Study area
This study area is part of the Pearl River Estuary (PRE). The PRE is located in the south of the Tropic of Cancer, with a humid, hot and rainy tropical and subtropical monsoon climate. The annual mean temperature is 22.2 °C and the annual average precipitation is 1154.8–2702.2 mm [32]. PRE is a bell-shaped semi-enclosed sea on the Guangdong coast of South China, occupying an area of about 2500 km2 [28]. The average annual inflow of the RRE into the sea is 1.124 × 1011 m3 [32]. Water depths in the PRE vary from 2–5 m in the western region, to about 15 m in the eastern region [28]. The PRE is perennially affected by an irregular semidiurnal tide, with a tidal coefficient between 0.94 and 1.77 [32]. The ranges of water temperature, salinity, alkalinity and pH are 16.64–30.09 °C, 0–35‰, 1.80–2.40 meq/L and 6.5–8.58, respectively [32,33,34].
Sampling and analysis of sediment properties
Undisturbed surface sediments of 0–3 cm were carefully gathered from 21 sites (Fig. 1) during summer tide in June 2018 using a triangular plastic spade. Five subsamples were obtained at each site and these were thoroughly mixed to achieve a typical sample. The samples from each site were then divided into half, with each part put separately into a clean plastic bag. One half of each sample was then freeze-dried, cleaned of visible debris such as rock and plant fragments, ground in a clean mortar, passed through a 2000-μm stainless-steel sieve for homogenization, and kept at − 20 ℃ until analysis of DGT-labile metals, organic matter (OM), and inorganic carbonate (CaCO3). The other part of each sample was stored at − 4 ℃ for the determination of particle size. Details of how to determine sediment properties (OM, CaCO3, and particle size) are provided in the Additional file 1.
DGT experiment
Procedure of DGT extraction
The Chelex DGT device used in this study is composed of a binding gel (0.40 mm thickness), a diffusive gel (0.90 mm thickness), and a filter membrane (0.80 mm thickness) (EasySensor Ltd, www.easysensor.net). These three layers are assembled with the dual-mode holder, which consists of an “O-shape” ring, a recessed base, and a hollow base that offers structural support and accommodates the gels [35]. Target metals diffuse through the membrane filter and diffusive gel, and are accumulated on the binding gel [15]. Afterwards, the metals associated with the binding gel are extracted using acid solution for further analysis.
The DGT application procedure to measure homogenized sediments complies with the method proposed by Wang et al. [9]. Accordingly, the method involves three sequential steps: (1) pretreatment of samples; (2) DGT deployment on samples; and (3) DGT retrieval.
In initial step, sample of about 10 g was weighted, placed in a beaker, added deionized water to reach 70–80% water holding capacity, mixed thoroughly. The well stirred sediment was covered with plastic wrap to prevent water evaporation and incubated for 48 h at a constant ambient temperature.
In next step, a small amount of rehydrated sediments homogenized was added to the open cavity of the DGT device using a plastic spoon, shaken gently to ensure the sediment settles fully and was in contact with the surface of the filter membrane, with more sediment then added to completely fill the cavity. The loaded DGT device was then transferred into a plastic, semi-opened bag for incubation for 24 h at a constant ambient temperature, with 1–2 mL deionized water added to the bag to maintain moisture during deployment.
In last step, 24 h later, the DGT core was unscrewed and pulled out of the base using another base. The surface of the filter membrane was rinsed with deionized water and the binding gel retrieved and placed in a centrifugal tube. 1.8 mL 1 M HNO3 was added to the tube to immerse the gel and then the tube was closed and kept at 4 °C for 16 h. The elution was collected for analysis of DGT-labile metals. The concentrations of Cd, Pb, Ni, Cu, Zn, Co, Fe, and Mn were determined by inductively coupled plasma mass spectrometry (ICP-MS). Absolute difference between two independent determination results obtained under repeatability conditions not exceeded 10% of arithmetic mean value. The research work related to DGT was performed in the laboratory of EasySensor Ltd.
Models for calculating DGT-labile metal concentration
The target metal observes Fick’s 1st law of diffusion in the diffusive layer. The target metal concentration from the DGT extracting solution can be transformed to a mass of metal (M) through Eq. (1) as follows:
$$M = \frac{{C_{e} \left( {V_{e} + V_{g} } \right)}}{{f_{e} }},$$
(1)
where Ce is the target metal concentration in the extracting solution; Ve is the volume (mL) of the extracting solution; Vg is the volume (mL) of gel (in this study, Vg is 0.2 mL); and fe is extraction rate, which is 0.938 for Cd, 0.955 for Pb, 1.05 for Ni, 1.03 for Cu, 0.88 for Zn, 0.975 for Co, 0.889 for Fe, and 0.967 for Mn [9]. The DGT-labile concentration is then calculated via Eq. (2):
$$C_{\text{DGT}} = \frac{M\vartriangle g}{{DAt}},$$
(2)
where △g is the thickness of the diffusive layer and equal to 0.9 mm in this study; D is the diffusion coefficient of the target metal, which for the studied metals relies on the EasySensor user manual (http://www.easysensor.net/col.jsp?id=109); A is the exposed surface area of the DGT device (3.14 cm2); and t is the deployment time, which is equal to 48 h in this study.
Toxicity data assemblage
The acute toxic data relevant to Cd, Pb, Ni, Cu, Zn, Co, Fe, and Mn were retrieved from the USEPA ECOTOX database that includes algae, crustaceans, and fish (https://cfpub.epa.gov/ecotox/). These aquatic species are from freshwater and saltwater media types.
In order to reduce errors in any species with varying acute toxicity data, and to reflect the specific metal toxicity to aquatic biota, the median values of metals in algae, crustaceans, and fish species were calculated as shown in Table 1. For fish, the EC50 values of Cd, Pb, Cu, and Zn are used in this study. Due to the paucity of EC50 data for the other metals in fish species, the LC50 data of Ni, Co, Fe, and Mn are conducted in this study. Detailed information about the values of toxicity data and aquatic species are given in the Additional file 1.
Table 1 The acute toxicity data of the studied metals and values of predicted no-effect concentration (PNEC) (μg/L) Ecotoxicological risk model
Single metals
The risk quotient (RQ), which is the ratio between the measured metal concentrations in the environment (MEC) and the predicted no-effect concentration (PNEC), is used to evaluate the ecotoxicological risk for each DGT-labile metal as given in Eq. (3):
$${\text{RQ}} = \frac{{{\text{MEC}}}}{{{\text{PNEC}}}}.$$
(3)
Table 1 explains how to calculate the PNEC and gives the values obtained in this study. The RQ is considered a robust method for assessing single pollutant ecotoxicology risk and is widely used in environment studies [19, 36, 37]. An RQ value less than 1 indicates no potential ecotoxicological risk, while an RQ value higher than 1 suggests a potential ecological risk and that the environmental risk posed should not be excluded [28, 38]. A larger value indicates a greater potential risk.
Mixture of metals
The RQ for the mixture based on MEC/PNEC ratios (RQMEC/PNEC) and the RQ for the mixture based on toxic units (RQSTU) are acceptable and extensively used models [19, 38, 39]. The two models are as follows in Eqs. (4)–(5):
$${\text{RQ}}_{{\text{MEC/PNEC}}} = \mathop \sum \limits_{i = 1}^{n} {\text{RQ}}_{i} = \mathop \sum \limits_{i = 1}^{n} \frac{{{\text{MEC}}_{i} }}{{{\text{PNEC}}_{i} }},$$
(4)
$${\text{RQ}}_{{{\text{STU}}}} = \max \left( {S{\text{TU}}_{{{\text{algae}}}} {\text{, STU}}_{{{\text{crustaceans}}}} {\text{, STU}}_{{{\text{fish}}}} } \right) \times \text{AF} = {\text{max}}\left( {\mathop \sum \limits_{i = 1}^{n} \frac{{{\text{MEC}}_{i} }}{{{\text{EC}}50_{{i, {\text{algae}}}} }}, \mathop \sum \limits_{i = 1}^{n} \frac{{{\text{MEC}}_{i} }}{{{\text{EC}}50_{{i, {\text{crustaceans}}}} }}, \mathop \sum \limits_{i = 1}^{n} \frac{{{\text{MEC}}_{i} }}{{{\text{EC}}50_{{i, {\text{fish}}}} }}} \right) \times {\text{AF,}}$$
(5)
where AF is the assessment factor (= 1000) [38].
In both models, if the RQs exceeded 1, the environmental risk posed by the mixture should be considered. The risk assessment was constructed considering firstly all the metals, and later only the metals with RQ below 1, to evaluate the potential risk of this mixture too [33].
Statistical analysis
Prior to factor analysis (FA), the normal distribution of each sediment-property variable and DGT-labile metal concentration was tested, which showed that only OM fitted normality. Therefore, the contents of Cd, Ni, Cu, Co, Fe, and CaCO3 were log-transformed, and the concentrations of Pb, Zn, Mn, and the median grain size were transformed with the Box–Cox method. These transformed parameters then fitted normal distributions. The transformed parameters and OM were ultimately standardized for FA. Data transformations were conducted with StatSoft STATISTICA 8.0 and FA was implemented with IBM SPSS Statistics 19.0.