Dissolved and Particulate Beryllium Isotopes in the Pearl River Estuary: Their Geochemical Behavior in Estuarine Water and Potential Contributions From Anthropogenic Sources

The ratio of cosmogenic 10Be and its stable isotope 9Be has been used as a proxy of long-term continental weathering fluxes and denudation rates, but transport processes of these isotopes from river water to estuarine water and seawater, as well as interference of potential anthropogenic source of 9Be on natural 10Be/9Be around populated estuaries are not well constrained. Here, we present results of 10Be and 9Be concentrations of dissolved and reactive particulate phase in the Pearl River Estuary (PRE) and its eight major outlets. The concentrations of Cu, Cd, and Pb are also measured, allowing us to assess their contamination levels and anthropogenic source together with 9Be by the geo-accumulation index (Igeo–reac) and enrichment factor (EF). A wide distribution pattern of dissolved 10Be (137–1,194 at/gwater) and 9Be (0.781–8.31 × 10–12 g/gwater) among these outlets is observed. The distribution coefficients (Kd) of both isotopes between sediment and water are in the order of 105, and on average only 5% of 10Be exists as dissolved form. Compared with total meteoric 10Be deposited on the river basin, 23% of the meteoric 10Be is retained while 38% of 10Be finally escape the estuary and is transported into coastal seawater. Despite the high contamination levels of Cu and Cd, the lower Igeo–reac and EF values of 9Be indicate that 9Be is hardly polluted by anthropogenic source. Thus, the 10Be/9Be in the PRE area is mainly associated with natural processes instead of human activities.


INTRODUCTION
Beryllium isotopes have played an important role in many geoscience studies. For example, the 10 Be/ 9 Be ratio in authigenic phase of marine sediment has been used as a tool to reconstruct the history of Earth's magnetic fields using the inverse relationship between production rate and geomagnetic dipole values (Frank et al., 1997;Christl et al., 2003;Valet et al., 2014; three decades and there has been no new data in this area. It is unknown if anthropogenic beryllium is released together with other heavy metals which have polluted the PRE (Zhuang et al., 2018). If so, the normalization protocol of 10 Be by 9 Be in many studies, such as Earth surface erosion processes and the reconstruction of the geomagnetic field parameters, would be challenged. Since beryllium is highly particle-reactive, both dissolved and particulate data of 10 Be and 9 Be from the PRE are needed in order to gain a full picture of their sources and behavior in transport process.
In this study, we measure distributions of 10 Be and 9 Be, along with heavy metals (Cu, Cd, and Pb) in dissolved and reactive phases of suspended sediments among eight outlets of the PR and within the Lingdingyang Bay, to assess the partition of beryllium isotopes between particles and water, the export flux of riverine 10 Be into the South China Sea (SCS), and evaluate impacts of potential anthropogenic 9 Be around industrial cities on 10 Be/ 9 Be ratios in the estuarine waters.

Study Site
The annual discharge of the major tributaries of the PR, the West River, the North River, and the East River, is 219.7, 42.1, and 23.4 × 10 9 m 3 year −1 , respectively (Dai et al., 2009). The suspended sediment flux of the West River (66.8 × 10 6 t year −1 ) is higher than those of the North River (5.4 × 10 6 t year −1 ) and the East River (2.5 × 10 6 t year −1 ), disproportionately higher than the discharge ratio (Zhang et al., 2011). The West River basin is typically a karst area dominated by carbonate (Cai et al., 2008).
The PR system drains into the northern SCS through eight outlets (river mouths), including Humen (HM), Jiaomen (JM), Hongqimen (HQ), and Hengmen (HE) in the northern part, and, Modaomen (MD), Jitimen (JT), Hutiaomen (HT), and Yamen (YM) in the south (Figure 1). The broadly defined PRE consists of three sub-estuaries, Lingdingyang, Modaomen, and Huangmaohai. Four of the outlets (HM, JM, HQ, and HE) sit along the western shore of the northern part of the Lingdingyang Bay. The Lingdingyang Bay receives river water mainly from the East River, the North River, and small branches of the West River, all of which account for more than half of the PR discharge (Pan et al., 2020). The Huangmaohai Bay sits in the west of the PRE, with two branches of the West River injected via Hutiaomen and Yamen. Between these two bell-shaped sub-estuaries sits the shallow Modaomen Bay, which also receives the water of West River and empties into the northern SCS via Modaomen and Jitimen outlets.

Sample Collection
The river waters and estuarine waters of the PRE were sampled during May 3, 2017 and May 8, 2017 by fish boats. Sampling stations include eight outlets of the PR, and a meridional section of four stations inside the Lingdingyang Bay (Figure 1). At each station, waters were sampled by a 3 L home-made polymethyl methacrylate sampler attached with an iron block at the middle point of river. The sampler was launched repeatedly until the FIGURE 1 | Study area and sampling sites of the PRE. The sampling sites at eight outlets and the Lingdingyang Bay are marked with red triangles and blue squares, respectively. total amount of ∼20 L water for each sample was collected. Both surface (0.5 m depth) and bottom (above 0.5-1.5 m of river bed) waters were sampled and collected in acid-cleaned HDPE bottle. This single-point sampling procedure at the middle point of the vertical profile was reported to cause less than 10% error compared with five-point procedure (Ni et al., 2008). Immediately after sampling, the pH value and salinity were measured with calibrated FiveGo portable pH meter and portable salinity meter, respectively. Each sample was filtered through a 142-mm diameter 0.4 µm nitrocellulose filter, and acidified with 20 mL concentrated HCl to pH < 2 within 2 h of sampling. All used filters were sealed in Ziploc bags and taken to lab, where the suspended sediments were freeze dried for 48-72 h, scraped from the filter and weighed to calculate the suspended load in water. For each filtered sample, an aliquot of 2 L was separated to an acid-cleaned 2 L bottle for 9 Be and heavy metals measurements. All water samples were not ultra-filtered, so our "dissolved" phase include both "truly dissolved" and colloidal beryllium.

Dissolved Beryllium Isotopes
For the determination of the dissolved 10 Be concentrations, the ∼18 L acidified water samples were added with 0.4 mg 9 Be carrier (AccuStandard R ICP-MS standard solution) and 120 mg Fe carrier, shook and left for equilibrium for 48 h. Then the pH value was adjusted to 8.5 with ammonium hydroxide, Be(OH) 2 was co-precipitated with Fe(OH) 3 and Mg(OH) 2 after equilibrium and setting for another 48 h. The precipitate was collected in a 50 mL centrifuge tube by decanting and centrifuging, and rinsed by ultrapure water (18.2 M MilliQ R water) twice. The cleaned precipitate was dissolved in 3 mL 10.2 M HCl, and the chromatography separation procedure of 10 Be was slightly modified from Simon et al. (2016b). Briefly, the Fe was separated by passing samples through a 12 mL column filled with 6 mL Dowex R 1 × 8 (100-200 mesh) anion exchange resin, the beryllium fraction was collected by 15 mL 10.2 M HCl and evaporated to dryness. After redissolved in 3 mL 10.2 M HCl, the anion resin step was repeated once, evaporated to dryness and redissolved in 1 mL 1 M HCl. Boron and aluminum were separated by loading redissolved sample solution onto a 12 mL column containing 5 mL Dowex R 50 × 8 (100-200 mesh) cation exchange resin. After eluent of 15 mL 1M HCl, the beryllium fraction was collected in the next 35 mL 1M HCl, evaporated to dryness and redissolved in 10 mL 1M HNO 3 solution. The Be(OH) 2 was precipitated at pH of 8.5, rinsed by pH 8 water for three times, redissolved in 0.1 mL concentrated HNO 3 and converted to BeO at 850 • C for 1 h. The BeO was mixed with Nb powder and pressed into cathode-holder for AMS measurements. Blank samples were made by co-precipitate 0.4 mg 9 Be and 120 mg Fe in 1 L ultrapure water and experienced the same chemical procedures.
The preconcentration and purification method to analyze dissolved 9 Be concentration was adjusted from Tazoe et al. (2014). Briefly, ∼250 g acidified water sample was neutralized to pH ∼7 and passed through a column contained 1 g Wako-Gel C-300 silica gel, the beryllium ( 9 Be) was retained on the column. The matrix residue such as Na, Ca, and Mg were separated by 15 mL ultrapure water and 15 mL 5 mM EDTA solution. The residual EDTA was removed by another 15 mL ultrapure water, and 9 Be was eluted by 15 mL 0.5 M HNO 3 and collected in a PFA vessel. The eluent was evaporated to dryness and redissolved in 1.2 mL 0.5 M HNO 3 . We compared our beryllium recovery with UV-treated samples but didn't find significant difference, thus the organic matter doesn't seem to affect the beryllium recovery of our water samples. The recovery rate of preconcentration method for dissolved 9 Be were tested by adding known amount of 9 Be to 50 mL pure water, resulting yields of 99.2 ± 2.3%.

Suspended Sediments
The "reactive" authigenic phases [amorphous and crystalline Fe-Mn (hydr)oxides] of sediment are main carriers of 10 Be and non-silicate-bound 9 Be, and the grain size effects are removed by normalizing 10 Be by 9 Be (Bourlès et al., 1989;Wittmann et al., 2012). To extract the reactive phase of suspended sediment, the simplified sequential leaching procedure was adopted following that of Wittmann et al. (2012) which was developed from protocols of Tessier et al. (1979) and Wiederhold et al. (2007).
Suspended sediment of 0.1-0.5 g was used for leaching, the volumes of following leaching solutions were based on 0.5 g sediment. The amorphous oxyhydroxides phase was leached by 15 mL of 0.5 M HCl for 24 h at room temperature (water bath of 25 • C and keep shaking at 200 rpm). The residue was separated by centrifuging, then rinsed by another 10 mL 0.5 M HCl, vortexed for 30 s and centrifuged. The 15 mL HCl leachate and the following 10 mL rinsing HCl were merged as leachate I, which contains the amorphous oxides fraction and the exchangeable fraction. The corresponding residue I was used for the extraction of crystalline oxide phase. Residue I was further leached by 10 mL 1M HCl in 1M hydroxylamine hydrochloride solution for 4 h at water bath of 80 • C (keep shaking at 250 rpm). After centrifuging, the residue was rinsed by 5 mL 0.5 HCl, vortexed for 30 s and centrifuged to separate residue II, and the 5 mL rinsing solution was incorporated with leachate II (the crystalline oxides fraction). Aliquots of 2 mL leachate I and II were separated to analyze 9 Be and heavy metal concentrations. Since our sequential extraction steps to achieve amorphous and crystalline phases are different from that of Poulton and Canfield (2005), the two phases we extracted are not completely corresponding to theirs.
To determine the reactive 10 Be concentrations, the leachate I and II were spiked with 0.4 mg 9 Be carrier, added with 5 mL concentrated HNO 3 , evaporated to dryness and redissolved in 1.5 mL 10.2 M HCl. The subsequent column chromatography procedures were the same as for water samples.

Measurements
The 10 Be measurements of all samples were performed at the French AMS national facility ASTER (CEREGE). The 10 Be standard used was CEREGE in-house standard (STD-11, 10 Be/ 9 Be = 1.191 ± 0.013 × 10 −11 ) which can be traced back to NIST 4325 Standard Reference Material (Braucher et al., 2015). Blanks of measured atomic ratio were at 10 −15 levels and deducted from measurement results. The 10 Be concentrations were calculated from measured 10 Be/ 9 Be ratio and mass of 9 Be carrier added (Simon et al., 2016b). Uncertainties of measured 10 Be/ 9 Be are around 3% and propagated to calculated 10 Be concentrations. The 9 Be concentrations were measured by a graphite furnace atomic absorption spectrometer (GF-AAS, Thermal Scientific ICE 3500) at Peking University. The detection limit of GF-AAS for 9 Be measurement is 0.0044 µg/L. The leachates were diluted for ∼50 times before measurements. For both water sample and leachate, standard addition method was used to eliminate the matrix effects. Each measurement was repeated four times and the standard deviation was below 3%. Both 9 Be and 10 Be concentrations in amorphous and crystalline fractions were measured separately.
The concentration of Cu, Cd, Pb in leachates were also analyzed by GF-AAS, and the concentration of Fe was measured by Flame-AAS. To determine the dissolved Cu, Cd, Pb concentration, ∼10 mL filtered water samples were added with 2 mL concentrated HNO 3 , evaporated to dryness and redissolved in 0.5M HNO 3 to fulfill 2-10 times enrichment. The recovery rate of Cu, Cd and Pb is 99 ± 1, 97 ± 2, and 96 ± 4%, respectively. The uncertainties (2σ) of heavy metal measurement are typically below 3%.

Calculation of Geo-Accumulation Index and Enrichment Factor
To evaluate contamination levels of heavy metal, the geoaccumulation index (I geo ) was first introduced by Müller (1969), and defined as the following equation: where C n is the measured concentrations of 9 Be and heavy metals in sediments, B n is the geochemical background values.
The background values of Cu, Cd, and Pb in the study area are selected as 17, 56, and 36 µg/g (MEP (Ministry of Environmental Protection of the People's Republic of China), 2009), while the background value of 9 Be is adopted as 1.19 µg/g (CEMS (China Environmental Monitoring Station), 1990). For each metal, the I geo is assessed for its contamination condition, from uncontaminated (I geo ≤ 0) to extremely contaminated (I geo ≥ 5) conditions (Hanif et al., 2016;Zhuang et al., 2018). To calculate the I geo−reac , the C n is adopted as the concentrations of metals in the reactive phases of the suspended sediments from the bottom samples. Since both the definition of I geo and the reported background values are based on the complete digestion of the sediments, the evaluation criterion of calculated I geo−reac with concentrations in reactive phase should be roughly 1 level higher than that of I geo . Thus, the uncontaminated criterion is I geo−reac ≤ −1, and I geo−reac ≥ 4 would represent extremely contaminated condition. The enrichment factor (EF) has been used to assess contamination level as well as to discriminate anthropogenic sources of heavy metals (Rahn and McCaffrey, 1979). Because the anthropogenic source of Fe is negligible compared to its high abundance in nature (Zhou et al., 2007), Fe is widely used as one of the reference elements in calculation of EFs, besides Al and Ti (Reimann and Caritat, 2000;Gu et al., 2018). In this study, we choose Fe as the reference element. The concentrations of heavy metals in sediments are normalized to the concentrations of the reference element (Fe), and compared to that of background values as equation (2): where Fe is selected as a normalizing element (C M /Fe) sample and (C M /Fe) background are the concentration ratio of heavy metals and Fe in suspended sediments from bottom water and background value, respectively. The background value of Fe from Guangdong Province soils is 28.8 mg/g (CEMS (China Environmental Monitoring Station), 1990). We assume that the elemental ratio in silicate residue is approximate to that of reac phase to assess the contamination level with our results of reactive phases.

Dissolved 9 Be and 10 Be Concentrations
The concentration of dissolved beryllium isotopes are listed along with salinity and pH values in Table 1. The units of dissolved concentrations of 9 Be ([ 9 Be] diss ) and 10 Be ([ 10 Be] diss ) are adopted as g/g water and at/g water , respectively. The [ 9 Be] diss range from 0.781 to 8.31 × 10 −12 g/g water (i.e., 87-921 pmol/kg). They are higher than the average value (0.66 × 10 −12 g/g water ) for the shallow groundwater of the PRE reported by Zhang et al. (2011), but their extremely high concentrations up to >40 × 10 −12 g/g water reported at several sites were not observed here. The Station HM in the north of the Lingdingyang Bay has the maximum [ 9 Be] diss of both surface and bottom water. The [ 9 Be] diss of the four outlets in the northern part of the Lingdingyang (i.e., HM, JM, HQ and HE) generally decreases  Kusakabe et al. (1991) in the salinity range of <18.0, and much higher than those in shelf seawaters of the SCS (S > 25). Compared with the riverine [ 9 Be] diss data worldwide, the [ 9 Be] diss of eight outlets are generally higher than those of Ganges-Brahmaputra estuarine water (Brown et al., 1992b), but are one magnitude lower than those in acidic Orinoco and Amazon basins (Measures and Edmond, 1983;Brown et al., 1992a). The [ 10 Be] diss range between 137 and 1,194 at/g water , and have a similar distribution pattern as the [ 9 Be] diss with higher concentrations in the bottom waters of Stations HM and L1. The [ 10 Be] diss in the Lingdingyang Bay are also comparable to those of Kusakabe et al. (1991), with [ 10 Be] diss of 550 at/g water in the Lingdingyang except for an abnormally high [ 10 Be] diss value of 22,060 at/g water . The bottom enrichment characters of [ 9 Be] diss and [ 10 Be] diss are similar to those previously observed by Yang et al. (2003) near the Yangzi River estuary. For all stations except JT in the Modaomen Bay, the [ 10 Be/ 9 Be] diss vary from 1.08 to 3.11 × 10 −9 , while Station JT has the highest [ 10 Be/ 9 Be] diss of 5.80 × 10 −9 due to its lowest [ 9 Be] diss value. The inconsistent beryllium isotope distributions of eight outlets reflect their distinct weathering environments of source areas, or impact of potential local anthropogenic contribution.

Reactive 9 Be and 10 Be Concentrations
The concentrations of 9 Be and 10 Be, as well as 10 Be/ 9 Be ratios in the reactive phases (sum of amorphous and crystalline oxides) are listed in Table 2. The subscripts of am-ox, x-ox, and reac refer to the fractions of amorphous oxides, crystalline oxides and reactive phase, respectively. The amount of suspended sediment collected from surface waters of Stations L3, L4, HE, and MD were not sufficient for leaching (<0.1 g), and the am-ox leachate of L2 for 10 Be measurements was unavailable due to an incidental sample loss during the preparation. The 9 Be concentrations in the reactive phases ([ 9 Be] reac ) of all samples range between 1.26 and 2.25 µg/g, higher than the background value of soils in Guangdong Province (CEMS (China Environmental Monitoring Station), 1990). The mean of [ 9 Be] am−ox (1.11 ± 0.14 µg/g) accounts for, on average, 62% of the [ 9 Be] reac (1.78 ± 0.25 µg/g). The 10 Be concentrations in the reactive phases ([ 10 Be] reac ) range from 1.37 to 2.97 × 10 8 at/g, with [ 10 Be] am−ox taking up 67% of the [ 10 Be] reac on average. The [ 10 Be] reac is slightly higher than the particulate 10 Be concentration of 0.25-2.48 × 10 8 at/g reported by Kusakabe et al. (1991) in the PRE (Locations 30 to 34). The ( 10 Be/ 9 Be) reac range between 1.25 and 2.28 × 10 −9 , and the average ( 10 Be/ 9 Be) am−ox (2.07 × 10 −9 ) is ∼21% higher than the ( 10 Be/ 9 Be) x−ox (1.71 × 10 −9 ). The exceptions are that ( 10 Be/ 9 Be) x−ox at JM and HQ are observed slightly higher than their ( 10 Be/ 9 Be) am−ox . Among the three sub-estuaries of the PR, the Lingdingyang Bay generally has higher [ 9 Be] reac and [ 10 Be] reac , but lower ( 10 Be/ 9 Be) reac ratios. The [ 9 Be] reac and [ 10 Be] reac of four outlet stations in the northern part of the Lingdingyang (i.e., HM, JM, HQ, and HE) show marked differences. The average [ 9 Be] reac of Stations HM and JM is ∼18% higher than the average value of all samples, while the corresponding [ 10 Be] reac only exceeds ∼4%, giving rise to distinctively low ( 10 Be/ 9 Be) reac ratios. Station HQ has the highest [ 10 Be] reac among all outlets, as well as a ( 10 Be/ 9 Be) reac ratio much higher than those of HM and JM. As an exception in the Lingdingyang, the [ 9 Be] reac , [ 10 Be] reac , and ( 10 Be/ 9 Be) reac of Station HE are all lower than the average values. From Stations L1-L4 (with increasing salinity), the [ 9 Be] reac and [ 10 Be] reac in the bottom suspended sediments decreased by 15 and 10%, respectively.
Comparing to the northern outlets in the Lingdingyang Bay, the southern four outlets in the Modaomen and Huangmaohai Bay have generally lower [ 9 Be] reac and [ 10 Be] reac , resulting in higher ( 10 Be/ 9 Be) reac ratios. There is one exception that Station HT in the Huangmaohai Bay has distinctively lower ( 10 Be/ 9 Be) reac ratios both in surface and bottom suspended sediments. Just a few kilometers to the west of Station HT, Station YM has an average ( 10 Be/ 9 Be) reac ratio of 2.23 × 10 −9 , ∼32% higher than that of HT. It should be pointed out that YM is also an outlet of the Tanjiang River, rising from west Guangdong Province and also belonging to the PR system.

Concentrations of Cu, Cd, and Pb in Dissolved and Reactive Phases
The concentrations of Cu, Cd, and Pb in both dissolved and reactive phases are summarized in  Apart from the heavy metals, the concentrations of reactive Fe are also analyzed to derive the metal/Fe ratio. Because the anthropogenic source of Fe is negligible compared to its high abundance in nature (Zhou et al., 2007), Fe is widely used as a reference element in calculation of EFs (Reimann and Caritat, 2000). The [Fe] reac is nearly constant among all samples, with values of most samples around 46 mg/g. Hence, it is appropriate to use Fe as a normalizing element in our study. Wittmann et al. (2012) reported that the amorphous oxides of suspended sediments carried 47-74% of 10 Be and 15-27% of 9 Be relative to the total detrital Amazon River sediments. Compared to the crystalline oxides which record the 10 Be/ 9 Be from upper reaches of the river, the amorphous oxides phases have higher specific surface areas and exchange with dissolved phase more actively (Wittmann et al., 2015). Here we calculate distribution coefficient (K d ), as well as partition ratio (k) between amorphous oxides of suspended sediments and dissolved phases, to present proportions of beryllium isotopes carried by suspended sediments. In a certain volume of water, the partition ratio represents the ratio of total amounts of 10 Be (or 9 Be) carried by suspended sediment and water: k( 10 Be) am−ox = m( 10 Be) am−ox m( 10 Be) diss = K d ( 10 Be) am−ox = m sediment m water (3) where m( 10 Be) am−ox and m( 10 Be) diss represent total amounts of 10 Be in amorphous oxides and dissolved phases, C( 10 Be) am−ox and C( 10 Be) diss are their concentrations in corresponding phases, and m sediment /m water is the suspended load calculated from mass of dried filter and sampled water. The K d ( 10 Be) am−ox is calculated as a ratio of 10 Be concentrations in sediments (amorphous oxide phase) and water (You et al., 1989;Aldahan et al., 1999). Thus, both the k of 10 Be and 9 Be are calculated. A higher k( 10 Be) am−ox means that the suspended sediments carried more 10 Be relative to dissolved phase. Since the concentrations of 10 Be and 9 Be in amorphous oxides have minor variations, the k( 10 Be) am−ox and k( 9 Be) am−ox mainly depend on their dissolved concentrations and the suspended load.

Partition of Beryllium Isotopes Between Amorphous Oxides and Dissolved Phases
The distribution coefficients K d ( 9 Be) am−ox and K d ( 10 Be) am−ox are site dependent, ranging from (1.56-10.8) × 10 5 and (1.32-9.76) × 10 5 g water /g, respectively. Both of them are within the limits of K d values at neutral and alkaline conditions observed by You et al. (1989) in the laboratory batch experiments. The bottom water of Stations L1, L3, and HM with high suspended load is corresponding to low K d values (Figure 2). Most of the calculated partition ratios k( 10 Be) am−ox and k( 9 Be) am−ox are in the range from 3.00-33.5 to 2.98-30.4, with average values of 21.1 and 20.1, respectively. Therefore, only 5% of 10 Be occurs in the water as dissolved form, much less than those in form of amorphous oxide. The extremely high FIGURE 2 | Partition coefficient (K d ) of 9 Be and 10 Be versus suspended load. Three bottom waters (HM, L1, and L3) with high suspended load and low K d values are marked. k( 10 Be) am−ox and k( 9 Be) am−ox , (71.7 and 116) for the bottom water of Station L3 are attributed to the sampling depth that was too close to the river bed and caused a higher suspended load. As shown in Figure 3A, the k( 10 Be) am−ox -k( 9 Be) am−ox relationship is close to the 1:1 line on log-log coordinates. Excluding the extremely high value of L3, the k( 10 Be) am−ox is, on average, 20% higher than the corresponding k( 9 Be) am−ox .
Similar to K d , the k( 10 Be) am−ox and k( 9 Be) am−ox generally increase with pH at pH > 7 (Figure 3B), with exception of HM. The lowest k( 10 Be) am−ox and k( 9 Be) am−ox were found at Stations JM and HM, where the river waters are too clear to obtain enough suspended load and precise beryllium concentrations in reactive phase. From Station L1 to Station L4, the k( 9 Be) am−ox show minor variations and k( 10 Be) am−ox increases ∼50%, despite the greater increase in pH and salinity. Indeed, the K d of 10 Be and 9 Be in bottom water have increased by three times from L1 to L4. The reason for decreased amplitude of k value is that the concentration of suspended sediments of L4 is dropped to FIGURE 3 | (A) The partition ratio (k) of 9 Be and 10 Be between amorphous oxides and dissolved phases. The blue dash line represents 1:1 line. (B) The k-pH plot of both 9 Be (black squares) and 10 Be (blue dots), the data of HM with pH < 7 are not shown. The extremely high values of L3 were excluded in these plots due to suspended load bias. ∼1/4 compared with L1. Thus, the k instead of K d could reflect the capacity of suspended sediments in transporting beryllium isotopes more directly.
Since most of the suspended sediments are trapped in the estuary, the k( 10 Be) am−ox from river outlets are higher than the actual percentage of particulate 10 Be that transport from river to seawater. Moreover, the exchange between suspended sediments and dissolved phases cannot reach equilibrium in several weeks (Li et al., 1984), which may introduce large error to the 10 Be budget estimation based on K d or k.

Export Flux of Riverine 10 Be Into Estuarine Water
At steady state, the 10 Be flux transported by river water into estuary and the SCS should be equal to the deposition flux of meteoric 10 Be that the river basin received. We apply the approach of Wittmann et al. (2015) which was based on framework of von  to estimate the balance of 10 Be flux between atmospheric deposition and riverine export. Based on the assumption of constant concentrations, discharges and sediment loads, a very rough estimation of 10 Be export flux can be achieved. The equations to calculate total fluxes of atmospheric input (J 10 Be atm ) and riverine exported (J 10 Be riv ) of 10 Be are listed as follows: where A riv is the area of the PR basin, about 4.50 × 10 11 m 2 ; F 10 Be met is the meteoric flux of 10 Be, which is derived from ECHAM5-HAM general circulation model as 250 at/m 2 /s and equals to 7.89 × 10 9 at/m 2 /year (Heikkilä et al., 2013); λ is the decay constant of 10 Be, while t is the storage time of sediments. If the storage time is short enough, the right term of equation (5)   in crystalline oxides are also included. Thus, the equation (5) is transformed as sum of eight outlets to equation (6) (4) and (6) are 3.55 × 10 21 and 2.74 × 10 21 at/year, respectively. The exported depositional flux ratio of 10 Be (J 10 Be riv /J 10 Be atm ) is 0.77, that is, 23% of the meteoric 10 Be that is deposited on the PR basin is retained in the basin. Wittmann et al. (2015) pointed out that 10 Be deposited on inactive areas and the decay during sediment storage may also explain the 23% gap between the atmospheric input and export.
As the major Bay of the PR, the Lingdingyang receives 53% discharge of the PR from HM, JM, HQ, and HE. The riverine 10 Be export from these four outlets is 2.05 × 10 21 at/year, accounts for 75% of the total riverine export. The data of Station L4  in the southern end of Lingdingyang are adopted to estimate total amounts of 10 Be that escape the Lingdingyang Bay. The discharge of Lingdingyang is the sum from the northern four outlets. Thus, the estimated export 10 Be flux of Lingdingyang is 0.99 × 10 21 at/year. The ratio of 10 Be flux input into the Lingdingyang Bay and export into the SCS is 49%, i.e., about half of the riverine 10 Be is trapped in the Lingdingyang Bay. Assuming that the Modaomen and Huangmaohai Bay have similar trapped 10 Be proportions as the Lingdingyang, and multiplied by the J 10 Be riv /J 10 Be atm ratio of 0.77, the fraction of 10 Be that export into the SCS from the PRE compared to the corresponding deposition of meteoric 10 Be was 38%, although most riverine 10 Be may be further retained in shelf sediments. Therefore, our results are consistent with the current knowledge that marginal seas and continental margins are major sinks of marine 10 Be (Anderson et al., 1990;Yang et al., 2003).

Contamination Assessment of Beryllium and Heavy Metals
The I geo−reac values are calculated to evaluate the contamination levels of Be ( 9 Be), Cu, Cd, Pb in the PRE. Results of I geo−reac for all metals in suspended sediments are shown in Figure 4. The order of average I geo−reac of the four metals is Cd (2.45 ± 1.46) ≥ Cu (2.39 ± 1.21) > Pb (0.471 ± 0.469) > Be (−0.048 ± 0.209). Most stations of the PRE are highly polluted by Cu and Cd, while 9 Be and Pb belong to uncontaminated to moderately contaminated levels.
The highest I geo−reac of Cu is found at Station JT (5.16, extremely contaminated), while HM, JM, and HE also have I geo−reac higher than 3 (heavily to extremely contaminated). Heavily contaminated levels (2 < I geo−reac ≤ 3) are observed at middle part of the Lingdingyang (L2 and L3). The lowest I geo−reac value of Cu was found at MD (0.876, moderately contaminated), which sits in the same sub-estuary with JT. This distinct contrast of I geo−reac values from two adjacent outlets indicate that the contamination of Cu is more likely originated from local anthropogenic sources than from upper reaches. The high contamination levels of Cu at HM and JT was also observed by Liu et al. (2014), who attributed these to industrial discharges, copper mining factories and locally high copper consumption.
The highest contamination level for Cd are observed at JM (4.68) and MD (4.21). Indeed, the I geo−reac of Cd for all eight outlets are above 2, indicating a widespread contamination of Cd among these outlets. The contamination of Cd is mainly sourced from industrial sewage (electroplating and electronic industries) and runoff from agriculture sites in South China (Cheung et al., 2003;Zhuang et al., 2016). The I geo−reac values of Cd for most estuarine stations in the Lingdingyang are below 1, thus the majority pollution of Cd is not delivered to estuarine waters.
Compared to Cu and Cd, Pb is less polluted in the PRE, with highest I geo−reac found at JM (1.52) and JT (1.39). The I geo−reac values of Pb for the rest stations are all below 1, belonging to moderately contaminated conditions. Unlike heavy metals, all I geo−reac values of 9 Be are around zero, and no high outliers are found. Thus, the PRE area is generally uncontaminated to moderately contaminated by anthropogenic sources of 9 Be.
In order to differentiate between natural and potential anthropogenic sources of 9 Be and heavy metals, the EF values are calculated and shown in Figure 5. We classify the EF values into the following three levels based on criteria adopted by Gu et al. (2012Gu et al. ( , 2018: mainly natural source (EF < 0.5); natural or anthropogenic sources (0.5 ≤ EF ≤ 1.5); mainly anthropogenic source (EF > 1.5). For the four investigated metals, the ranking of EF values in the PRE is Cd (7.77 ± 6.19) ≥ Cu (7.16 ± 7.57) > Pb (1.34 ± 0.46) > Be (0.896 ± 0.124). Accordingly, Cd and Cu mainly come from anthropogenic pollutions, while Pb and 9 Be are at the natural or anthropogenic sources level. The spatial distribution of EF of 9 Be is similar to that of I geo−reac , and there are no outliers with distinct high EF of 9 Be among all stations.
In summary, the 9 Be in the PRE are mainly originated from natural sources, none of these investigated stations show evident anthropogenic contaminations. The case of Pb is most similar to 9 Be, with anthropogenic sources observed at only two stations. Unlike 9 Be and Pb, the Cd and Cu are widely enriched in suspended sediments, and have extremely high level of contamination at several stations. These results confirm that anthropogenic source of 9 Be are negligible in the PRE area, hence the beryllium isotopes are eligible in various studies of Earth surface processes.

CONCLUSION
In this study, distribution of beryllium isotopes in both dissolved and reactive phases of suspended sediments from the PRE are presented. The [ 9 Be] diss and [ 10 Be] diss range between 0.78-8.31 × 10 −12 g/g water and 137-1,194 at/g water , respectively. We calculated the distribution coefficients (K d ) and partition ratio (k) of 10 Be and 9 Be between amorphous oxides phase of suspended sediments and the dissolved form. The K d of both isotopes are at 10 5 level, and low K d values are found in three bottom waters with high suspended load. The average k values of 10 Be and 9 Be are 21.1 and 20.1, respectively. This means that only ∼5% of 10 Be exist in the water as dissolved form, in contrast to the vast inventory of amorphous oxides. In spite of the intense removal of dissolved 10 Be and 9 Be in the estuary, their proportions in the water and suspended sediments only show minor fluctuations.
The total fluxes of atmospheric meteoric 10 Be deposited on the PR basin, and the 10 Be fluxes export through eight outlets were estimated as 3.55 × 10 21 and 2.74 × 10 21 at/year. Thus, 23% of deposited atmospheric 10 Be are retained in the river basin, and only 38% of deposited atmospheric 10 Be could escape the PRE and be transported to the SCS. The calculated I geo−reac and EF suggest that the PRE area is highly polluted by Cd and Cu, less polluted by Pb, while Be is hardly sourced from human activities. Therefore, our results demonstrate that the 10 Be/ 9 Be ratios in the PRE and the northern SCS are mainly controlled by natural processes, and are thus suitable for geoscience studies such as denudation rates by meteoric 10 Be and tracing water masses by ( 10 Be/ 9 Be) seawater .

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.

AUTHOR CONTRIBUTIONS
WK and LZ designed the research and undertaken data analysis and manuscript writing. WK performed the sampling and experiments. GA, DB, and KK responsible for the 10 Be measurements. All authors contributed to the article and approved the submitted version.