Distribution and fractionation of rare earth elements in suspended particulate matter in a coastal river, Southeast China

Background In the river system, the geochemistry of rare earth elements (REEs, a series of elements from La to Lu) in suspended particulate matter (SPM) is generally controlled by rock weathering processes and hydrochemical characteristics, as well as being affected by anthropogenic activities. However, the variations of geochemical characteristics and behaviors of REEs in SPM with a salinity gradient from the inland river to the estuary have been short of a systematic understanding. Methods The REE concentrations, Post Archean Australia Shale (PAAS)-normalized REE, La/Yb, La/Sm, and Sm/Yb ratios of SPM were investigated in the Jiulongjiang River, which is a coastal river mainly flowing through granite rocks in Southeast China. The correlation relationships between physicochemical parameters (including water pH, total dissolved solids (TDS), HCO3– concentrations, and the concentrations of major elements of SPM) and PAAS-normalized REE ratios of SPM were analyzed to determine the factors that affect the REE concentration and fractionation of SPM in the different regions of Jiulongjiang River, including the main stream and tributary of Beixi River, Xixi River, Nanxi River, and estuary. Additionally, the Ce, Eu, and Gd anomalies of SPM were estimated. Results The average ∑REE concentration of SPM (352 mg/kg) in the granite rock basin was twice higher than the mean value (175 mg/kg) of the world’s rivers. The PAAS-normalized REE ratios of SPM in the main rivers including Beixi River (main stream), Xixi River, and Nanxi River were near due to the same lithologic distribution. In the tributary of Beixi River, the input of low-weathered carbonate minerals which contain very few REE caused the lower REE concentrations of SPM. The PAAS-normalized REE ratios of SPM in the estuary were significantly lower than those in the main rivers, which was mainly attributed to the significant REE removal with the increment of salinity. The enrichment of LREE relative to HREE in SPM increased with decreasing water pH in the main rivers. In the estuary, the preferential removal of dissolved LREE occurred compared to HREE with the increment of salinity. The negative Ce and Eu anomalies of SPM occurred in both the main rivers and estuary region and rare Gd pollution was present in the basin. Additionally, human activities caused the increment of REE concentrations and more negative Ce anomaly at some specific sites, such as dam effect and agricultural pollution. Conclusions The REE concentrations and fractionations of SPM in river water mainly depend on lithologic distribution and riverine pH, while they are affected by salinity in the estuary.

In the river system, REEs in rocks mobilize into riverine suspended particulate matter (SPM) and dissolved loads during the rock weathering processes. The two forms of REEs can be interconverted through the physical and chemical processes, such as adsorption, colloid flocculation, dissolution, and complexation (Brookins, 1989;Sholkovitz, Landing & Lewis, 1994;Xu & Han, 2009). Generally, the REEs of SPM (59-289 mg/kg in the world's rivers) are far more than those in dissolved loads (13-2,484 µg/kg) (Goldstein & Jacobsen, 1988), for example, da Silva et al. (2018) reported that more than 95% of total riverine REEs were carried by SPM in the Ipojuca River. Rock weathering products and re-suspended sediments are the primary sources of SPM (Roussiez, Aubert & Heussner, 2013;Vercruysse, Grabowski & Rickson, 2017). In some sites with intensive human activities, a large part of REEs in SPM is likely attributed to anthropogenic sources (Bau & Dulski, 1996;Louis et al., 2020). Thus, the REEs of SPM in the river, even other water bodies, record the integrated information about the geochemical characteristics of REE in SPM controlled by weathering processes and human disturbances since the Anthropocene (Linders et al., 2018;Rogers et al., 2019).
The alterations of hydrochemical characteristics caused by the input of weathering products and anthropogenic activities can substantially affect the concentration and fractionation of REE in SPM (Jones et al., 2016;Smith & Liu, 2018). For example, water pH, redox potential, and organic matter concentration significantly influence the dissolution and enrichment of SPM-associated REE in an inland river (Goldstein & Jacobsen, 1988;Johannesson et al., 2004;Migaszewski, Galuszka & Dolegowska, 2019). However, the REE removal in the estuary region with the increment of salinity significantly reduces the REE concentration of SPM (Elderfield, Upstill-Goddard & Sholkovitz, 1990;Nozaki et al., 2000). On the other hand, the REE fractionation can cause the variational concentrations of different REEs (Louvat & Allegre, 1998), due to the alterations of water physicochemical properties (Elias et al., 2019;Michaelides et al., 2010). Generally, the fractionations of REEs in SPM are also closely related to the water physicochemical properties, such as water pH, salinity, and the colloids of Fe/Mn oxides/hydroxides (Dagg et al., 2004;Krickov et al., 2020;Quinn, Byrne & Schijf, 2006). For example, LREE is preferentially adsorbed by clay mineral particles compared to HREE, resulting in HREE depletion relative to LREE in SPM (Chelnokov, Bragin & Kharitonova, 2020). Xu & Han (2009) reported that MREE and Ce are preferentially adsorbed by Fe/Mn oxides/hydroxides, while other LREEs are easily absorbed by clay colloids. Compared to LREE, HREE is more easily separated with SPM into the dissolved load with increasing river water acidity (Migaszewski, Galuszka & Dolegowska, 2019). Elderfield, Upstill-Goddard & Sholkovitz (1990) reported that the preferential removal of LREE occurred with the increment of salinity in the estuary region. Additionally, the redox potential-dependent Ce anomaly (Alderton, Pearce & Potts, 1980), lithology-dependent Eu anomaly (Han et al., 2009), and anthropogenic pollutiondependent Gd anomaly (Bau & Dulski, 1996;Louis et al., 2020) are the useful indexes of REE fractionation. Overall, the geochemical characteristics (including concentration and fractionation) of REEs in SPM can provide much useful information about weathering processes, hydrochemical characteristics, and anthropogenic activities (Cholet et al., 2019;Smith & Liu, 2018).
REEs of SPM are useful environmental tracers to indicate weathering processes and anthropogenic activities. The Jiulongjiang River, which is a coastal river mainly flowing through granite rocks in Southeast China (Liu & Han, 2020), is a perfect area to study the variations of geochemical characteristics and behaviors of REEs in SPM with a salinity gradient from the inland river to the estuary. Thus, the concentration and fractionation of REEs in SPM in river and estuary were investigated and the effects of weathering processes and anthropogenic activities on them were analyzed in the present study. This study aims to: (1) investigate the concentrations and fractionations of REEs in SPM in river and estuary, (2) determine their controlling factors from weathering processes and anthropogenic activities, and (3) understand the differences in the concentrations and fractionations of REEs in SPM between in river and estuary.

Study area
The Jiulongjiang River (24 • 05 N-25 • 55 N, 106 • 50 E-118 • 20 E), which is situated at the southeast coastal margin of the Chinese mainland, has the basin area of about 14,740 km 2 (Liu & Han, 2020). The main stream flows across the Longyan and Zhangzhou cities with a channel length of 260 km and an annual water discharge of 14 km 3 . The Jiulongjiang River is comprised of the three main rivers, including Beixi River, Nanxi River, and Xixi River. The basin is mainly controlled by the sub-tropic monsoon climate. The period with the majority of rainfalls (∼80%) and the highest air temperature are simultaneously concentrated in the wet season from April to September. In the basin, mean annual precipitation (MAP) ranges from 1400 mm to 1800 mm and the mean annual temperature (MAT) varies from 19.9 • C to 21.1 • C (Liu & Han, 2020). Magmatic rocks and clastic sedimentary rocks are widely distributed in the basin, while metamorphic rocks and limestones are exposed rarely (Li, Qiu & Yang, 2014). The red soils, which are classified into Alisols (WRB, 2014), account for >90% area of the basin. The region of upper reaches mainly flows through mountain areas with high forest coverage and slight agricultural and industrial perturbations, while other regions are affected by different degrees of human activities (Huang et al., 2013). Furthermore, the Wananxi Reservoir regulates the discharge of river water in the upper reaches of the Beixi River.

Sample collection and in-field measurements
A total of 42 river water samples were collected in the high-flow season (July) of 2014 ( Fig. 1 and Table 1). The river water samples of 1-21, 24-33, and 34-37 were collected along with the Beixi River, Xixi River, and Nanxi River, respectively. The sites of 22, 23, and 38-42 were located in the estuary region. The physicochemical parameters of pH values and total dissolved solids (TDS) concentrations in river and estuary water were measured using a water quality monitor (YSI multi-parameter probe) on site. The HCO 3 − concentrations were determined using titration with diluent HCl solution (Liu & Han, 2021). All river water samples were hoisted by a clean plastic bucket from the center of the river on bridges and ferries. The samples were collected at a depth about 0.5 m below the water surface. Water samples were stored in a 50L LDPE bag, which had been pre-cleaned with 10% nitric acid. The collected river waters were filtered through a 0.22 µm cellulose acetate membrane (Millipore) to separate SPM samples within 24 h. Compared to the 0.45 µm size, the 0.22 µm size was more widely employed to distinguish SPM and dissolved solutes in natural river water (Zeng & Han, 2020;Zeng, Han & Yang, 2020). An important reason is that the 0.22 µm filter can separate most particles or colloids associated with anthropogenic sources, such as living substances and biopolymer aggregates (Jackson & Burd, 2015). The SPM samples absorbed on the surface of the filter membrane were collected through washing by deionized water and then dried on a hotplate at 55 • C.

Analytical methods
The SPM samples were digested to dissolved loads according to the modified method by Li & Han (2021). The procedure was showed in detail, dried SPM sample was weighed 100 mg, put into a cleaned PFA digestion tank digested, added one mL pure HF and three mL pure HNO 3 , heated keeping at 140 • C for 3 days (Li et al., 2020). Finally, the solid sample was transformed into a solution, and then stored in a 2% HNO 3 solution for analyzing the concentrations of REEs and major elements. For the entire procedure, the blank samples and standard samples (GBW07404 and GBW07120) were treated the same as the samples, to monitor procedural reliability. The recovery rate of Ce was more than 90% and over 95% for other REEs during the digestion procedure. The concentrations of REEs (including La, Ce, Pr, Nd, Sm, Eu, Gd, Tb, Dy, Ho, Er, Tm, Yb, and Lu) in SPM  were determined by an inductively coupled plasma mass spectrometry (ICP-MS, ELAN DRC-e, Perkin Elmer, Waltham, Massachusetts, USA), and the concentrations of major elements (including Al, Ca, Fe, K, Mg, Mn, Na, and Ti) were analyzed by an inductively coupled plasma optical emission spectroscopy (ICP-OES, Optima 5300DV, PerkinElmer, Waltham, Massachusetts, USA) (Liu, Han & Li, 2021). To restrict the mass interferences of BaO + on Eu during the instrument testing process, a liquid-liquid extraction method was employed to remove Ba efficiently, as well as other matrix elements (Shabani & Masuda, 1991). The precision and accuracy of measurement were evaluated by repeated analysis of standard solution with multi-element, which were better than ±3% for major elements and better than ±5% for REEs, respectively.

Chemical index of alteration
The chemical index of alteration (CIA) of SPM can indicate the degree of weathering (i.e., the evolution degree from luminum-silicate minerals (especially feldspar) to clay minerals) in the source region (Nesbitt & Young, 1982). The higher CIA value means the more intensive leaching of Na, K, and Ca in silicate minerals, that is the stronger chemical weathering. The CIA > 80% indicates strong chemical weathering; 60% < CIA < 80% indicates moderate chemical weathering; CIA < 60% indicates weak chemical weathering. The CIA can be obtained by the calculation of molecular proportions, as the formula: where CaO* (mol/kg) is the molecular mass concentration of CaO incorporated in the silicate fraction of the rock. A correction is made for carbonate and apatite content (McLennan, 1993). In brief, if CaO/Na 2 O (mole ratio) > 1, the mole fraction of CaO* is replaced by the mole fraction of Na 2 O; if CaO/Na 2 O < 1, CIA is calculated directly using the mole fraction of CaO.

Statistical analysis
Boxplot was used to show the ranges of the concentrations of REE, LREE, MREE, and HREE of SPM in different regions of the Jiulongjiang River basin. The normal distribution of the sample data set were tested via the Shapiro-Wilk test before Pearson correlation analysis. Pearson correlation coefficient determined the relationship between physicochemical parameters and REE fractionation proxies of SPM. All statistical analyses were performed by the SPSS 18.0 software (SPSS Inc., Chicago, IL, USA) and all graphs were drawn by SigmaPlot 12.5 software (Systat Software GmbH, Erkrath, Germany).

Hydrochemical characteristics
The pH, TDS, and HCO 3 − concentration of river water at different sampling sites in the Jiulongjiang River are showed in Table 1. The pH values ranged from 6.7 to 7.9 (mean 7.2 ± 0.3), indicating the neutral and slightly alkaline river water. The TDS varied intensively with a range of 27.1-720 mg/L (mean 111 ± 107 mg/L). Moreover, the higher TDS mainly occurred in the water of the estuary region. The HCO 3 − concentrations ranged from 7.5 mg/L to 49.0 mg/L (mean 22.1 ± 9.1 mg/L) and showed an increasing trend downstream. Particularly, the water of tributary in the Beixi River had higher HCO 3 − concentrations compared to the main stream.

The concentrations of major elements in SPM and CIA values
The concentrations of the major elements in SPM are shown in Table 2. The average Al, Ca, Fe, K, Mg, Mn, Na, and Ti concentrations in SPM were 151 ± 22.6 g/kg, 4.8 ± 7.8 g/kg, 49.6 ± 20.2 g/kg, 17.3 ± 3.7 g/kg, 4.3 ± 2.3 g/kg, 1.2 ± 0.5 g/kg, 2.1 ± 1.9 g/kg, and 2.9 ± 0.6 g/kg, respectively. Generally, the concentrations of the major elements slightly varied among the sites in the river, except for several sampling sites. For example, the No. 21 and 34 sites had relatively high Ca, Mg, and Na concentrations. However, Al and Fe concentrations at the 34 site were significantly lower than those at other sites, but not at the 21 site. Furthermore, most metal elements, besides Mn, were at a relatively high concentration at the 37 site. The CIA values varied among the sites in the river with a range of 83.2%-95.7% (89.6 ± 4.8%), except for the exceptionally low value of 63.7% at the 34 site (Table 3). Particularly, the CIA values in the tributary of the Beixi River were slightly lower than those in the main stream.

REE concentrations of SPM
The concentrations of single REE in SPM at different sampling sites are shown in Table 4. The total REE ( REE) concentrations of SPM ranged from 111 mg/kg to 1,292 mg/kg with an average of 376 ± 191 mg/kg (Fig. 2), which were higher than those of the UCC (146 mg/kg) (Taylor & McLennan, 1985), the local soils of Fujian Province (198 mg kg −1 ) (Chen et al., 1992), the SPM of world rivers (175 mg/kg) (Viers, Dupre & Gaillardet, 2009), and the PAAS (185 mg/kg) (Taylor & McLennan, 1985)). The LREEs accounted for the largest proportion of total REEs in SPM at all sites. The LREE concentrations of SPM ranged from 95.5 mg/kg to 1093 mg/kg with an average of 324 ± 164 mg/kg. The MREE and HREE concentrations of SPM varied from 12.5 mg/kg to 162 mg/kg (mean 42.2 ± 22.9 mg/kg) and from 2.6 mg/kg to 36.9 mg/kg (mean 10.2 ± 5.1 mg/kg), respectively. For spatial distribution, the REE LREE, MREE and HREE concentrations of SPM in the main stream of Beixi River were similar to those in the Xixi River and Nanxi River (Fig. 2). However, these concentrations in the tributary of Beixi River were significantly lower than those in the main stream, but near to those in the estuary region.

PAAS-normalized REE patterns of SPM
The PAAS-normalized REE patterns of SPM at each site were showed in Table 5. The PAAS-normalized REE ratios of SPM at most sites were larger than 1, except for the parts of REEs at the 11, 34, and 40 sites. The pattern of average SPM in the main stream of Beixi River was almost identical to that in the Xixi River and Nanxi River (Fig. 3). Additionally, the pattern in the tributary of Beixi River was similar to that in the estuary region. Overall, the enrichment of REEs in SPM was larger in the main stream than that in tributary and estuary. The (La/Yb) N ratio, (La/Sm) N ratio, and (Sm/Yb) N ratio of SPM at different sites are shown in Fig. 4. The (La/Yb) N and (Sm/Yb) N ratios at all sites were higher than 1, indicating the enrichment of LREE and MREE relative to HREE in SPM. The (La/Yb) N and (Sm/Yb) N ratios generally showed a decreasing trend downstream in the Beixi River. The (La/Sm) N ratio of SPM varied within a wide range of 0.7-1.3, indicating the strong spatial variability between LREE and MREE fractionation in the river.

Ce, Eu, and Gd anomalies in SPM
The Ce, Eu, and Gd anomalies in SPM at different sites are shown in Fig. 5. The Ce of SPM at most sites showed a slight negative anomaly (δCe: 0.8-1.0), except for the relatively large negative anomaly at the 1 and 28 sites with a δCe value of 0.70 and 0.43, respectively. The Eu of SPM at most sites showed negative anomaly (δEu: 0.7-1.0). Exceptionally, a positive Eu anomaly occurred at the 21 and 34 sites with a δEu value of 1.15 and 1.24, respectively. Additionally, the δEu value generally showed an increasing trend along with the flow direction of the main stream. The δGd values of SPM at most sites ranged from 1.5 to 1.6, except for a relatively higher δGd value at the 37 site (1.72) and a relatively lower value at the 28 site (1.40). This result indicates that the SPM in the Jiulongjiang River is almost not affected by anthropogenic Gd sources.

Controls of REE concentrations in SPM of river water by weathering processes
The clay minerals of SPM mainly derived from the weathering of rocks are important carriers of REE (Migaszewski & Galuszka, 2015). Thus, the REE concentrations of SPM are closely associated with the lithologic distribution in a basin. In the main stream of Beixi River, the PAAS-normalized REE ratios of SPM positively correlated with Al concentrations and CIA values, and negatively correlated with TDS, Fe, K, and Mg concentrations (Table 6). These results indicated that the REE concentrations of SPM were dominated by silicate weathering in the main stream, similar to the report from the downstream of Zhujiang River (Xu & Han, 2009). In the basin, the weathering of magmatic rocks and clastic sedimentary rocks contributes to abundant Al-silicate minerals (clay minerals) into soils. These clay minerals, which carry most REEs, are transferred into the river by soil erosion, resulting in the increment of REE concentration in SPM (Linders et al., 2018). Thus, the PAAS-normalized REE ratios positively correlated with Al concentrations in SPM.
In the tributary of Beixi River, the PAAS-normalized REE ratios of SPM positively correlated with Ca concentrations and negatively correlated with CIA values. These results indicated that the REE concentrations of SPM in the tributary were also affected   (Figs. 2 and 3). A similar result about the low REE concentrations of SPM in the karst basin has been reported by Han et al. (2009).
The CIA values of SPM in the tributary of the Beixi River were slightly lower than those in the main stream (Table 1). The result indicated that the SPM in the tributary consisted of many low-weathered carbonate minerals. However, the carbonate rock contains very few REEs, with a mean REE concentration of 8.05 mg kg −1 (Han, Yang & Zeng, 2021). Thus, the river flowing through the carbonate rock region is characteristic of the low REE and high Ca concentrations in SPM (Xu & Han, 2009).
The REE, LREE, MREE, and HREE concentrations, PAAS-normalized REE ratios of SPM in the Xixi River and Nanxi River were near to those in the main stream of Beixi River (Figs. 2 and 3). Moreover, the correlation relationships (although not significantly) of the PAAS-normalized REE ratios with major element concentrations and CIA values were also similar in the three rivers (Table 6). These results indicated that the REE concentrations in the Xixi River and Nanxi River were also controlled by the weathering of magmatic rocks and clastic sedimentary rocks. However, the non-significant correlation relationships were likely attributed to the influences from human activities. Compared to the wide forest upstream of the Beixi River, the Xixi River and Nanxi River mainly pass through farmlands and residential areas (Liu & Han, 2020). Human activities can directly cause the increment of REE into the river and disturb REE fractionation by affecting physicochemical parameters of river water (Da Silva et al., 2018;Yang et al., 2021).

Influences of salinity on REE concentrations in SPM of estuary water
The PAAS-normalized REE ratios of SPM in the estuary region were near to those in the tributary of Beixi River (Fig. 3). The result is not attributed to the weathering of carbonate minerals like that in the tributary of the Beixi River. Theoretically, the characteristics of REE concentrations in the estuary region should inherit from the main stream of Beixi river, Xixi River, and Nanxi River because the three main rivers have the larger water discharge      (Table 6), indicating the effects of silicate weathering on REE concentrations of SPM similar to the effects in the main stream of Beixi River. Compared to river water, the estuary water was lower in REE concentrations of SPM (Fig. 3), which was mainly attributed to REE removal. Elderfield, Upstill-Goddard & Sholkovitz (1990) found that significant REE removal (∼30%) occurred in the Connecticut, Delaware, Mullica, and Tamar estuaries. During the mixing of river water and seawater, the increment of salinity promotes the coagulation of colloidally associated Fe and Mn hydroxides (Migaszewski & Galuszka, 2015). The flocculations easily absorb REEs and transfer them into the sediments, resulting in significant removals of both dissolved and particulate REE. Thus, the coprecipitation of Fe and Mn hydroxides and REEs in the estuary explained the positive correlation (although non-significant) between the PAAS-normalized REE ratios and Fe (and Mn) concentrations of SPM (Table 6).

Effects of REE of human activities on REE concentrations in SPM
The exceptional REE concentrations of SPM at several specific sites were likely attributed to the intensive influence of human activities. For example, the REEs concentrations at the 1 site were significantly higher than those at the 2 and 3 sites (Table 4), possibly due  Xixi River and Nanxi River to river closure by the dam of Wananxi Reservoir. The dam can extend the residence time of upstream water, resulting in the full absorption of REEs by clay mineral particles and organic matter particles, i.e., high REE concentrations of SPM. Generally, river water pH values decrease after flowing across a dam (Wang et al., 2011). A similar phenomenon also occurred in the present study; the river water pH values decreased from 7.58 (at the 1 site) to 6.72 (at the 2 site) ( Table 1). As river water pH decreases, many REEs associated particles or colloids are released into dissolved loads (Goldstein & Jacobsen, 1988;Johannesson et al., 2004;Migaszewski, Galuszka & Dolegowska, 2019), resulting in the decrement of REE concentrations of SPM. The significantly higher REE concentrations at the 28 site compared to the nearby sites (Table 4) were likely attributed to agricultural fertilization. The concentrations of major elements at the 28 site did not show anomalous features (Table 2), which can exclude the effects of rock weathering on REE concentrations of SPM. However, the river water NO 3 − concentration at the 28 site was 4 times higher than the average value of Xixi River (Liu & Han, 2020), which was closely associated with the loss of agricultural N. Since the extensive addition of REE within fertilizer (Altomare, Young & Beazley, 2020;Dushyantha et al., 2020;Volokh et al., 1990), agricultural erosion can cause the increment of REE concentrations of SPM. The low REE concentrations at the 34 site compared to the near sites (Table 4) (Tables 2 and 3). The result was attributed to the regolith materials which contained low-weathered carbonate minerals were denudated and translocated into river, which explained the low REE concentrations at the site.

Eeffects of physicochemical properties on REE fractionations of SPM
River SPM is mainly derived from local soils by soil erosion (Linders et al., 2018). However, the PAAS-normalized REE patterns of SPM showed a huge discrepancy compared with local 0.6 0.8 Xixi River Nanxi River soils (Chen et al., 1992) (Fig. 3). The discrepancy is mainly attributed to REE fractionations between the colloidal (solid) pool and dissolved pool (Elderfield, Upstill-Goddard & Sholkovitz, 1990). Overall, the (La/Yb) N and (Sm/Yb) N ratios of SPM were near 1.5, and the (La/Sm) N ratio was near 1 in the basin (Fig. 4), indicating the enrichments of LREE and MREE relative to HREE in SPM. The enrichment of LREE in SPM is closely linked with the strong adsorption by clay minerals, whereas HREE prefers to form stable soluble complexes (Da Silva et al., 2018). The river water pH plays an important role in affecting the fractionation degree between LREE and HREE in SPM. In the main stream of the Beixi Table 6 Pearson correlation coefficient between physicochemical parameters and REEfractionation proxies of SPM in different regions of the Jiulongjiang River. The normal distribution of the sample data set were tested via the Shapiro-Wilk test before Pearson correlation analysis. Statistical significant correlation at the level of P < 0.05 (*) and P < 0.01 (**).  was likely attributed to slight Gd pollution. Gd concentration of SPM at the 37 site was 1.4 times higher, while the concentrations of other REEs were about 1.1 times higher, compared to that at the nearby 36 site. This result indicates that the 37 site is affected by single Gd point pollution. However, the relatively lower δGd value (1.40) at the 28 site was likely associated with agricultural activities. Compared to other sites in the Xixi River, the riverine NO 3 − concentrations at the 28 site were significantly 4 times higher (69 mg/L vs. 15 mg/L) (Han et al., 2020;Liu & Han, 2020), indicating strong agricultural activities. Moreover, the concentrations of REEs in SPM at the 28 site were about 3-4 times higher than those at the other sites in the Xixi River, indicating the anthropogenic input of REEs from agricultural soils. Overall, the 28 site is affected by agricultural Gd pollution although the δGd value is less than 1.6.

CONCLUSIONS
The concentrations and fractionations of REEs in SPM were investigated in rivers regions (including the main stream and tributary of Beixi River, Xixi River, Nanxi River, and estuary) of the Jiulongjiang River. There were similar REE concentrations of SPM in the main stream of Beixi River, Xixi River, and Nanxi River, these mainly flow through the granite region; while it was lower in the tributary of Beixi River, which is associated with widely distributed carbonates. However, the lower REE concentrations of SPM in the estuary are mainly attributed to the removal of REEs with the increasing salinity. Overall, the REE concentrations of SPM in riverine water are primarily controlled by lithologic distribution, while it is also affected by salinity in the estuary. Riverine pH plays an important role in affecting REE fractionation of SPM in the river, but the fractionation is mainly controlled by salinity in the estuary. These results indicate that the increasing salinity from the inland river to the estuary affects the concentrations and fractionations of REEs in SPM.