In this study, we focused on five marginal seas around the world with different water-exchange rates and levels of anthropogenic activity (Fig. 1) (Cui et al., 2021; Dobson et al., 2000; Lloyd et al., 2019). Enclosed seas (in this study, the Bohai Sea and the Gulf of Mexico) are typically characterized by particularly slow water exchange, which limits the diffusion and transportation of land-sourced pollutants and produces a limited self-purification capability. The three open-sea areas included in this study were the Changjiang Estuary, the Eastern Coastal Waters of the United States, and the Amazon Estuary. In this study, we defined an open-sea area relative to the enclosed-sea area, that is, a near-shore area with only one land boundary. Generally, the hydrodynamic conditions in open-sea areas are favorable for the diffusion of land-derived materials.

The East China Sea is one of the largest marginal seas of the northwestern Pacific, with complex circulation systems (Fig. 1a). The Tsushima and Kuroshio currents, with high temperatures and salinities, prevail in the eastern side of the East China Sea. On the western side, the Zhemin Coastal Currents, formed by the Changjiang River Plume with the intrusion of the Taiwan Warm Current, dominate. Due to increasing human activities in the Changjiang River Basin, nutrient input has increased significantly, leading to more frequent harmful algal blooms (HABs) (Ministry of Ecology and Environment, PRC, 2001, 2021; Zhou et al., 2022).

The Eastern Coastal Waters of the United States, located to the northeast of the Gulf of Mexico, are strongly influenced by the Gulf Stream, resulting in enhanced primary productivity (Ma and Smith, 2022). Numerous rivers, including the Hudson, Delaware, Susquehanna, and Joptank rivers, which flow through highly urbanized cities, discharge pollutants and have increased the frequency of HAB events in the past decades (e.g., Mizuta and Wikfors, 2020).

In the Amazon Estuary, the tropical Atlantic surface circulation, specifically the North Brazil Current running north-westward along the Brazilian coastline, exerts a dominant influence over the area (Fig. 1c). The Guyana Current is also active in this sea area. HABs in the Amazon Estuary typically occur during the spring and summer, influenced by the seasonal flow variations of the Amazon River. Notably, HAB events have increased since 2008 (Dai et al., 2023; Hallegraeff et al., 2021; Subramaniam et al., 2008).

The Bohai Sea has been undergoing severe anthropogenic disturbance with coastal runoff from at least 40 rivers, including the Yellow River, Haihe River, and Liaohe River. The influx of significant amounts of pollutants, including sewage, has severely disrupted the local aquatic ecology (Ministry of Ecology and Environment, PRC, 2021). The circulation of the Bohai Sea exhibits seasonal variations (Fig. 1d), which has significantly affected pollutant transport (Wu et al., 2023; Zhang et al., 2018). The rate of seawater exchange with the Yellow Sea has strongly influenced the ecological environment within the Bohai Sea (Ju et al., 2020).

The circulation system in the southeast of the Gulf of Mexico (Fig. 1e) acts as its primary hydrodynamic force, stimulating the local primary productivity (Damien et al., 2021). In contrast, the northern part of the Gulf of Mexico is significantly influenced by nutrient input originating from the Mississippi–Atchafalaya Basin. Accelerated industrial growth and extensive fertilizer use have led to the excessive production of algae (e.g., Fu et al., 2020; Yingling et al., 2022).

The five selected sea areas cover varying degrees of human activity and hydrodynamic exchange capability, as summarized in Table 1. We obtained population density data for the surrounding basins from the LandScan (2020) high-resolution global population dataset (Lebakula et al., 2025), which we used as a proxy for anthropogenic pressure. In this context, the term “Human regulation” quantifies the intensity of human activities within the basin that regulate nutrient input and environmental stress on the marine ecosystem. Because long-term, internally consistent time-series datasets of riverine total nitrogen and total phosphorus fluxes were not available for all five study regions over the entire analysis period (1998–2020), we did not include quantitative nutrient loading as an explanatory variable in this comparative analysis.

Regarding the various chemical, hydrological, meteorological, and biological factors, satellite-derived datasets with Chl a, SST, PAR, and SWS data were used to evaluate temporal and spatial variation trends and interactions (Hong et al., 2023; Martinez et al., 2020; Xi et al., 2021). SST was used to identify periods in which the seawater was heated, and PAR and SWS were included as the key environmental indices known to influence phytoplankton growth and distribution. PAR directly constrains photosynthesis, whereas SWS affects vertical mixing, nutrient entrainment from deeper layers, and light availability by modulating the mixed-layer depth. Table 2 shows the types of data, data sources, temporal and spatial resolution, spanning periods, and download websites used in this study.

We selected these datasets, derived through algorithmic processing of raw satellite imagery, for their reliability and comprehensive temporal coverage, which made them particularly valuable for studying waters with complex optical properties, such as coastal and estuarine environments (Pisano et al., 2016; Zhang et al., 2006). We applied a consistent preprocessing method (see Sect. 2.3) to establish a unified satellite dataset from multiple satellites. Taking Chl a and PAR as examples, before constructing the time-series dataset, we divided the satellite Chl a and PAR data into the following three periods according to availability of different satellite products: T1 period: January 1998–December 2002, Chl a and PAR products were only available from the Sea-Viewing Wide Field-of-View Sensor (SeaWiFS); T2 period: January 2003–December 2010, overlap period, Chl a and PAR products were available from both the Moderate Resolution Imaging Spectroradiometer (MODIS) and SeaWiFS; and T3 period: January 2011–December 2020, Chl a and PAR products were available only only from MODIS. Because the two satellite sensors (SeaWiFS and MODIS) had different operational periods, we eliminated systematic errors before merging them into a long-term dataset. See Appendix A for an in-depth explanation of the data preprocessing methodology. We extracted the SST data from the AVHRR Pathfinder dataset (Pisano et al., 2016). Additionally, we incorporated SWS data from the Cross-Calibrated Multiplatform (CCMP) Wind Vector Analysis Product (Mears et al., 2019). All datasets used in this study were last accessed on 25 May 2026.

Figure 2 shows a flowchart of this study. To harmonize datasets with different spatial resolutions, we reprojected and resampled all the raster datasets to a common reference grid before further analysis. We used a MODIS grid file covering the study areas as a spatial reference, and resampled the Chl a, PAR, SST, and SWS datasets using bilinear interpolation with the GDAL Warp function. This procedure ensured that all variables had consistent projection, pixel size, and spatial extent. We then used the temporally matched monthly datasets for subsequent analyses. Regarding the availability of satellite datasets for Chl a, PAR, SWS, and SST (1998–2020), we determined the correlation and interaction between changes in Chl a concentration and seawater warming according to the following steps: (1) We analyzed the temporal trends of seasonal and annual Chl a concentrations in the five study areas using the Mann–Kendall (M-K) trend analysis method. (2) We calculated Chl a stability and the coefficients of variation for each pixel, the varying degrees of Chl a concentrations, and conducted correlation analyses and principal component analyses (PCA) to determine the predominant marine environmental factors affecting Chl a concentrations in each sea area. (3) We identified global warming periods using the SST dataset. (4) We calculated the image-by-image metric correlations between Chl a and anthropogenic rising SST for specific time periods of interest.

Figure 4 shows the seasonal and annual spatial distribution of Chl a concentrations for the five study areas. All the study areas showed distinct trends of decreasing Chl a concentrations as the distance offshore increased, and higher values were observed primarily in the estuaries and bays. In particular, areas with higher Chl a were more uniformly distributed in the Bohai Sea with limited water exchange, whereas the Gulf of Mexico, although morphologically classified as an enclosed sea, showed a more similar spatial distribution tendency to the open-sea areas because of its strong water exchange with the Atlantic.

Chl a concentration varied from greatest to least as follows: Bohai Sea > East China Sea > Gulf of Mexico > Eastern Coastal Waters of the United States ≈ Amazon Estuary. As a severely eutrophic enclosed-sea area, the Bohai Sea maintained high values between 1.6 and 2.5 mgm-3 throughout the year. Meanwhile, the East China Sea, which is a highly eutrophic open-sea area, exhibited a consistently higher annual Chl a, with spring values frequently exceeding 0.6 mgm-3. The other three regions all had year-round Chl a below 0.6 mgm-3. Chl a concentrations in the open-sea area exhibited greater susceptibility to seasonal variation than those in enclosed-sea areas with restricted exchange, as shown in Fig. 4f. Changes in the monthly average rate of Chl a concentration surpassed 50 %.

The M-K test revealed distinct temporal trends in Chl a concentrations across the five study areas (Fig. 5). Notably, a statistically significant overall increasing trend was suggested in annual and all seasonal Chl a concentrations in the Gulf of Mexico (p < 0.05). We observed a slight increase in Chl a in the Eastern Coastal Waters of the United States, with a rate of 0.0301 mgm-3a-1. We did not observe any statistically significant trends (p > 0.05) in the East China Sea, the Eastern Coastal Waters of the United States, the Amazon Estuary, or the Bohai Sea. A distinct continuous increase of Chl a concentration, however, was indicated in the summer in the sea area of the Eastern Coastal Waters of the United States.

Notably, both the Bohai and East China seas exhibited a two-stage change, with a decreasing tendency after a continuously increasing trend; however, the decline of the East China Sea was not significant. The turning point for the Bohai Sea occurred in 2013, which was earlier than that of the East China Sea in 2017.

The seasonal mean Chl a concentrations did not always follow the same temporal trend as the annual mean. In the Gulf of Mexico, however, all four seasonal means followed consistent increasing trends that aligned with the annual mean trend (Fig. 5). The correlation analysis between the annual and seasonal Chl a concentrations (Table 5) suggested that the temporal changes in annual Chl a in the Bohai Sea were controlled dominantly by the spring and summer Chl a concentrations. In the Gulf of Mexico, owing to its different hydrodynamic regime, the annual trend was more uniformly correlated across all seasons. In open-sea areas, winter and spring Chl a concentrations were strongly correlated with the annual mean and thus served as the primary drivers of the annual trend. The Amazon Estuary was an exception, and summer and winter Chl a concentrations were more important in determining the annual trend.

Figure 6 and Appendix E illustrate the spatiotemporal variation rates of Chl a from 1998 to 2020. Overall, coastal regions exhibited increasing Chl a concentration trends, whereas offshore areas showed declining trends, with the exception of the Bohai Sea (Fig. 6d). The Bohai Sea had the largest area with positive annual growth (33.79 %), whereas the Eastern Coastal Waters of the United States had the smallest area with positive annual growth (6.53 %). Seasonally, the Bohai Sea also exhibited the most extensive increase in Chl a in the summer (Fig. 6d), which contrasted sharply with the Eastern Coastal Waters of the United States, which had the smallest growth area in the summer (8.06 %, Fig. 6b).

While a general decline in Chl a concentration levels was revealed across most time frames in the Bohai Sea, in its central basin in the summer, Chl a concentrations exhibited a sustained increasing trend (Figs. 5d and 6d). This anomaly was likely due to an ecological lag effect of nutrient aggregation in the central basin, which was facilitated by internal hydrodynamical patterns and a weak exchange capability with outer oceanic currents. Such nutrient enrichment in the central basin, particularly under the conditions of elevated summer temperatures, promoted HABs even amid reduced land-based input.

This dynamic, however, was intricately linked to the specific dimensions of the enclosed-sea area and the intensity of human activities within. The Gulf of Mexico, an analogous enclosed marine ecosystem, did not exhibit comparable patterns of Chl a concentration fluctuations. Strong influences from the Gulf Stream and Cuban Current enhanced water exchange with the Atlantic, making its Chl a patterns more akin to those of open seas. In contrast, in open seas, increasing Chl a was primarily confined to coastlines and river plumes, such as the Changjiang Estuary in the spring (Fig. 6a) and the Amazon Plume in the summer (Fig. 6c).

The stability of Chl a, quantified by the coefficient of variation from 1998 to 2020 across the five study regions, is illustrated in Fig. 7 and Appendix F. Regions with obvious increasing or decreasing Chl a trends possessed lower stability. Overall stability varied by sea type. The open seas were less stable than the enclosed seas, and the East China Sea had the lowest stability among all studied waters.

Open seas included the East China Sea, the Eastern Coastal Waters of the United States, and the Amazon Estuary. All the open seas presented unstable Chl a conditions in the spring and the summer. Pronounced seasonal discrepancies in Chl a stability were widely detected in open ocean zones, as shown in Fig. 7a–c. Chl a maintained higher stability in the winter across all study areas. We observed low stability in the Taiwan Strait, Gulf Stream, and Amazon Plume, whereas we observed that Chl a concentrations were relatively steady year-round in the partial waters of the Changjiang Estuary and the area near the mouth of the Amazon River.

Enclosed seas showed distinct stability patterns, represented by the Bohai Sea and the Gulf of Mexico. The Bohai Sea generally maintained moderate-to-high Chl a stability, excluding coastal and estuarine areas. Its stability remained relatively high despite seasonal fluctuations, which differed from the seasonal instability seen in its open-sea counterparts. By comparison, more than 90 % of the Gulf of Mexico featured low or moderately low stability. Areas with intense human activity, such as the northern Gulf and waters around the Florida Peninsula, sustained poor stability throughout the year. We witnessed slightly improved stability in the winter only in the central Gulf, which we attributed to the diminished human interference and decreased SST (Durán-García et al., 2017). Collectively, hydrodynamic conditions and anthropogenic disturbances jointly regulated Chl a stability in both open and enclosed marine environments.

To identify the major environmental factors affecting the observed spatiotemporal changes in Chl a concentrations in the five study areas, we correlated and evaluated the three main environmental indices SST, PAR, and SWS with Chl a using Pearson correlation analysis and PCA (see Sect. 2.3.4). Overall, the correlations between Chl a and the various marine environmental indices varied significantly depending on the study area, as presented in Fig. 8. In comparison with the much stronger correlation between Chl a and SST in the East China Sea and the Eastern Coastal Waters of the United States, and between Chl a and PAR in the Amazon Estuary, we did not identify any evident correlations between Chl a concentrations and the environmental indices in the two enclosed seas. In addition, we did not observe a distinct correlation between Chl a concentrations and SWS in any of the five study areas.

Among the three open seas, trends and degrees of correlations between Chl a and the environmental factors (SST, PAR, and SWS) differed markedly. All three of the open seas showed negative correlations between Chl a and SST. Specifically, the Amazon Estuary demonstrated a weaker negative correlation between Chl a and SST than the East China Sea and the Eastern Coastal Waters of the United States, but it exhibited a significant positive correlation between Chl a and PAR. We observed a stronger negative correlation between Chl a and SWS in the East China Sea, whereas both Eastern Coastal Waters of the United States and the Amazon Estuary exhibited insignificant correlations.

In contrast, in the two enclosed seas, Chl a exhibited only weak or negligible correlations with SST, SWS, and PAR, indicating that environmental factors exerted a comparatively minor influence on Chl a dynamics in these semi-enclosed basins.

Notably, varying degrees of intercorrelations among the three indices were indicated in the five sea areas. Distinct positive correlation between SST and PAR were indicated except in the Amazon Estuary. In addition, various degrees of negative correlation between PAR and SWS were indicated with the correlation coefficient of the East China Sea > the Amazon Estuary > the Eastern Coastal Waters of the United States ≈ the Bohai Sea, whereas the Gulf of Mexico demonstrated weak positive correlation. As for the correlations between SST and SWS, we observed a strong negative correlation only in the Amazon Estuary, whereas the other four sea areas exhibited weakly negative or negative correlations.

To clarify the key factors dominating Chl a concentration, we conducted PCA for each study area. As shown in Table 6, SST had a relatively high loading on the first principal component across all five sea areas. In the East China Sea, PAR had a loading comparable to SST, whereas in the Eastern Coastal Waters of the United States, PAR showed a higher loading than SST. In the Bohai Sea, PAR had a slightly lower loading than SST, which was much lower than SST in the Gulf of Mexico and the Amazon Estuary.

These results suggested that seasonal SST variation in the Eastern Coastal Waters of the United States, and the Bohai Sea, was closely associated with seasonal PAR variation. In contrast, in the Amazon Estuary, sustained solar irradiance near the equator may have weakened the coupling between SST and PAR, making SST more susceptible to oceanic water intrusion and hydrological processes. In the Gulf of Mexico, SST variation may be strongly influenced by ocean circulation, including the Gulf Stream, explaining the lower PAR loading on the first principal component. SWS showed larger loadings on the first principal component in the three open-sea areas than in the Bohai Sea and the Gulf of Mexico, possibly reflecting wind-driven vertical mixing and nutrient input from deeper waters.

In general, the loading on the first principal component was likely related to seasonal variation of SST, PAR, and SWS. The enclosed-sea areas of the Bohai Sea and the Gulf of Mexico were more susceptible to seasonal SST variation and less susceptible to SWS. For the open-sea areas of the Eastern Coastal Waters of the United States and the East China Sea, both the seasonal variations of SST and PAR have been important in enhancing phytoplankton biomass. In contrast, for the Amazon Estuary of Brazil, which is located at the equator, the guarantee of solar irradiance has made SST independent from PAR, and PAR has less significantly affected Chl a than SST and SWS.

The regulation of Chl a concentrations across the five study areas is due to a complex interplay between physical forcing and biogeochemical cycles. The dominant driver has shifted from climatic factors in open oceans to anthropogenic nutrients in enclosed seas.

The results of this study revealed two contrasting Chl a responses to rising SST, which appeared to have been mediated by local environmental settings and nutrient availability. Generally, in widespread areas that showed positive correlations between SST and Chl a during stable SST periods (Fig. 9), this correlation appeared to be related to seasonal SST variation. Additional anthropogenic nutrient input combined with the rising SST may have stimulated phytoplankton proliferation, leading to increasing Chl a along the coast in all five study areas. Despite the control of pollutant emissions to the Bohai Sea and the East China Sea, we observed a continuous increase in Chl a in the central of the Bohai Sea and the Changjiang River Plume in the East China Sea, although there was reduced Chl a along coasts and estuaries, as shown in Fig. 6.

Meanwhile, a significant inhibitory effect of SST rise on Chl a during typical warming periods was indicated. We observed negative correlations between Chl a and SST in most offshore areas, which was consistent with saline seawater intrusion and the influence of large current systems, such as the Taiwan Warm Current in the East China Sea, the Gulf Stream along the Eastern Coastal Waters of the United States, and the North Brazilian Current in the Amazon Estuary. The negative impacts from rising SST on nutrient supply, however, have been concealed by land-derived nutrient-induced Chl a enhancement, especially in nearshore areas, and in the plumes of estuaries in the East China Sea and the Amazon Estuary. Notably, in the Amazon Estuary, despite a significant increase in Chl a in nearshore areas in 2020 compared with 2018, the overall Chl a concentrations followed a decreasing trend. This has meant that the falling trend in Chl a in the offshore region because of global warming effects may no longer be compensated for by additional anthropogenic nutrient input combined with rising SST in the coastal region and the estuary.

For the enclosed-sea areas, the Gulf of Mexico has exhibited a much lower eutrophication level than the Bohai Sea. Its larger spatial extent and stronger connectivity with the Atlantic Ocean, which has been mediated by the Gulf Stream and Cuban Current, enhanced water mass exchange and amplified warming impacts. Consequently, the Gulf of Mexico behaved more like an open-sea system, displaying significant Chl a declines during SST increases (Fig. 9). These opposing responses have underscored the need to better resolve subregional physical drivers, such as currents and upwelling, that may synergize with warming to shape Chl a trajectories.

The impact of global warming on Chl a concentrations was evident in all three open-sea areas, namely the Amazon Estuary, the East China Sea, and the Eastern Coastal Waters of the United States. We also observed significant decreases in offshore Chl a levels, strongly supporting a linkage to global warming rather than human activity. If SSTs continue to rise and the contribution of riverine nutrient input continues to decrease as a result of long-term pollution control measures, this ecological effect may extend to nearshore areas. Additionally, in the enclosed-sea area of the Gulf of Mexico with medium eutrophic characteristics, the negative influence of global warming on Chl a has also been identified. Rising SSTs and oceanic currents have led to a marked decline in Chl a, particularly in areas where ocean currents enter the Gulf of Mexico. The negative correlation between SST and Chl a in the central Gulf further supports the detrimental impact of global warming on Chl a levels.

The impacts of rising SSTs under global warming in estuaries and along coastal waters, however, have been less pervasive than those from eutrophication. Eutrophication-induced phytoplankton blooms have been expanding into offshore sea areas, although Chl a levels may decline along the coast because of restrictions on anthropogenic nutrient input. In highly eutrophic confined sea areas, such as the Bohai Sea, Chl a levels appear to have been less impacted by rising SSTs under global warming. Despite a decline in Chl a concentrations in coastal waters of the Bohai Sea, the central basin maintains high levels of Chl a because of anthropogenic nutrient input, which currently masks the potential impact of global warming. This lagging effect may be temporary and subject to future nonlinear threshold responses. In the highly eutrophic open sea of the East China Sea, although there has been a significant decline in the rate of Chl a increase in estuaries and along the coast, Chl a levels in the Changjiang River Plume have continued to rise. This suggests that in sea areas with sufficient nutrients and suitable salinity, rising SST may further stimulate HABs. A similar phenomenon has been observed in the plume of the Amazon Estuary, where the area of high Chl a has expanded. Despite the widespread occurrence of decreasing Chl a offshore, the continuous proliferation of phytoplankton in the plume has compensated for the intrusion of warmer and oligotrophic pelagic waters.

The findings of this study contribute to our understanding of the ecological response of marine environments to both human activity and global warming. By continuously controlling the input of land-derived nutrients, the frequency of HABs can be reduced, but high Chl a levels may persist in estuaries and along the coast over the long term. From a broader perspective, nearshore areas will gradually experience greater impacts on nutrient supply from global warming. This study (1) has demonstrated the complex and spatially variable interactions between primary production, SST, and nutrient input and exchange under the dual pressures of changes in human activity and coastal development combined with global warming; and (2) has highlighted the need for further research on the impact of changing coastal and terrestrial nutrient input, including where these have been reduced because of environmental protection measures, on primary productivity in both enclosed- and open-sea areas. Unresolved physical–biogeochemical feedback, such as warming-driven stratification and deoxygenation, have limited mechanistic interpretations of these trends. Framing future research within the Global Carbon Project, GO₂NE, and marine ecosystem regime shift theory will help determine whether marginal seas are approaching ecological tipping points. Linking such insights to Sustainable Development Goal 14 and the United Nations Ocean Decade will be essential for adaptive coastal management and sustainable blue economy strategies under accelerating global change.