linear relationship after removing seasonal variations using four harmonic components. The model assumed that the Chl-a at time t was a combination of a constant term (\(\:{\mu\:}_{k}\)) for a phase lag (k) in a month, harmonic constituents with coefficients \(\:({a}_{1,j},{\:a}_{2,j})\) at 4 frequencies from 1 to 4 cycles per year (cpy) (St)42,43,44,45.

$$\:{Chla}_{t-k}={\mu\:}_{k}+{S}_{t-k}+{\alpha\:}_{k}{R}_{t}$$
(1)
$$\:{S}_{t-k}=\sum\:_{j=1}^{4}\:\left[{a}_{1,j}\:sin\left(2\pi\:{f}_{j}t\right)+{a}_{2,j}\:cos\left(2\pi\:{f}_{j}t\right)\right]$$
(2)

where \(\:\alpha\:\) is the Chl-a concentration change rate, mg m– 3/(m3 s–1), to river discharge rate R (m3 s– 1), \(\:{f}_{j}\) is the frequency (cpy) of harmonic constituents, and k is the phase lag in a month. To determine the direction in which water particles from the YZR are advected by surrounding currents, we utilized the OSCAR near-surface current vectors (https://www.esr.org/research/oscar/oscar-surface-currents/)30,45,46,47,48. The 1/3° resolution OSCAR surface current data with a 5-day temporal interval were utilized for the period 2003–2022. A climatological mean current field was derived from this dataset and subsequently applied in Lagrangian particle tracking experiments to investigate the dispersal patterns of YZRD. It is important to note that OSCAR surface currents represent averaged velocity fields over the upper 30 m of the ocean. During summer, when vertical stratification is strong, these currents may not fully reflect surface-layer dynamics (e.g., 0–10 m). Therefore, the particle-tracking experiment in this study should be regarded as a conceptual analysis aimed at identifying the potential pathways and general dispersal patterns of riverine waters, rather than a precise simulation of near-surface transport.