Instead, weakened coastal upwelling was associated with reduced chlorophyll off southeastern Vietnam, whereas wind-regulated retention of low-salinity river water coincided with higher chlorophyll on the northern shelf. An information-theoretic analysis found that cumulative heat stress provided no unique information within the selected driver sets in either region, although it remained involved in joint interactions with other physical variables. The research combined satellite observations, ocean and atmospheric reanalyses, and a biogeochemical hindcast to examine the periods from 1998 through 2024.

Sea-surface temperature came from NOAA’s Optimum Interpolation Sea Surface Temperature version 2.1 product at 0.25-degree resolution, and surface chlorophyll a was obtained from the Copernicus Marine Environment Monitoring Service GlobColour Level 4 merged ocean-colour product. Upper-ocean temperature, salinity, currents and mixed-layer depth came from the GLORYS12V1 eddy-resolving reanalysis at 1/12-degree resolution. ERA5 supplied 10-metre winds, surface downward shortwave radiation, precipitation, 500-hPa geopotential height and 850-hPa winds.

Vertically resolved chlorophyll a and supporting upper-ocean diagnostics came from the Global Ocean Biogeochemistry Hindcast and other Copernicus products. The fields were placed on a common 0.25-degree grid, and the variables were standardized before the attribution analysis. Marine heatwaves were identified using the Hobday framework: daily sea-surface temperature had to exceed the seasonally varying 90th percentile for at least five consecutive days, and events separated by no more than two days were merged.

Thresholds were calculated from a fixed 1982–2011 climatological baseline. As a result, the calculated frequencies for 1998–2024 do not adjust for long-term background warming. The study determined annual occurrence frequency, total heatwave days, mean event duration and cumulative intensity.

The Pearl River Estuary was a relatively high-frequency environment, with annual occurrence ranging from 3.41 to 5.37 events and a median of 4.57 events per year, compared with a basin-wide median of 3.78. The Vietnam Upwelling Area was distinguished by longer events, with a median duration of 16.63 days per event against 14.30 days across the basin. Both regions also had significantly higher median annual heatwave days and cumulative intensity than the basin-wide medians, with comparisons yielding p<0.01.

The chlorophyll a response varied across both season and location. Negative anomalies predominated in the South China Sea during spring, autumn and winter, but the summer composite showed a pronounced regional divide: positive anomalies in the Pearl River Estuary and adjacent northern shelf waters and negative anomalies in the western South China Sea. The summer anomaly patterns in the two regions also lay outside the variability observed under comparable non-heatwave summer conditions.

Because physical and chlorophyll a conditions vary strongly through the year, combining marine heatwaves from multiple seasons could obscure this contrast, so the mechanistic analysis concentrated on summer. Chlorophyll a is used as an indicator of phytoplankton conditions, but it is not a direct measurement of phytoplankton biomass or primary productivity. Moreover, the analysis concerned surface chlorophyll a and did not establish changes in the wider marine food web or in fisheries productivity.

To separate overlapping physical influences, the researchers applied Synergistic–Unique–Redundant Decomposition, or SURD. Grounded in Shannon information theory, the method measures how uncertainty in a future chlorophyll a anomaly is reduced when the past states of selected drivers are known. It partitions the captured information into unique contributions from individual drivers, redundant information shared by drivers, and synergistic information available only from their joint states.

Within this framework, “causality” refers to lagged information-theoretic predictive causality under the chosen variables and response interval, rather than direct experimental causation. The causality leak represents target uncertainty not assigned to the selected drivers and may reflect missing processes, observational uncertainty, finite-sample effects or data discretization. The reported unique, redundant and synergistic percentages are relative contributions within the information captured by each selected set, not fractions of total chlorophyll a variance or target entropy.

For the Vietnam Upwelling Area, the selected predictors were sea-surface salinity, sea-level anomaly, Ekman pumping velocity and cumulative sea-surface temperature anomaly. Ekman pumping velocity served as an indicator of wind-driven vertical transport. Among these variables, Ekman pumping provided the largest unique contribution, accounting for 12.9% of the captured non-leak information, while salinity and Ekman pumping shared 18.9% as redundant information.

Their common signal was consistent with weaker upwelling: reduced upward transport of saline subsurface water was associated with surface freshening. A four-variable synergy involving salinity, sea level, Ekman pumping and cumulative heat stress represented 20.9% of the captured information, the largest single non-leak term. The absence of a unique contribution from accumulated heat does not indicate that thermal effects were irrelevant, because temperature remained involved in the joint terms.

Instead, the decomposition identified vertical dynamics as the dominant separable physical signal associated with declining surface chlorophyll in the selected driver set. The Vietnam Upwelling Area has a climatological summer belt of relatively cool surface water, enhanced Ekman pumping, lower coastal sea level and high surface chlorophyll. In the biogeochemical hindcast, the climatological chlorophyll maximum lay at approximately 50–70 metres, and summer upwelling normally carried high-chlorophyll water toward the surface.

During marine heatwaves, positive temperature anomalies extended into the subsurface and strengthened stratification, while positive sea-level anomalies indicated thermocline deepening. At the same time, negative Ekman pumping anomalies coincided with the core of the chlorophyll decline, consistent with weaker wind-driven upwelling. Reduced upwelling had two connected physical consequences: less cold subsurface water reached the surface, limiting surface cooling, and less nutrient-rich deep water was transported into the euphotic zone.

The hindcast’s water-column mean anomaly was only −0.006 mg m−3, much smaller than its −0.077 mg m−3 surface anomaly, supporting a contribution from vertical redistribution. Nevertheless, negative anomalies extended through roughly the upper 60 metres, showing that redistribution alone may not explain the full surface response. In the Pearl River Estuary, the selected drivers were sea-surface salinity, sea-level anomaly, cumulative sea-surface temperature anomaly and the zonal wind component.

Salinity provided the largest unique contribution, at 28.1% of the captured information, while the joint state of all four variables supplied the largest term, at 30.6%. Accumulated heat again had no detectable unique contribution. The salinity pattern was consistent with enhanced nearshore retention of low-salinity Pearl River plume water: the strongest negative salinity anomalies occurred within the plume-affected area, while more saline offshore water did not show enough climatological or anomalous intrusion to explain that pattern.

Local precipitation did not increase during the heatwaves, and neither did precipitation over the nearby Pearl River Delta around regional heatwave establishment, arguing against anomalous rainfall as the main source of freshening. Instead, weakened winds reduced offshore Ekman transport, and elevated nearshore sea level could have narrowed the estuary-to-shelf water-level gradient, slowing the plume’s offshore spread. The retained fresh layer lowered surface density and strengthened salinity-driven stratification.

In this nutrient-rich estuarine system, the authors inferred that plume retention may have helped preserve favourable nutrient conditions while heat and stratification were associated with increased surface chlorophyll. Consistent daily nutrient and biological datasets with suitable spatial coverage were unavailable for both regions over 1998–2024, so nutrients were not included in the SURD driver pool. The estuarine explanation is therefore physically consistent but not directly demonstrated.

Considerable uncertainty also remained unexplained in the Pearl River Estuary, where optical complexity, suspended sediment, light limitation, grazing, changes in phytoplankton communities and satellite retrieval errors may affect surface chlorophyll independently of physical transport. Satellite chlorophyll can also change through physiological photoacclimation without a proportional change in phytoplankton biomass, further limiting biological interpretation. The selected physical drivers captured a larger share of uncertainty in the Vietnam Upwelling Area, where the surface response was more closely aligned with wind-driven vertical dynamics.

Both regions shared a similar large-scale atmospheric background: an intensified and westward-extended western North Pacific subtropical high, associated with stronger incoming shortwave radiation and weaker summer monsoon winds. Rather than directly determining the chlorophyll response, this common circulation pattern coincided with different regional pathways. The results indicate that marine heatwaves in marginal seas cannot be interpreted through surface temperature alone, and that direct nutrient, optical and plankton observations are needed to test the inferred mechanisms more fully.

Subject of Research: Regional physical controls on surface chlorophyll responses to South China Sea marine heatwaves

Article Title: Contrasting physical drivers of chlorophyll a responses to summer marine heatwaves in the South China Sea

Article References: Fang, Y., Liu, C., Chen, Y., Li, P., & Li, M. (2026). Contrasting physical drivers of chlorophyll a responses to summer marine heatwaves in the South China Sea. Ocean Science, 22(5), 3347-3367. https://doi.org/10.5194/os-22-3347-2026

Image Credits: AI Generated

DOI: 10.5194/os-22-3347-2026

Keywords: marine heatwaves, South China Sea, surface chlorophyll a, Pearl River Estuary, Vietnam upwelling, Ekman pumping, river plume retention, ocean stratification, information-theoretic attribution, physical-biological coupling, Contrasting, physical