The data presented in this paper comprise 30 stations sampled over a total of 5 days (January 8–12, 2020) in three geographically distinct areas of the PRE and adjacent sea (21–23°N and 113–115°E). Including the Huangmao sea (HMS) subregion, which is located on the west side of PRE and the Daya bay transect (DYB), which is located on the easternmost transect, are located away from the mouth of the PRE. The first transect is located at the PRE. The other two transects are nearshore environments on the west/east side of the PRE, respectively.

All three transects used the same sampling strategy by collecting water samples from a CTD Rosette with Niskin bottles and deploying optical instruments at each station.

The mean circulation over the PRE and the adjacent continental shelf is dynamic and heavily influenced by winds, tides, and buoyant effects of river discharge. Tidal forces, primarily semidiurnal (M2) and diurnal (K1) forces, dominate the region and reach a magnitude of approximately 1.0 m within the PRE, as outlined by Zu and Gan (2015). In terms of broader seasonal circulation patterns around the northern shelf of the South China Sea (SCS), previous studies have provided a clear understanding of the effects of monsoonal winds (Fang et al., 1998; Xue and Chai, 2002; Wong et al., 2003; Gan et al., 2009a) ( Figure 1A ). During the summer months, the East Asian summer monsoon drives a north-eastward coastal current along the PRE coast. Conversely, in winter, when the northeasterly monsoon takes effect, the surface currents shift, flowing southwestward across the continental shelf (illustrated in Figure 1B ). The surface temperatures and currents during these periods have been investigated, with particular attention given to the northern SCS.

Water samples were collected at discrete depths (0, 5, 10 and 20 m) via a CTD rosette multi-bottle array system. Water samples of approximately 0.5–1 L were filtered using a 0.7–μm glass fiber filter (Whatman GF/F), and filters were immediately stored in liquid nitrogen. Then those water samples were analyzed for: (1) Chl and selected pigments via High Performance Liquid Chromatography (HPLC), (2) particulate organic carbon (POC) via an elemental analyzer (Flash EA-Delta V), and (3) suspended particulate matter (SPM) that is determined gravimetrically following the protocol of Tilstone et al. (2002), based on van der Linde (1998) and Neukermans et al. (2012). (4) Dissolved organic carbon (DOC) via an OI700 Analytical total organic carbon analyzer by wet-oxidation. (5) Water samples for nutrient analysis were filtered through GF/F filters onboard and frozen immediately at − 20°C until analyzed. After thawing at room temperature in the laboratory, they were analyzed by an AA3 nutrient auto-analyzer using colorimetric methods (Knap et al., 1996) with detection limits of 0.02 and 0.02 μmol L⁻¹ for nitrate plus nitrite and soluble reactive phosphate, respectively.

Immediately after CTD casts optical profiles of the water column were obtained via spectral optical sensors (WET Labs ac-s absorption-attenuation meter, ECO BB9 backscattering meter and Sequoia LISST–100X type C, hereafter referred to as ac-s, BB9 and LISST).

A 10–cm path length ac-s was used for the acquisition of in situ profiles of the total absorption and attenuation coefficients. The ac-s was cleaned and calibrated in the laboratory following the ac-s User’s Guide before and after each campaign. The absorption and attenuation signals were corrected for temperature and salinity effects according to Pegau et al. (1997). Correction for incomplete recovery of the scattered light in the absorption tube of the ac-s was performed via the proportional method described by Zaneveld et al. (1994). BB9 was coupled with the ac-s in situ measurements of the total backscattering coefficient [bp(λ)] at seven wavelengths (412, 440, 488, 530, 595, 695, and 714 nm), as were the CDOM and chlorophyll fluorescence.

All particle size distribution (PSD) measurements were performed in this study using a LISST—a commercially available instrument that measures the light scattering of a particle suspension at small forward angles and uses this information to estimate the PSD (Agrawal and Pottsmith, 2000). The particle volume concentrations in the 36 size bins (with a size range of 2.5–500 μm) were processed via the LISST-SOPv5.1 software program provided by the manufacturer. The total volume concentration (VC), total cross-sectional area concentration (AC), mean particle size (DA), median particle diameter of the volume distribution ( D v 50 ) and slope of the PSD (ξ) were subsequently calculated. The specific calculation process for the LISST measurements can be found in Wang et al. (2016).

Chl fluorescence measured by WET Labs ECO BB9 was further validated on the basis of paired HPLC Chl (Chl_HPLC) measurements and fluorescence (Chl_Fluo) match-ups. The point-by-point comparisons for paired observations revealed little noise in the data distribution, except for a few points in low-value regions ( Figure 2A ). To reduce the uncertainty of Chl_Fluo, it was calibrated against the Chl_HPLC data.

Beam attenuation coefficients are important internal optical properties of seawater and are widely used to explore the characteristics of particulate matter, e.g., the particulate organic carbon (Stramska and Stramski, 2005). Here we investigated the relationship between the c_p(660) and POC, which suggested that the c_p(660) can be used to quantitatively estimate the POC in the GBA dry season ( Figure 2B ). Using the empirical relationships developed in this study, we characterized the vertical distribution of POC in the GBA on the basis of in situ c_p(660) measurements.

Satellite products were used to analyze the spatial and temporal variability in the physical and dynamic characteristics of the northern SCS. Ocean Surface Current Analyses Real-time (OSCAR) provides global surface current products directly calculated from satellite fields, including sea surface height, wind, and sea surface temperature (SST) (Bonjean and Lagerloef, 2002). The datasets are available at 5–day intervals, at 1/4°×1/4° spatial resolution and are obtained from the National Oceanic and Atmospheric Administration (NOAA) site (www.oscar.noaa.gov). Daily 4 km Chl and SST composites were calculated from MODIS/Aqua Level 1 granules downloaded from the National Aeronautics Space Administration (NASA) Ocean Color website (https://oceancolor.gsfc.nasa.gov). Coincident daily MODIS Chl and SST data (level 2) at the original resolution (approximately 1 km) were also obtained from the same NASA group.

We conducted hydrographic surveys along three cross-shelf transects in the PRE and adjacent seas to investigate the distribution of hydrographic properties ( Figures 1C , 3 ). During the winter, the seawater in the GBA experienced vertical mixing and horizontal stratification. This phenomenon was attributed primarily to the low river discharge and the vigorous stirring and mixing induced by strong northeasterly winds. Three cross-shelf transects were all shown a pronounced horizontal density gradient ( Figures 3G–I ).

The hydrographic vertical section of the DYB was shown considerable mixing from the surface to the bottom with temperature values varying between 19 and 22°C and salinity values fluctuating between 31 and 33 psu ( Figures 3A, D ). Notably, a very narrow flow band was found between 20 and 60 km offshore of the DYB, characterized by significantly lower temperatures (19 ± 0.2°C) and maximum Chl and POC (0.5 mg m⁻³ Chl and POC values of approximately 100 mg m⁻³, respectively). The temperatures greatly increased offshore (to ~23°C), which was accompanied by a rapid decline in the Chl to ~0.1 mg m⁻³. This distinct temperature and Chl front were also evident in the MODIS/Aqua L2 Chl and SST products on January 8, 2020 ( Supplementary Figure S1 ). The front consistently separated nearshore waters with lower temperatures (~19°C) and higher Chl (~0.5 mg m⁻³) from offshore waters with higher temperatures (~23°C) and lower Chl (~0.1 mg m⁻³) levels.

In the PRE section, the distribution of horizontal salinity gradients also trended positively offshore is largely due to the influence of freshwater runoff occurring near the coast, with salinity ranging from 5–32 psu. Moreover, the concentration gradients of key biogeochemical parameters, namely, Chl, POC, and CODM, consistently were greater near shore. Specifically, the concentrations of Chl, POC, and CODM were greater near the coast and gradually decreased as we moved offshore ( Figures 4B, E, H ).

The horizontal density gradient observed along the HMS transect was driven primarily by variations in salinity ( Figures 3F, I ), whereas the temperature exhibited a uniform mixing pattern in the horizontal direction ( Figure 3C ). The salinity gradient was strongly influenced by freshwater runoff near the coast. Moving offshore, there was a gradual decrease in the concentration of both Chl and CDOM, in contrast to the trend in the PRE. The changes in POC demonstrated a similar trend to those in Chl and CDOM ( Figures 4C, F, I ).

Three distinct water masses were identified on the basis of their temperature and salinity characteristics. The temperature versus salinity (T–S) diagrams, derived from field measurements, revealed temperature values ranging from 19 to 23°C and salinity values ranging from 7 to 33.5 psu in the PRE and adjacent seas. T–S diagrams provided valuable insights into the characteristics of three end-member water masses ( Figure 5 ): estuarine diluted water masses (EDW), water masses influenced by the GCC (GCCW), and dense shelf water masses (DSW).

The EDW represented a mixture of the three end-member water masses, exhibiting a broad range of salinity values spanning from 10 to 30 psu. On the other hand, the GCCW was characterized by cold temperatures, with an average temperature of only 19.45°C. The DSW stands out due to its relatively high temperature and salinity, with average values of 21.14°C and 33.54 psu, respectively.

In three stations biogeochemical signatures were particularly pronounced, including station A3, closest to the PRE, station B7, a mid-shelf station and station A14, an offshore station. The lower salinity surface plume waters of station A3 had elevated concentrations of Chl and CDOM, with Chl of approximately 3 mg m⁻³ ( Figure 6A ). Conversely, the weakly stratified offshore station (A14) had Chl of less than 0.4 mg m⁻³. Notably, the lower salinity site (A3) presented higher Chl and CDOM at the surface, resulting in significant attenuation of the light field ( Figures 6A, C ). This attenuation may have implications for primary production, potentially restricting it to near-surface waters.

At the mid-shelf station (B7), relatively high and vertically uniform Chl were observed ( Figure 6B ). The surface layer at this station had a low water temperature of 19.4°C. Similarly, the weakly stratified surface layer at station A14 ( Figure 6C ) presented a low surface Chl of approximately 0.3 mg m⁻³. However, as depth increased, the Chl gradually increased, peaking at approximately 0.9 mg m⁻³ at a depth of 30 m.

The vertical profiles of the total attenuation coefficients, particulate backscattering coefficient median particle diameter and particle size distribution (PSD) slopes ( D v 50 and ξ) ( Figures 6D–F ) at the same three representative stations revealed strong horizontal and vertical variability. These high values of attenuation and backscattering were all associated with lower-salinity waters. At the stations near the PRE (A3), and at the c_p(660), the values were greater than those at the other stations, which increased with depth, with maximum values recorded at the bottom. At the offshore stations, surface values of c_p(660) and b_bp(440) sharply decreased to 0.35 m⁻¹ and 0.003 m⁻¹ (A14), respectively.

The distributions of biogeochemical properties across the different water layers of the three distinct water masses are shown in Figure 7 . Generally, there was a decreasing trend in the concentrations of all biogeochemical properties from coastal to offshore regions, with no significant variation observed with depth. The EDW had the highest concentrations of POC, DOC, and nutrients (DIN and DIP), highlighting its unique estuarine characteristics. Interestingly, compared with the other two water masses, the GCCW had higher Chl ( Figure 7A ; Table 2 ). This phenomenon can be attributed to the GCC transporting nutrient-rich seawater from the north ( Figures 7G, H ) (Hu et al., 2023). Furthermore, compared with the other two water masses, the GCCW had lower DOC (0.83 ± 0.23 mg m⁻³) ( Figure 7C ; Table 2 ).

As particle size distributions are highly important for both optical properties in the water column and are indicative of phytoplankton size-classes, they were analyzed for each water mass. The total volume concentration (VC) and total cross-sectional area concentration of particle (AC) of the EDW were higher than those of the other water masses, with average values of 71.76 μL L⁻¹ and 2.55 m⁻² ( Table 2 ), respectively, and the values were greater closer to the bottom layer ( Figures 8A, B ). The mean particle diameter weighted by area (DA) and D v 50 gradually increased with increasing distance from the shore, and the difference within the water layer was small ( Figures 8C, D ). This distribution of DA characteristics revealed that the particulate matter in the water body was dominated by phytoplankton in the offshore regions, whereas the suspended sediment particles were dominant near the coast. The PSD slopes in the EDW region were relatively low (< −3.8), whereas the values were high in the GCCW and DSW regions (> −3.0) ( Figure 8E ).

The distributions of particle scattering and backscattering at 530 nm were studied in the three water masses ( Figure 9 ; Table 2 ). The ratio of backscattering by particles to total scattering by particles is an indicator of particle composition, for example, phytoplankton and detrital material have a low index of refraction relative to inorganic minerals (Boss et al., 2004). Both b_p(530) and b_bp(530) decreased from the coast to offshore ( Figures 9A, B ). The backscattering ratio of the surface layer ( Figure 9C ) decreased from the estuary to the open sea, indicating that the concentrations of inorganic minerals gradually decreased.

Figure 10 shows ternary plots representing the distributions of pico-, nano-, and micro-phytoplankton percentages in the GBA based on the basis of the methodology of Brewin et al. (2015). In winter, the phytoplankton communities in the GBA were predominantly composed of micro- and/or nano-phytoplankton. Their contribution to the Chl exhibited significant variability, ranging from over 80% in the GCCW to less than 60% in the DSW ( Table 2 ). Although pico-phytoplankton contributed mainly to the surface layer of the DSW during winter, their proportion of Chl remained below 30% in most samples.

The substantial fraction of micro-phytoplankton and the low ratio of POC to Chl in the GCCW indicated the dominance of large phytoplankton, such as diatoms ( Figure 10B ; Supplementary Figure S2 ). Conversely, DSW exhibited the opposite pattern, characterized by a large proportion of pico-phytoplankton and high POC: Chl values, indicating the prevalence of small phytoplankton in the phytoplankton biomass ( Figure 10B ; Supplementary Figure S2 ).

Physical processes, including summer upwelling, alongside the Guangdong coastal currents and tides, exerted a substantial influence on bio-optical and biogeochemical properties in the PRE and adjacent waters. Over the last few decades, many studies have focused on the summer upwelling phenomena and the interactions between the Pearl River plume and the coastal currents (Gan et al., 2009b, 2010; Gu et al., 2012). In addition, tidal dynamics induce variations in nutrient supply, light availability, and sediment resuspension, impacting the growth and composition of phytoplankton communities (Tao et al., 2020). However, few studies have investigated the vertical distribution of bio-optical properties and their response to the Guangdong coastal currents during the winter dry season in the GBA.

The GCC, characterized by strong currents along the coast, can transport nutrients, suspended particles, and dissolved matter, strongly influencing the distribution and composition of phytoplankton and particulate matter (Yang and Ye, 2021; Hu et al., 2023). The GCC also can frequently bring strong ocean fronts, which play crucial roles in dynamic and chemical processes (Yang and Ye, 2021). These ocean fronts, characterized by sharp gradients in temperature, salinity, and other properties, serve as dynamic boundaries where water masses with different characteristics interact. During our investigation, the study region was shown obvious frontal structure, with the estuary area related to the river plume, and the shelf area related to the GCC in the open sea. During our investigation, we observed a clear frontal structure in the study region, evidenced by both temperature and salinity distributions ( Figure 4 ) as well as remote sensing images ( Supplementary Figure S 1 ). The estuarine region was influenced by the river plume, while the shelf area was associated with the GCC in the open sea. The interactions at these fronts trigger intense mixing and exchange of nutrients, organic matter, and other chemical compounds, influencing the distribution and productivity of marine ecosystems. In our study, we identified three distinct water masses with markedly different bio-optical and biogeochemical properties ( Figure 7 ; Table 2 ).

On the other hand, the Pearl River, as a major freshwater source, influences the salinity structure and nutrient input, shaping the spatial and temporal patterns of phytoplankton and biogeochemical properties. Our analysis of nutrient concentrations with salinity indicated that dissolved inorganic nutrient concentrations decreased as salinity increased ( Supplementary Figure S3 ). The linear correlations were significant at the 95% confidence level, and the correlation coefficients ranged from 0.55–0.77 (N=98), suggesting that terrestrial input is the major cause of high dissolved inorganic nutrient levels in coastal waters. In contrast, the relationships between nutrients and temperature were very weak.

The variability of bio-optical properties in the PRE is closely related to tidal forces. These tidal forces are primarily semidiurnal (M2) and diurnal (K1) forces, reaching amplitudes of approximately 1.0 m within the estuary (Zu and Gan, 2015). However, our sampling frequency is quite low, lacking hourly observations, which makes it challenging to differentiate the effects of tides from other processes. In the future, we should increase the sampling frequency to better differentiate the effects of tidal currents from other physical processes that influence changes in bio-optical properties.

The findings of this study have significant implications for the development and refinement of bio-optical algorithms used to retrieve key constituents of seawater in complex coastal environments, such as the GBA. The SPM exhibited substantial spatial variations (up to 23.5 mg m⁻³ in the EDW), which can significantly influence the accuracy of retrieving optical water quality parameters from ocean color algorithms in these regions (Qin et al., 2007; Blondeau-Patissier et al., 2009). Another major constituent is CDOM, which can be present in high concentrations in the GBA ( Figure 4 ). Remote sensing ocean color algorithms for estimating Chl often lack sufficient correction for CDOM effects (Carder et al., 1999). These algorithms typically use constant values for the spectral slope coefficient, assuming that it varies unpredictably (Maritorena and Siegel, 2005; Blondeau-Patissier et al., 2009). Our findings indicate that the modelling and development of ocean color algorithms based on the assumption of constant average values for parameters such as the backscattering power law coefficient or spectral slopes of CDOM, which are commonly used in global algorithms, limit their applicability in similar coastal regions.

In recent years, the occurrence of hypoxia in the PRE and adjacent waters has been associated with high primary production and algal blooms due to increased nutrient concentrations in waters discharged by the Pearl River (Wang et al., 2010; Zhao et al., 2021). This highlights the importance of reliable monitoring of algal blooms, which requires improved knowledge of the optical characteristics of seawater constituents, as presented in this text. Furthermore, the development of better optical models that relate these constituents to inherent optical properties such as absorption and backscattering, as well as apparent optical properties such as remote sensing reflectance, can be improved from our results (D’Sa et al., 2006). The complete coverage of the optical diversity of as many natural waters as possible is also important for the classification of optical water types (Shanmugam, 2011; Bi et al., 2023).