Potentially toxic elemental contamination in Wainivesi River, Fiji impacted by gold-mining activities using chemometric tools and SOM analysis

Potentially toxic element (PTE) contamination in Wainivesi River, Fiji triggered by gold-mining activities is a major public health concern deserving attention. However, chemometric approaches and pattern recognition of PTEs in surface water and sediment are yet hardly studied in Pacific Island countries like Fijian urban River. In this study, twenty-four sediment and eight water sampling sites from the Wainivesi River, Fiji were explored to evaluate the spatial pattern, eco-environmental pollution, and source apportionment of PTEs. This analysis was done using an integrated approach of self-organizing map (SOM), principle component analysis (PCA), hierarchical cluster analysis (HCA), and indexical approaches. The PTE average concentration is decreasing in the order of Fe > Pb > Zn > Ni > Cr > Cu > Mn > Co > Cd for water and Fe > Zn > Pb > Mn > Cr > Ni > Cu > Co > Cd for sediment, respectively. Outcomes of eco-environmental indices including contamination and enrichment factors, and geo-accumulation index differed spatially indicated that majority of the sediment sites were highly polluted by Zn, Cd, and Ni. Cd and Ni contents can cause both ecological and human health risks. According to PCA, both mixed sources (geogenic and anthropogenic such as mine wastes discharge and farming activities) of PTEs for water and sediment were identified in the study area. The SOM analysis identified three spatial patterns, e.g., Cr–Co–Zn–Mn, Fe–Cd, and Ni–Pb–Cu in water and Zn–Cd–Cu–Mn, Cr–Ni and Fe, Co–Pb in sediment. Spatial distribution of entropy water quality index (EWQI) values depicted that northern and northwestern areas possess “poor” to “extremely poor” quality water. The entropy weights indicated Zn, Cd, and Cu as the major pollutants in deteriorating the water quality. This finding provides a baseline database with eco-environmental and health risk measures for the Wainivesi river contamination.


Introduction
In recent years, potentially toxic element (PTE) contamination in water and sediments of the aquatic system has attracted global attention owing to their persistence and non-biodegradable nature. Moreover, these PTEs are known to be toxic and pose a major risk to human and ecological health (Ali et al. 2016;Ali et al. 2018;Islam et al. 2018;2020a;Pandey et al. 2019;Ustaoğlu and Islam 2020;Amiri and Berntdsson 2020;Kumar et al. 2021a, b;. Considerable quantities of hazardous chemicals especially PTEs such as chromium (Cr), nickel (Ni), cobalt (Co), cupper (Cu), mercury (Hg), arsenic (As), cadmium (Cd), and lead (Pb) have been discharged into the river system due to the rapid expansion of emission sources, global population growth, improper wastewater disposal, and agricultural runoff (Türkmen et al. 2011;Srebotnjak et al. 2012, Islam et al. 2014Kabir et al. 2021;Sohrabi et al. 2021). Therefore, the fate of PTEs in the aquatic ecosystems is of particular importance as PTEs bind with suspended particulate or organic matter and ultimately end up bound with the riverine sediments (Varol and Sen 2012). However, the spread of PTEs to the aquatic bodies especially surface sediments occurs naturally and anthropogenically via chemical weathering motor vehicles exhaust gases, mining activities, toxic elements containing fertilizers and pesticides, various urban, and industrial activities (Tokatlı et al. 2015;. Furthermore, continuous accumulation of PTEs in the aquatic system severely degrades the environmental quality and unfavorably affects the biosphere structure, triggering changes in the biogeochemical cycles (Alizadeh et al. 2017;Maria Cavadas Morais Couto et al. 2019). The behavior of PTEs in the riverine freshwater is a function of the substrate and suspended sediment composition and the water chemistry (Mohiuddin et al. 2012). During transport, PTEs may undergo various changes in their speciation due to changes in sediment and water chemistry, reductive dissolution, precipitation, sorption, and complication process (Mohiuddin et al. 2011;Mohiuddin et al. 2012;. These changes not only affect the fate and transport of the PTEs, but also their toxicity. Therefore, the distribution, possible ecological risks, and source identification of PTEs in the riverine water and sediment are critical for efficient environmental risk perspective (Tepe et al. 2017;Kormoker et al. 2019a).
Investigation of PTEs in the surface sediments could be utilized to evaluate the anthropogenic effects and risks posed by improper waste discharge to the riverine ecosystems (Islam et al 2020a;Liu et al. 2018). Standard reference dataset on PTEs in aquatic media is very scarce in Fiji and is considered as a restricted access in generating the association between PTE accumulation and their source distribution (Kumar et al. 2021a, b). To the best of our knowledge, thus far, limited attentions have been paid to PTE evaluation in surficial water and sediments of Wainivesi River adjacent to Wainivesi gold-mining areas. Hence, accurate source distribution and ecological risk evaluation of PTEs in riverine sediments are urgently needed. Some sediment quality indices have been recognized and extensively applied by many researchers in order to evaluate PTE pollution and ecological risks in sediments from riverine aquatic environments (Ali et al. 2018;Hu et al. 2018;Islam et al. 2020b;Xiao et al. 2019). These indices are mainly classified into three categories, such as contaminant identification, background enrichment, and ecological risk (Kumar et al. 2020;Rinklebe et al. 2019;Song et al. 2017). Several methods have been used to determine the pollution level in sediments, the probable ecological risk index (RI), and the pollution load index (PLI) to assess the toxic effects of PTEs on sediments (Kormoker et al. 2019b;Islam et al. 2022). Therefore, this research has focused on the indexical and chemometric tools to evaluate the pollution levels of PTEs and source apportionment.
In recent decades, Tailevu Province, Central Division, Fiji has experienced fast urbanization due to its rapid industrial and gold-mining growth. Consequently, today the rapid unplanned urbanization and industrialization with the anthropogenic inputs of PTEs in this town are a matter of concern. The Wainivesi River basin is the most popular river system that crosses through the north of Navunisole town, posing a socio-economic benefit to its nearby riverine dwellers (Maata and Singh 2008). In Fiji, gold-mining activities have been evolved in the fourth century, and since that time, gold has been mined using the mercury amalgamation approach from alluvial sedimentary rock of Wainivesi River waterways and subsurface sediments along the side of dried-up valleys (Diarra and Prasad 2020). The economy of Fiji depends on gold-mining development, and gold is the country's most export product (Mines 2019;Diarra and Prasad 2020). In Fiji, most of the gold formed by goldamalgamation procedure derived from goldfields distributed in the Central Division of the country (Maata and Singh 2008), which is drained by the Wainivesi River system. About 7549 ha of landmasses has five tailings dams where all sludge waste particles from Votukoula gold-mining area are deposited (Kumar et al. 2021a). In this river system, PTEs come primarily from mineral ores processing and the disposal of tailing and wastewater across gold mines. This result in numerous detrimental effects on the geochemistry and ecosystem of the receiving water bodies (Grimalt et al. 1999; Mora et al. 2019). Thus, assessment of PTEs accumulation in the Wainivesi River sediment is a topic of concern for sustainable PTE management in Fiji.
At the solid gold tailings some sulfide mineralization occurs to produce acids that could trigger leaching of PTEs from gold tailing dams and thus makes a probable risk to the surrounding environment (Hadzi et al. 2018;Adewumi and Laniyan 2020). The gold-mining industries favor PTE mobility in the water bodies of Wainivesi River, Fiji and are the major cause for PTE pollution of this riverine ecosystem (Mora et al. 2019). Recently, gold-mining activities in Fiji have observed a pipeline failure around gold tailing dams around the Nasivi River which has since been demarcated as a key disaster event. Gold mining that includes on-site ore processing that produces large quantities of tailings can have major impacts off-site on groundwater, riverine streams, wildlife, and the inhabitants (Zabowski et al. 2001). A few studies have stated that inhabitants near the gold-mining area have raised their disquiet on the elevated level of contamination of PTEs in surface water and drinking water (Ackley 2008;Matakarawa 2018;Kumar et al. 2021a, b).
Multivariate statistical tools including principal component analysis (PCA) Saha et al. 2020), e.g., dimensionality reduction, is commonly used to aid the interpretation of complex datasets of water and sediment. To obtain more exact and robust outcomes, this study employed principal component analysis (PCA) for suitable source identification of PTEs. However, PCA is a linear tool that can cause an erroneous outcome when dealing with nonlinear datasets (Astel et al. 2007). In this study, to overcome the drawbacks of linear dimensionality reduction, we employed the self-organizing map (SOM) which is a robust nonlinear projection approach that portraits complicated high-dimensional datasets into a two-dimensional space and concurrently (Wang et al. 2020). Although combined use of SOM and cluster analysis can help to illustrate the diverse classes of PTEs, these analyses can reasonably be compared quantitatively (Nakagawa et al. 2020). The SOM is generally applied to classify and pattern recognition of hydro-geochemical features due to its high performance, and the use of PCA may further verify SOM's outcomes to identify the potential sources of PTEs in water and sediment (Nguyen et al. 2015;Wu et al. 2021;Amiri and Nakagawa 2021). Besides, SOM can exhibit local nonlinear associations among the observed parameters, whereas PCA only reveals global linear associations between parameters. The advantage of coupling receptor models (SOM and PCA) in the present study may be an appropriate tool for pollution source identification in the Fijian context. Numerous studies in the gold mine regions on the contamination and risks of PTEs via different environmental media have been well documented around the globe, e.g., in China (Huang et al. 2020), Brazil (Pereira et al. 2020), Nigeria (Adewumi and Laniyan 2020), Ghana (Hadzi et al. 2018), Bolivia (Pavilonis et al. 2017, Papua New Guinea (Kapia et al. 2016). However, only limited information is available on PTE contamination from the gold-mining activities around the Wainivesi River in Fiji so far (Matakarawa 2018;Kumar et al. 2021a). Therefore, this is the first study to appraise the contamination levels of PTEs in water and sediments of Wainivesi River, Tailevu Province, Fiji.
In the present research, we hypothesized that inhabitants of the Tailevu Province, Central Division, Fiji are concerned about the high levels of PTEs through aquatic media, and the local people are suffering from freshwater shortage, and most of people rely on the river water for domestic, industrial, and animal feeding purposes. Due to the PTE potential risk in water and sediment of Wainivesi River through the improper disposal of effluents from gold mining, this riverine water, sediment, and ecosystem may be polluted. Due to the surrounding area of the gold mine to one of the major rivers in Fiji (Wainivesi River), the possible movement of mine-induced PTEs to this river is of paramount concern. Thus, there is a crucial requirement to concurrently monitor the level of PTE contents and related eco-environmental risks through aquatic media from the gold-mining activities, to safeguard against the diffusion of diseases and fatalities. The main goals of this study are to estimate the pollution level of PTEs in river water and surficial sediments, to appraise the water quality and possible source patterns of the PTEs in the water bodies and to evaluate the ecological risk and possible source of PTEs in sediments. The novelty of this research is that it is the preliminary inclusive study on riverine water and sediments resource pollution affected by gold-mining area of central Fiji.

Study area description
This work was performed at Wainivesi River which passes via Wainivesi Mine that is situated 10.14 km northwest of Korovou town, in the inland of Tailevu. The various sampling locations are displayed in supplementary Table S1 and the sampling sites are outlined in Fig. 1. The location of the Wainivesi mine close to the Wainivesi River makes it an important region to assess the PTEs. Korovou town is one of 25 cities in Fiji and ranks 23 in the Fijian population. In the Wainivesi area most people are heavily dependent on farming. Major type of farming is vegetables and root crops. Also most farmers are commercial dairy farmers. Most of the cattle drink water from Wainivesi River. The climate is mostly tropical rainforest, experienced by a distinctive rainy period from six months (November-April) and a dry period is noticed from other six months (May-October). The annual mean precipitation ranged from 2100 to 4200 mm during the rainy period. The Wainivesi River runs from the hills that flow downstream sides which passes Korovou town. When heavy precipitation happens in the Korovou town, a waterway of the Wainivesi River gets huge household domestic waste material, which noticeably attributed to flood that accumulated with mud and other debris particles. The Tailevu Province, Fiji is witnessing a precise gold occurrence in the history of the resources industry.

Sample collections and procedures
Benthic surficial sediment and water samples were collected from 8 diverse sample sites of Wainivesi River, Tailevu, Fiji considering the presence of mining outlets ( Fig. 1, Table S1). Sample sites were ~ 500-1000 m apart from each other, and the sample processes encompass ~ 4.8 km of Wainivesi River. From each sampling site, vertical sediment columns (15 cm) were collected by using a polyvinyl chloride (PVC) pipe corer (Tamim et al. 2016;Khan et al. 2021). Sediment columns were then divided into three portions in terms of their vertical depth (0-5 cm, 5-10 cm, 10-15 cm) followed by preserving in zip-lock bags with proper labeling (Tamim et al. 2016;Khan et al. 2021). Sediment samples were oven dried (60 °C) and were sieved through a 2-mm nylon sieve to remove the botanical debris, lumps and stones. To obtain homogenized powder form, samples were then pulverized using a ring mill. However, representative surface water sampling collection procedures from each sample sites were identical as those of prior studies Islam et al. 2020a). Concisely, two sets of combined water samples from each sample sites were obtained in which a single set of sample was preserved as it was while the another set were filtered and acidified (con. HNO 3 ; pH < 2) for subsequent laboratory analysis ).

Analytical processes
Nitr ic acid-perchlor ic acid-hydrof luor ic acid (HNO 3 -HClO 4 -HF) based digestion procedure for the sediment samples was identical as those mentioned in Khan et al. (2015a, b). Briefly, ~ 100 mg of sediment samples was thermally decomposed by sequential acid treatments. Elemental concentrations in the sediment samples were measured by atomic absorption spectrophotometer (AAS) (Perkin Elmer: AAnalyst 400, USA). Data quality assuring protocols including instrumental calibration, procedure blank monitoring, using double deionized water for dilution and replicate measurements of relevant reference materials were same as those in Ahsan et al. (2019). For river-water samples, physical and chemical parameters including temperature, pH, oxygen reduction potential (ORP), electrical conductivity (EC), total dissolved solids (TDS), and dissolved oxygen (DO) were measured in the un-acidified samples by portable Horiba-52 Multi-meter (Japan) at the sample locations. On the other hand, AAS-based elemental analysis was executed by using the acidified water samples. Analytical processes and quality ensuring processes were essentially same as those of our previous studies Hasan et al. 2021;Kumar et al. 2021a). For quality control purpose, PTE abundances from replicate estimations (n = 3) of standard materials NIST-SRM-8704 (sediment) and NIST-SRM-1640 (normal water) along with their limits of detection (LOD: 3σ) and limits of quantification (LOQ: 10σ) are outlined in Table S2.

Water quality and eco-environmental indices assessment
An integration of particular and various indices was employed for assessment of PTE contamination level (Table 1). Eco-environmental indices such as I geo (Muller, 1979), EF (Kumar et al. 2021a), and CF (Hakanson 1980) were used for the evaluation of PTE contamination. On the other hand, PLI (Tomlinson et al. 1980), C d (Hakanson 1980), modified degree of contamination (mC d ) (Abrahim and Parker, 2008), and RI  were chosen for the integrated evaluation of pollution in sediment. Moreover, chemometric tool such as the entropy water quality index (EWQI) developed by Shannon (1948) was utilized for surface water in this study. In the current work, indexical approaches such as I geo , EF, CF, C d , mC d , PLI, and RI were adopted for appraising the sediment's contamination and possible ecological risks, while EWQI was employed for assessing the water quality status in the Wainivesi River.  applied element abundances of upper continental crust (UCC) as the standard dataset for computing these indices, while Rubio et al. (2000) recommended the application of regional base line datasets. Furthermore, the method of establishing limit datasets is to compare the major PTE contents in polluted and non-polluted sediments that are analogous in terms of mineralogical and textural perspective (e.g., Hornung et al. 1989). However, the regional changes of sampling collection sites between the polluted and non-polluted sediments may trigger the considerable variations in element compositions. Therefore, to have the optimal baseline data, deep sedimentary layer can be used (Siegel et al. 1994) to measure those aforementioned indices. According to our study, deep sediment datasets from sample site (6C) were employed for calculating these indices for sediments. The calculation results of geoaccumulation index (I geo ) and contamination factor (CF) will assist and help to provide a robust evaluation. Conversely, EWQI was applied in the information entropy theory, in which the entropy was regarded for taking into consideration as information. The adaption of this entropy theory to identify the weight of each parameter can lessen omission caused by ignoring the relative arbitrary weight (Pei-Yue et al. 2010;). In our work, the EWQI is employed to show the current status of the river water quality due to its recognition, strength, and reliability (Islam et al. 2020b). All the calculated formulas, brief explanation, and importance of these eco-environmental indices are outlined in Table 1 Self-organizing map (SOM) analysis Kohonen (1982) pioneered for visualizing and clustering learning based algorithm to solve non-linear issue, which is recognized as self-organizing map (SOM) analysis (Wang et al. 2020). In general, artificial neural network (ANN) method is used for imagining multi-dimensional datasets into lower-dimensional image (Céréghino et al. 2009). After training stage, SOM can rapidly make a map using the new datasets. Furthermore, having some random components based on training and identical topology it can show different results (Wehrens 2009;Lenard et al. 1999). In our work, SOM model was done for the assessment of surficial water and sediment samples of Wainivasi River, Fiji. SOM model provides a good graphical image of association between parameters. To visualize positive association, it exhibits analogous gradient pattern while anti-parallel pattern depicts negative association. To develop the graphical representation, projection in two-dimensional datasets is the calculation grid size in the self-learning procedure (Kohonen et al. 2001;Park et al. 2003). However, the k-means process is a robust tool for clustering SOM model (Hentati et al. 2010;Jin et al. 2011). To estimate the number of SOM nodes, heuristic formula was adopted (Vesanto and Alhoniemi 2000;Mari et al. 2010;Li et al. 2019).
Here, m is the SOM node number, and n is the input dataset number .
By using the SOM analysis to the log-transformed and standardized analyzed datasets (n = 32 for sediment and water) 49 (~ 42) output neurons were acquired. A total of 49 neurons were selected according to the heuristic Eq. (1) (Vesanto and Alhoniemi 2000). The neurons are represented on 9′7 rectangular grids with 49 neurons in this work. To select the optimal number of clusters, the DBI values were calculated for SOMs from 1 to 6 clusters after the training process, as presented in supplementary Fig. S1. The details procedure of SOM analysis can be found elsewhere (Kumar et al. 2021a), and SOM map was prepared using MATLAB (version R2020a) platform through neural clustering toolbox.

Statistical analysis
The multivariate statistical analyses such as principal component analysis (PCA) and hierarchical cluster analysis (HCA) were carried out to detect plausible sources using Statistical Package for the Social Sciences (SPSS, version 23) and Origin Pro (version 2015). To avoid the uncertainty of the PTE distribution, Kolmogorov-Smirnov (K-S) and Shapiro-Wilk (S-W) tests were employed by SPSS (SPSS, IBM, USA) (Amiri et al. 2017). PCA and one-way analysis of variance (ANOVA) test were executed at the significant level of p ≤ 0.05. Cluster analysis was adopted to detect the likelihood of the similarity of the PTEs in this work (Saha et al. 2020). It is a very popular and convenient multivariate approach that can handle the dataset based on the initial features. The relevant graphs were plotted using Origin Pro (version 2015) and MATLAB (version R2020a). Since the study sites and data were not large in size, the inverse distance weighted interpolation model was adopted for the spatial map of the PTEs for environmental media such as water and sediment using ArcGIS (version 10.4) ).

Characteristics of potentially toxic elements in sediment and water
The variation of different PTEs in sediment samples of Wainivesi River, Fiji is presented in Table 2. The average analysis of data (in mgkg −1 ) depicted an order of heavy metal accumulation in sediment that was Fe (50,417 ± 16,773) > Zn (1013 ± 1286) > Pb (112 ± 62) > Mn (89 ± 29) > Cr (58 ± 22) > Ni (49 ± 26) > Cu (28 ± 25) > Co (23 ± 8) > Cd (10 ± 13) respectively. The data indicated that Cd was marginally accumulated in the sediments, while Fe got extreme enrichment in the river basin. The percentage of relative standard deviation (%RSD) for the heavy metal distribution in sediments at different sampling points showed that abundance of Co, Fe, and Mn did not significantly vary, while the rest of the metals (i.e., Cd, Zn, Co, Pb, Ni, and Cr) significantly varied (%RSD: 39-128%). This study showed the metals were heterogeneously distributed at different sampling points suggesting the sources of these metals in sediments were mainly anthropogenic. The concentrations of the studied PTEs in the sediment samples were compared with the threshold values of the sediment quality guidelines (SQGs): Probable Effect Level (PEL), Threshold Effect Level (TEL), Severe Effect Level (SEL), Effect Range Low (ERL), Lowest Effect Level (LEL), and the Effects Range Medium (ERM) values (Rudnick and Gao 2003;Persuad et al. 1993;Macdonald et al. 1996;US-EPA 1999).
The average Cr concentration in sediment samples was noticed to be higher than LEL (Persuad et al. 1993), TEL (Macdonald et al. 1996), TRV (US-EPA 1999), while lower than the SEL (Persuad et al. 1993), PEL (Macdonald et al., 1996), ERL (Long et al. 1995), and ERM (Long et al. 1995). This study revealed that 95.8% samples were fallen between the LEL and SEL, while 58.3% samples were below the TEL. Therefore, it could be suggested that the present levels of Cr in the study area do not have significant impact on the ecological health. However, if a higher level of Cr would appear, then this could have adverse impact on the microbial community (Fashola et al. 2016). Nonetheless, the average Cr concentration was below the maximum acceptable Cr concentration in surface soil and the reported Cr level (486 mg kg −1 ) in gold mine tailing in Oman (Abdul-Wahab, S.; Marikar 2012). Subsequently, the average Cr concentration in sediment samples was lower than the several river sediments reported in literature (Tamim et al. 2016;Zhang and Wang 2001;Mathis and Cummings 1973;Pehlivan 2010;Garzanti et al. 2010;Pattan et al. 2008); Kim et al. 2010), which indicates that the sediment in Wainivesi River, Fiji was not severely contaminated with Cr.
This study revealed that the 100% Wainivesi River sediment for Mn concentration was much lower than the two SQG threshold values (LEL and SEL) (Table 1), which was consistent with the reported results for different river sediment in the world: Bangladesh (Tamim et al. 2016), China (Zhang and Wang 2001), Russia (Sorokina and Zarubina 2011), Turkey (Pehlivan 2010), Angola (Silva et al. 2016), Germany (Brügmann 1995), Ireland (Jones and Jordan, (1979), Malaysia (Elias et al. 2018), Congo (Atibu et al. 2016). Therefore, it has been suggested that the sediments in the study area were free from Mn contamination with respect to the SQG threshold values.
Nickel (Ni) concentration in 12.5% and 58.3% of the samples exceeded the SEL and PEL thresholds respectively Moreover, 87.5% of the samples were between LEL and SEL threshold and 33.3% of the samples were between TEL and PEL thresholds ( Table 2). The average Ni concentration was much higher than all the threshold values (LEL, SEL, TEF, PEL, ERL, ERM, and TRV) for the SQGs (Persuad et al. 1993;Macdonald et al. 1996;Long et al. 1995;US-EPA 1999). These results showed that Ni may be potentially harmful to the sediment-dwelling organisms. It is noteworthy to mention that Ni has been shown to adversely interfere with the bacterial cells. Studies have shown that Ni can damage the bacterial cell by (1) substituting fundamental metal in metalloproteins; (2) connecting to synergist deposits of non-metalloenzymes; (3) and by inducing oxidative pressure that enhances DNA fragmentation (Fashola et al. 2016). However, the average Ni concentration for this study was found to be higher than the reported results for sediments in many countries in the world (Tamim et al. 2016; Persuad et al. (1993) c Macdonald et al. (1996) d Long et al. (1995) e US-EPA (1999) f % of sample fall in different category g Rudnick and Gao (2003) h Turekian and Wedepohl (1961) Silva et al. 2016;Brügmann 1995;Jones and Jordan 1979;Elias et al. 2018;Atibu et al. 2016). The higher level of Ni was observed in the study area because of gold mining as well as Ni existing in gold-bearing ore as pyrrhotite (Fe (1−x) S), which can contain up to 5% of Ni (Fashola et al. 2016). In addition to gold mining, Ni may be acquired from the synthetic industry waste, burning, metallurgy, and metal plating (Palansooriya et al. 2020).
The average Cu concentration in the sediment samples was observed to be higher than LEL (Persuad et al. 1993), TEL (Macdonald et al. 1996), TRV (US-EPA 1999), while lower than the SEL (Persuad et al. 1993), PEL (Macdonald et al., 1996), ERL (Long et al. 1995), and ERM (Long et al. 1995). This study revealed that 95.8% samples were fallen between the LEL and SEL thresholds, while 58.3% samples were below TEL thresholds. The average Cu concentration in the sediments for this study was in line with the reported mean concentration of unpolluted soil (5 to 70 mg kg −1 ) (Kabata-Pendias and Pendias 2001) and lower than the Cu level (92.2 mg kg −1 ) in the gold mine tailings in Ghana (Bempah et al. 2013). However, the average Cu levels in Wainivesi River sediment (Fiji) were lower than the reported results of the river sediments in several countries in the world (Turekian and Wedepohl 1961;Zhang and Wang 2001;Sorokina and Zarubina 2011;Mathis and Cummings 1973;Silva et al. 2016;Garzanti et al. 2010;Pattan et al. 2008;Kim et al. 2010;Elias et al. 2018;Atibu et al. 2016).
Iron (Fe) concentration was found to be higher than many river sediments in the world: Bangladesh (Tamim et al. 2016), China (Zhang and Wang 2001), Russia (Sorokina and Zarubina 2011), Turkey (Pehlivan 2010), Angola (Silva et al. 2016), Germany (Brügmann 1995), Ireland (Jones and Jordan, (1979), Malaysia (Elias et al. 2018), Congo (Atibu et al. 2016). Subsequently, SQG threshold values showed that 75% of the samples were higher than SEL and 20% were between LEL and SEL, which indicates that Fe may likely cause potential harm to sediment-dwelling organisms. High level of Fe might be resulting from industrial activities, especially mining industry.
Zinc (Zn) concentration in sediment samples was ranged from 175.5 to 4693.9 mg kg −1 with an average value of 1013.4 mg kg −1 , which was much higher than the entire threshold values (LEL, SEL, TEF, PEL, ERL, ERM, TRV) for sediment quality guidelines (Table 2). However, Zn concentration in 75%, 71%, and 46% of the samples exceeded the SEL, PEL, and ERM thresholds respectively. Moreover, 29.2% of the samples were between TEL and PEL threshold, and 54.2% of the samples were between ERL and ERM thresholds. Zinc speciation is significant in deciding its harmfulness to microorganisms (Fashola et al. 2016). High concentrations of Zn show diverse inhibitory or harmful impact on cell activity and bacterial cell development. Furthermore, it was reported the presence of high Zn concentration, nitrification processes catalyzed by Nitrosospira sp. was reduced 20% (Mertens et al. 2007). Therefore, it has been suggested that Zn has a significant impact on the environment and ecological health. The average Zn concentration in sediments for this study was several folds higher than the reported results (8.9 to 65.7 mg kg −1 ) in the gold mine tailings in South Africa (Mitileni et al. 2011) and 77.56 mg kg −1 in Ghana (Bempah et al. 2013). Further, the average Zn concentration for this study was found to be higher than the reported results for the river sediments in many countries in the world (Tamim et al. 2016;Zhang and Wang 2001;Sorokina and Zarubina 2011;Pehlivan 2010;Silva et al. 2016;Brügmann 1995;Jones and Jordan 1979;Elias et al. 2018;Atibu et al. 2016). This may have occurred due to possible Zn leaching from gold ore bodies (Fashola et al. 2016), as Zn occurs in gold ore bodies in the form of sphalerite (ZnS), which is often associated with galena, and thereby accumulating in the sediments (Emenike et al. 2020).
A wide range of Cd concentration (1.3 to 45.9 mg kg −1 ) was found in the study area with an average value of 10.2 i Tamim et al. (2016) j  k Zhang and Wang (2001) l Sorokina and Zarubina (2011) m Mathis and Cummings (1973)   mg kg −1 , which was higher than the typical Cd concentration (1 mg kg −1 ) in unpolluted soil (USEPA 2001), and the reported Cd concentration (6.4 to 11.7 mg kg −1 ) in gold mine tailings in Tanzania (Bitala et al. 2009). On the other hand, the average Cd concentration was found to be higher than the entire threshold values (LEL, SEL, TEF, PEL, ERL, ERM, TRV) for the SQGs (Macdonald et al. 2000;Persuad et al. 1993;Macdonald et al. 1996;Long et al. 1995;US-EPA 1999) guidelines (Table 2). Furthermore, Cd concentration surpassed the SEL, PEL, and ERM standards in 25%, 37.5%, and 43% of the samples, respectively. Subsequently, 62.5% of the samples were between ERL and ERM thresholds, and 75% of the samples were between ERL and ERM. The outcomes represent the probability of destructive impacts caused by Cd on the biota. Cadmium is a harmful heavy metal to most organisms, and it influences numerous metabolic activities of soil microorganisms like nitrogen mineralization, carbon mineralization, CO 2 evolution, and protein activities (USEPA 2001). It was observed that the average Cd concentration for this study was found to be higher than the reported results for the sediments in many countries in the world (Tamim et al. 2016;Zhang and Wang 2001;Sorokina and Zarubina 2011;Pehlivan 2010;Silva et al. 2016;Brügmann 1995;Jones and Jordan 1979;Rahman and Islam 2009;Elias et al. 2018;Atibu et al. 2016). A few anthropogenic exercises bring Cd into the environment. Gold-mining activities are one of them as well as they occur in gold-bearing ore bodies as an isometric minor component in sphalerite and their focus relies upon the grouping of the sphalerite in the mineral body (Fashola et al. 2016).
The average Pb concentration in sediment was found to be lower than the threshold values of SEL, ERM, while higher than that of the LEL, TEL, PEL, ERL, and TRV. Pb concentration in 50% of the samples was greater than the PEL threshold, while 95.8% of the samples were between LEL and SEL thresholds. Moreover, 50% of the samples were between TEL and PEL thresholds, and 79.2% of the samples were between ERL and ERM thresholds (Table 2). Like as other fundamental divalent metals (Mn 2+ and Zn 2+ ), Pb 2+ might interfere with nucleic acids, proteins, and the alterations of the osmotic balance in the bacterial cells (Fashola et al. 2016). Therefore, excessive Pb in sediments might have adverse impact on ecological health (Rahman et al. 2014(Rahman et al. , 2022. The average Pb concentration for this study was found to be higher than the Pb in surface soils (32 mg kg −1 ) worldwide average (Kabata-Pendias and Pendias 2001) but was in line for Pb concentration (80 to 510 mg kg −1 ) in gold mine tailing (Ogola et al. 2002). The average Pb concentration was found to be greater than the river sediments in Bangladesh (Tamim et al. 2016), China (Zhang and Wang 2001), Russia (Sorokina and Zarubina 2011), Turkey (Pehlivan 2010), Angola (Silva et al. 2016), Germany (Brügmann 1995), Ireland (Jones and Jordan, (1979), Malaysia (Elias et al. 2018), and Congo (Atibu et al. 2016). This high Pb could be due the excessive use of gasoline additives, pesticides, as well as chemical fertilizers (Aboud and Nandini 2009), sand extraction (Khan et al. 2017;Madzin et al. 2015), and mining activities (Emenike et al. 2020). The excessive Pb in the study area may be attributed to the gold mining since Pb occurs in the form of galena (PbS) in gold ore (Ogola et la. 2002).
The PTE concentrations in surface water of Wainivesi River, Fiji were analyzed by AAS-based analytical technique, and the statistics data (i.e., minimum, median, mean, maximum, standard deviation) for each studied metal was presented in Table 3, in comparison with the different reported results in the literature and the threshold values set by different recommended bodies (ECR 1997;WHO 2011;EU 2003;USEPA 2001;EPA 2001). The average metal contents in the water samples showed the following increasing trend: Cd (12 ± 10) < Co (34 ± 15) < Mn (45 ± 32) < Cu (47 ± 38) < Cr (104 ± 24) < Ni (115 ± 20) < Zn (183 ± 295) < Pb (190 ± 16) < Fe (1632 ± 1335) µg L −1 respectively. The %RSD for the heavy metals distribution in water samples at different sampling points in Wainivesi River, Fiji showed a wide ranged (45-161%) for the metals of Co, Ni, Cu, Zn, Cd, and Pb, whereas %RSD values for the rest of the metals (Cr, Mn and Fe) were not varied significantly. However, it has been observed that the sources of most PTEs were anthropogenic origin. Moreover, our results indicated that most of heavy metals were not homogeneously distributed in the study area.
The pH of the water samples was within the standard level set by the WHO (pH: 6.5-8.5), showing alkaline in nature (Table S1). The low organic components may trigger the probable high pH values at these respective locations in the designated area. The average electrical conductivity (EC) (our work: 0.10; limit values: 0.25-2.5 mS cm −1 ) and total dissolved solids (TDS) (our work: 0.065; limit values: 1 g L −1 ) were compared to those of suggested standard limit values set by international laws (WHO 2011;ECR 1997;EU 1998;EPA 2001), while dissolved oxygen (DO) (our work: 9.49; limit values: 6 mg L −1 ) was comparatively greater than the permissible levels (ECR 1997). The comparison of PTEs in the Wainivesi River, Fiji with the baseline values of the current work river and other rivers across the globe is displayed in Table 3.
The maximum Cr concentration (122 µg L −1 ) was found at the sampling points of W7 ranging from 55 to 122 µg L −1 . This study showed that Cr concentration for 100% of the sampling points in Wainivesi River exceeded the thresh   (MHC, 2007). Subsequently, the average Cr concentration was observed higher than the reported results for river water in Bangladesh (Islam et al. 2020a, b;Bhuyan et al. 2017), China (Li and Zhang 2010;Gao et al. 2019;Wang et al. 2017;Meng et al. 2016;Xiao et al. 2014), India (Giri and Singh 2014), Vietnam (Thuong et al. 2013), Nigeria (Eneji et al. 2012), Turkey (Varol and Şen 2012), Crotia (Dragun et al. 2009), Pakistan ( Kazi et al. 2009). Therefore, it has been suggested that Wainivesi River is not suitable for drinking purposes. A reasonable cause of Cr in these locations might be disintegration in ultramafic volcanic stone by enduring and diagenesis measures, which prompted the elevated Cr levels (Zhitkovich 2011). The average Mn concentration (45 µg L −1 ) in surface water samples of Wainivesi River were found to be below than the threshold values recommended by the several organizations (ECR 1997;WHO, 2011;BIS, 2012;EU, 2003;EPA 2001;MHC, 2007). However, the maximum Mn concentration was found at the sampling point of W1, which is near to mine area and 50% water samples (W8, W4, W2, and W1) contained higher Mn concentration than that of the recommended bodies. However, the current climatic condition in this area upholds Mn transport as manganese hydroxide by framing carbonate layers (Maata and Singh 2008). While pH esteems from 6.62 to 7.83 in a large portion of the waters, the dissolvability of Mn will in general improve (Islam et al. 2020b). In view of this proof, higher Mn concentrations found in surface waters may have a geogenic beginning, related with chemical weathering and disintegration of mineral in the bedrocks (Kumara et al. 2021). The ranking order of the average Mn concentration in surface water was W3 < W5 < W6 < W7 < W8 < W4 < W2 < W1 respectively ranging from 5 to 96 µg L −1 . The average Mn concentration in surface water was in line with the reported result for the Nasivi River (Fiji), which is located near the Vatukoula Goldmine region (VGR), Fiji (Kumara et al. 2021).
The Ni concentration in 8 different sampling points was found in a wide range (72 to 140 µg L −1 ), and > 90% water sample contained higher Ni concentration compared to the threshold values set by the ECR (1997), WHO (2011), EU (2003, EPA (2001). The finding of this study is consistent with the reported result for the Nasivi River (Fiji), which is located near the Vatukoula Goldmine region (VGR), Fiji (Kumara et al. 2021). Similar to Mn, the highest Ni concentration was found at the sampling point of W1, which was 2 times higher than the drinking water guideline value set by WHO (1996). Furthermore, the average Ni concentration was found to be higher than the several river waters in the world: China (Han River: Li and Zhang 2010;Major Rivers: Gao et al. 2019;Huaihe River: Wang et al. 2017;Jialu River: Fu et al. 2014;Qiantang River: Su et al. (2013); Dan River: Meng et al. 2016;Tarin River: Giri et al. 2014), India (Giri and Singh 2014), Vietnam (Thuong et al. 2013), Nigeria (Eneji et al. 2012), Turkey (Varol and Şen 2012), Crotia (Dragun et al. 2009), Pakistan ( Kazi et al. 2009). The high level of Ni in Wainivesi River water could be due to leaching of Ni from gold-bearing ore as pyrrhotite (Fe (1−x) S), which can contain up to 5% of Ni (Fashola et al. 2016).
The highest Zn concentration (753 µg L −1 ) in surface water was found at the sampling point of W1, while the lowest Zn concentration (21 µg L −1 ) was found at the sampling point of W6 having an average value of 183 µg L −1 . The ranking order for Zn concentration in the study area was W8  Markich and Brown (1998) w Kazi et al. (2009)  < W2 < W4 < W1 < W6 < W7 < W3 < W5 respectively. The average Zn concentration in Wainivesi River waters was found to be below the threshold value set by the ECR (1997), WHO (2011), BIS (2012), EU (2003, EPA (2001), andMHC (2007). This finding suggested that the surface water in the study area might be free from Zn contamination as well as there is no coal combustion and waste incineration in the study area, which might be prime sources of Zn in surface water/environment .
This study revealed that >90% and >100% water samples of Wainivesi River, Fiji contained higher Cd and Pb level compared to the threshold values (Table 2) (2007). Among all the studied heavy metals, these two elements were serious concerns as a result of their excessive presence in water sample, and both metals and their compounds are generally toxic pollutants. The main sources of these two elements are natural. For instance, normally a huge amount of Cd (~around 25,000 tons per year) is released into the environment. Nearly 50% of this Cd delivered into streams through enduring of rocks, and some Cd is released through jungle fires and volcanoes. The rest of the Cd is delivered through anthropogenic inputs. Cd can be transported over significant stretches when it is absorbed by sludge. This Cd-rich sludge can contaminate surface waters just as soils (Islam et al. 2020a). Furthermore, water contamination containing Pb compounds results from Pb minerals in the mining activities. Additionally, with mining, Pb compound "tetra-ethyl lead" is applied as an added substance in gasoline in numerous nations (Rahman et al. 2019). This natural Pb compound is immediately converted to inorganic lead and winds up in water, in some cases even in drinking water.

Water quality assessment of PTEs in surface water
The Entropy Water Quality Index (EWQI) is based on the average weighted of the considered parameters that are able to determine the water quality (Singh et al. 2021). Entropy along with other relevant information delineates measurements that express a long-term behavior of random processes (Gorgij et al. 2017). When the concentration of PTEs is frequently above the drinking water standard, the water is unsafe for human consumption and agricultural usage (WHO 2011). Site samples are linked to domestic waste, household sewage, and industrial effluents, such as gold mine-derived waste material, all of which are heavily polluted due to Ni, Cd, and Pb concentrations in the surface water. To identify the water quality for the drinking purpose, the critical limit for EWQI estimation is set at EWQI = 100, implying that 100% of the tested samples are over the crucial value (> 100). Only 12.5% of the analyzed samples are of moderate grade (EWQI: 100-150), indicating that they can be utilized for residential, irrigation, and industrial purposes rather than drinking. In the research area, 87.5% of sampling sites provide water that is of poor to extremely poor quality, making it unfit for any use. The presence of elevated levels of contaminants such as Cd, Zn, and Cu in water samples may be the cause of poor water quality that might be resulted from anthropogenic inputs such as fast industrialization, urbanization, and the disposal of industrial waste materials. This study showed that the highest metal concentration was found at the sampling point of W4 followed by W8 > W6 > W2 > W7 > W1 > W5 > W3, while the descending trend for entropy water quality index (EWQI) was followed as W8 > W2 > W4 > W1 > W6 > W7 > W3 > W5 ranging from 148 to 468 (Table 3). Kumar et al. (2021a) reported that the surface water contamination was caused by the massive volume of gold mine wastes and related household products released into the Nasivi River. The EWQI results revealed the state of water quality concerning PTE pollution in the river system. Due to the elevated contents of PTEs in the river water, samples from the river were defined as unfit for consumption purposes and stated the origins of elements from towns that were from anthropogenic inputs (Xiao et al. 2019). The spatial distribution map of EWQI revealed that in comparison to the rest of the locations, the northern and northwestern areas have higher EWQI values (Fig. 2). Overall, the majority of the sites with good-quality water and the study locations are located in the southern and southeastern areas. Hence, in these riverine areas with poor water quality, appropriate management strategies for de-silting and constant monitoring are urgently needed (Xiao et al. 2019). Figure 3a outlined the EF values of the studied PTEs from the sediments of the Wainivesi River system in Fiji. The calculated EFs of the selective PTE contents like Cr, Mn, Fe, Co, Ni, Cu, Zn, Cd, and Pb are 1.298, 1.028, 1.0, 1.369, 2.793, 2.119, 2.936, 3.303, and 1.838, respectively, presented in Table S4. We can organize the EFs of the PTEs according to their mean in the following order: Cd > Zn > Ni > Cu > Pb > Co > Cr > Mn > Fe. Our result shows that except for Cr, Mn, Fe, and Co, the remaining PTEs had EF values greater than 1.5 (1.5 ≤ EF < 3), demonstrating minor enrichment of the PTEs in sediment and implying anthropogenic origins (Ustaoğlua and Islam 2020;Islam et al. 2020a). On the contrary, metals like Cr, Mn, Fe, and Co had low EF values, implying geogenic inputs as the key factors of these toxic elements (Ustaoğlua and Islam 2020;Varol and Sen 2012;Gao and Chen 2012;Abrahim and Parker 2008). Wang et al. (2008) reported that EF values of 0.5 EF 1.5 indicate that the origin of the elements is purely geogenic, but an EF value of > 1.5 indicates that the elements are from anthropogenic sources. Shirani et al. (2020) reported that dysprosium (Dy), Pb, and Ni are highly enriched compared to other metals, indicating an intense anthropogenic activity in the study area which was similar to our study.

Eco-environmental appraisals of PTEs in sediment
I geo is a critical ecological index that is used to differentiate between natural and human-caused metal sources, as well as to assess the contamination level of sediment samples (Islam et al. 2020a). The analyzed I geo values of the considered PTEs were included in supplementary Table S5 and delineated in the following descending order based on their mean (relative to the background dataset): Ni (0.748) > Pb (0.138) > Cu (0.016) > Co (− 0.061) > Zn (− 0.077) > Cr (− 0.119) > Cd (− 0.142) > Fe (− 0.466) > Mn (− 0.533) (Fig. 3b). The highest I geo value was found for Cd (3.801), while the lowest I geo value was found for Fe (− 2.455). Cd and Zn were observed as "strongly contaminated" in sampling sites 1 and 2, while Cu was observed as "moderately to heavily contaminated" in sampling sites 1and 2 while Ni was investigated as "moderately contaminated" in some sites 4, 6, 7, and 8. To compare with other studies, Song et al. (2019) observed similar findings in China's Zhauyua city gold mine. The river sediments were regarded as contaminated with Cd, Pb, Zn, and Ni, which already exceed the acceptable levels considering the background values in an earlier section (Table 2). Moreover, the I geo values for the majority of the PTEs indicated that the studied locations were moderately to severely contaminated. Previous studies from Souza et al. (2017) and (Huang et al. 2020) found a significant content of PTEs in sediments and tailing dams in gold mining and processing regions which was consistent with our findings. According to Müller (1981), the sediments of the Wainivesi River system were unpolluted to moderately polluted (0 < I geo < 1) for all PTEs.
The analyzed contamination factor (CF) values for the PTEs were depicted in Fig. 3 (Fig. 3c). In most sites, the CF values for Cd and Zn were > 2.5, indicating a high degree of contamination, while the CF values for other PTEs were 1 ≤ CF < 3, indicating a moderate level of contamination at all sites. The CF values for the examined PTEs at all sites showed a moderate to a high level of contamination, which could be attributed to the Wainivesi River system in Fiji absorbing a substantial volume of municipal, residential, and gold mine effluent. A significant level of Cd was observed in a few sites such as sites 1 and 3 (> 10.0), indicating a high level of contamination which could be attributed to the station receiving a large amount of municipal, agricultural, household, and factory effluent (Fu et al. 2009). Moreover, Cd could be attributed to the study area from the existence of a large-scale gold mine and wastewater drainage in the vicinity. Islam et al. (2015a, b) conducted a similar investigation in a developing country's urban river and reported that the CF value of Pb in sediment samples was 4.3 and 3.7 during the winter and monsoon seasons, respectively. Domestic wastewater drainage, industrial effluents such as mine effluents, and atmospheric deposition are the main sources of the higher level of Pb in the sediment. However, our CF records were higher than that of Hossain et al. (2021) who reported that Halda river sediment remained in uncontaminated condition except for Pb. Figure 4a summarizes the derived pollution load index (PLI) values of PTEs in sediments. The PLI value is greater than one, then the investigated region can be considered to be contaminated (Tomlinson et al. 1980) (Table S2). The PLI values for all sampling sites in the study site varied from 1.00 to 2.46, indicating that the sediment in the study river had a high level of pollution (PLI > 1). PLI values estimated from the six PTEs ranged from 0.8 to 3.9, with a mean of 1.95. PLI values for all sampling locations calculated and displayed in Fig. 4a to comprehend the actual distributions of the integrated PTE loads to assess the sequential contamination status. The PLI can provide some insights into the sediment quality to the general public. It also provides crucial information to policy makers on the state of pollution in the study region (Suresh et al. 2012).
The findings of the ecological risk index (RI) are displayed in Table S7. The value of RI in sediments according to descending order of Cd > Ni > Cu > Pb > Zn > Cr > Co > Mn and the RI for sole PTE was observed in the low to very high-risk category. ) was noticed as the highest and noteworthy PTE in all sites, and the risk factors of it were low to very high-risk category for the aquatic environment, where the remaining PTEs were observed to be a considerable risk for the aquatic environment. After integrating the RI of each PTE (Fig. 4b) with its ranking category, it is observed that the analyzed PTE revealed considerable to very high-risk group. In general, the major sources of Cd and Ni in surface sediment are the intense uses of phosphate fertilizers to the agro-farming site beside the river basin and waste disposal material from the rural area (ATSDR 2008). The RI values were 40.9 to 733.8, which show moderate to very high ecological risk in the sampling sites (Fig. 4b).
Similar to the PLI and RI, C d and mC d exhibited similar spatial patterns (Fig. 4c,d) in the studied sites. The higher amount of C d was noticed at sites 1 and 2 (> 40.00) indicating a very extreme degree of contamination which might be due to receiving the enormous amount of rural waste, agro-farming practices, and mine-derived wastewater at these sites (Fu et al. 2009) (Table S6).
C d values ranged from 8.0 (7.1) to 62.4 (2.1) with a mean value of 22.36, which suggested that the Wainivesi river basin possess considerable to high level of contamination in the northwestern site (Fig. 4c). In contrast, values mC d varied from 0.9 (7.1) to 6.9 (2.1) with a mean value of 2.48 (n = 24), which implied that the tested river system was slightly to moderate heavily contaminated in mostly norwestern part near the gold mine where Cd, Zn, and Ni were the key contributors (4d).

Quantifying source apportionment of water and sediment
In this research, to quantify plausible sources of PTEs in the surface water and sediment from the Wainivesi River, Fiji, hierarchical cluster heat map (HCHM) and principal component analysis (PCA) were performed in Fig. 5. To confirm the appropriateness of the dataset for PCA, the Kaiser-Meyer-Olkin (KMO) and Bartlett's sphericity (BS) tests were validated before performing PCA. The KMO test score was 0.69 and 0.73 for water and sediment samples and the confidence level of BS test at p < 0.05, indicating the both dataset in this study was appropriate for PCA. For water sample dataset, three PCAs which elucidated 89.75% of the total variance were attained via PCA (Table S8 and Fig. 5a). The first component (PC1) responsible for 54.22% of the variance and had weak positive loading on Zn (0.41), and Mn (0.43) while weak negative loading on Cr (− 0.41) and Co (− 0.40). Zn showed noteworthy spatial variation, while Mn did not show noteworthy spatial changes, and the highest contents of Zn were found at W-1 due to the lithogenic origin and wastewater release from mine effluent from gold-mining activities. Results showed that Co and Cr may be derived from geogenic contents (Guan et al. 2018). The runoff from the hilly region may be a plausible transport way of these contaminants into river water. Zn and Mn concentrations are mostly driven by geogenic sources. Hence, PC1 was contributed to natural sources.
The second component (PC2) explained for 25.52% of the variance and had moderate positive loadings on Fe (0.64) and Cd (0.55) (Table S8). In majority of water samples, contents of Fe and Cd were five times and four times greater than the WHO standard limit values. Fe is among the most abundant elements in the Earth's crust, thus may be derived from weathering of the rock-water interaction (Bodrud-Doza et al. 2016). The mean concentrations of Fe and Cd exhibited significant variations among the sampling sites, implying that there can be both mixed sources (geogenic and anthropogenic) for these PTEs along the river to change their spatial patterns. Indeed, the highest concentrations of Fe were recorded at site W4 due to both geogenic and anthropogenic sources such as mine waste discharge and farming activities. The recorded Cd at site W1 was due to atmospheric deposition and electroplating effects ). Long-term use of Cd as a component in fungicides and algaecides (Bodrud-Doza et al. 2016) has also been observed to be beneficial to agro-farming field. Mining industry, for example, generates and discharges wastages containing high levels of Cd in the effluent . Fe and Cd are controlled by both geogenic sources and anthropogenic inputs. Hence, the PC2 was attributable to diverse sources.
The third component (PC3) accounted for 9.99% of the variance had strong positive loadings on Ni (0.73) and weak negative loadings on Pb (− 0.36) and Cu (− 0.38) (Table S8). Ni and Cu derived from anthropogenic inputs, especially, waste, and agro-farming fields (Cheng et al. 2020;Islam et al. 2020a). Ni and Pb concentrations were two times and nineteen times higher than the guideline values for threatening aquatic organisms and for drinking water purpose (Table 3). Motor exhaust contributes a significant amount of Pb to the water body, in addition to oil leakage from boats and steamers (Namngam et al. 2021). These findings indicated that these PTEs are mainly controlled by anthropogenic processes. Hence, PC3 was attributable to anthropogenic sources. Overall, the PCA showed two main sources of PTE contamination in the aquatic environment were humaninduced and geogenic contents.
For sediment sample dataset, three PCAs which explained 93.96% of the total variance were attained via PCA (Table S8 and Fig. 5b). The first component (PC1) was responsible for 55.20% of the variance and had weak positive loading on Mn (0.40), Cu (0.40), Zn (0.39), Cd (0.38), and Pb (0.42). In most of sampling sites, Zn, Pn, and Cd were about fifteen times, five times, and thirtyfour times higher than their guideline values. Fertilizers, electroplating, Cd containing alloys, foils, oils, and other applications use Cd that deposited into riverbed sediment (Namngam et al. 2021). Pb can come into the river sediments from urban waste leach out comprising Ni-Cd batteries from automobile factories, batteries, and smelting electroplants (Fang et al. 2019). Earlier cited work revealed that Pb and Zn in the riverine basin were likely attributed to release from industrial sewages and wastes (Islam et al. 2020a;Namngam et al. 2021). Currently, this is occurring daily at the study site near the gold mine as dumpling sites for their wastewater materials. This finding is in line with sediment pollution in other regions of the world (Emenike et al. 2020). On the other hand, the sources of Mn and Cu could be related with lithogenic inputs including soil erosion that enhances during the wet period due to the surficial run-off (Islam et al. 2020b). Mn may be discharged into the water via biogeochemical process of pyroclastic sediment (Ahmed et al. 2019). Hence, PC1 was contributed to mixed sources.
The second component (PC2) explained for 31.25% of the variance and had moderate loadings on Co (0.51), Cr (0.48), and Ni (0.50) (Table S8). In sediment samples, their mean contents were lower than the standard values for protecting aquatic lives (Table 3). This finding suggested that these PTEs are mainly controlled by geogenic processes viz. municipal and mine waste releases (Emenike et al. 2020; Hence, this component was contributed to geogenic sources. The third component (PC3) accounted for 7.51% of the variance had strong negative loading on Fe (− 0.75) and weak positive loadings on Co (0.43) and Pb (0.39) ( Table S8). Fe is related to a likely blend of organic matter overlapped on geogenic content (Emenike et al. 2020). Co may be originated from geogenic sources (Gao et al. 2016;Guan et al. 2018). In this study, Pb showed significant spatial variation, and Co did not show noteworthy spatial changes. The mean content of Pb was about five times greater than guideline value. This was due to substantial anthropogenic inputs. Furthermore, the 90th percentile contents of Cu and Pb were below the standard limit values for safeguard of freshwater aquatic system and for consumption purpose (Table S8). Pb has a long history of use as an anti-corrosive element in steels and gold industry, as well as an anti-knocking ingredient in gasoline and diesel fuels (Emenike et al. 2020). These outcomes showed that Fe and Co concentrations are influenced by geogenic source while Pb is driven by substantial anthropogenic sources. Thus, this last component was contributed to mixed sources. Overall, the PCA results suggested two major sources of PTE contamination in the sediment.
The two-way hierarchical cluster heat map (HCHM) and dendograms were generated in this research using the Ward linkage technique with Euclidean distance, and the outcomes are displayed in Fig. 6. The HCHM is often applied to verify the PCA findings. In this research, for water samples, the HCHM grouped the 9 PTEs into three clusters (Fig. 6a) in the vertical portion. The first (Pb, Fe, Cr, and Ni) containing sites 1 and 2 and second clusters (Co, Mn, Cd, and Cu) consisting of sites 3, 5, 6, and 7 could be ascribed to the mixed sources of both geogenic and human-induced sources (Emenike et al. 2020;Rakib et al. 2021). The third (Zn) cluster representing sites 4 and 8 indicates anthropogenic sources. Thus, the HCA results were in good agreement with the PCA outcomes.
For sediment samples, in the vertical portion, the dendrogram presented two clusters: cluster 1 was limited to Fe and Cu which consisted of sites 1 and 2 indicating geogenic source, while cluster two represents Cr, Zn, Cd, Co, Pb, Ni, and Mn containing the remaining sites, indicating intense anthropogenic sources (Fig. 6b). The Pb, Cd, Zn, Ni, and Cr in the river system were found to originate from Environmental Science and Pollution Research (2022) 29:42742-42767 42759 traffic-related sources, e.g., fuel consumption, wear of vehicle components, tire wear, vehicle emission (Tusher et al. 2020;Zhao et al. 2014) as well as industrial emission activities (Jiang et al. 2018;Xiao et al. 2019;Ali et al. 2021). Islam et al. (2018) and Islam and Al-Mamun (2017) also reported that the Cd and Pb originated from industrial activities, e.g., metal processing, smelting, and incineration waste. Such findings verified a comparable source of the selected PTEs in PCA results.

Spatial pattern of PTEs in water and sediment using SOM analysis
The component planes of SOM outcomes were outlined in Fig. 7. SOM planes were developed in color codes for exhibiting the significance of specific variables for each SOM class. The smaller the space of hexagon is, the more analogous the features of samples are. In the planes, a similar color revealed a positive association between variables, while different colors implied negative associations. For water dataset, each parameter resembled as illustrated in Fig. 7a. In the planes, three spatial patterns are evident to the surface water. First, Zn and Mn exhibited likely patterns to Co and Cr; their weight values enhanced from the lower to the upper sides. Second, contrasting other PTEs contents, Cd and Fe revealed horizontal gradient (elevated from the left to right side) in color patterns, indicating that both PTEs are controlled by mixed procedures from those affecting the main elements. Third, Ni and Pb values revealed more complicated color patterns than other elements with rising from the lower left to the upper right sides (Fig. 7a) which may be due to the impact from untreated waste material from rural area and gold mine-induced effluents, atmospheric precipitation, and roadside dust particle (Islam et al. 2015a;2020a). Cu was similar to Zn and elevated concentrated at the lower left corner which might be due to high agro-chemical fertilizer usage linkage with vegetable besides the river system (Dash et al. 2020). Interestingly, the pollution sources of PTEs in surface water from the SOM analysis showed an analogous pattern of PTE distribution in the clusters and PCA in this research. However, three spatial patterns were recognized in these PTEs, which can make them difficult to interpret in situations where PCA is employed to classify the 3 PCs in these PTEs. Therefore, SOM led to a detailed pattern recognition, which is in line with the outcomes of other studies (e.g., Amiri and Nakagawa 2021;Kumar et al. 2021a).
For surface sediment samples, three spatial patterns are obvious in the planes (Fig. 7b). First, Zn and Cd depicted similar patterns to Cu and Mn; their weight values enhanced toward the left to the right corners. These PTEs indicated a common association among the elements. Second, analogous to these PTEs, Cr and Ni are also alike color patterns and extremely distributed on the lower left corner. These toxic elements are derived from human-induced inputs, particularly, waste and extensive agro-farming production (Tepanosyan et al. 2021;Islam et al. 2020a). Third, Fe, Co, and Pb depicted complex color distributions than other PTEs, indicating extensive human-induced and geogenic inputs are liable for governing the sediment quality in the river basin. Such findings of SOM analysis also verified a comparable source of the selected PTEs in PCA outcomes in earlier section. Thus, the SOM model gives more robust pattern recognition of PTEs than PCA analysis.

Conclusions
This is the first inclusive study to investigate contamination levels of PTEs in surface water and sediment of Wainvesi River, Fiji induced by gold-mining activities. The results revealed that the contents of Cd, Pb, and Ni for water, and Zn, Pb, Cd and Fe for sediment exceeded the standard limits, which implied that the Wainivesi River is severely polluted by PTEs and may pose an adverse impact on this aquatic environment. The results of EWQI revealed that 87.5% of surface water poses poor to extremely low water qualities. The indexical approaches such as I geo , EF, PLI, CF, mC d , and C d confirmed that the sediments were heavily contaminated by PTEs at various degrees; however, diverse spatial changes were noticed for Cd and Zn. The value of RI in sediments followed the descending order of Cd > Ni > Cu > Pb > Zn > Cr > Co > Mn. The potential RI for PTEs showed low to very high-risk category where the highest level of ecological risk was observed in the designated area due to Cd and Ni.
The pollution sources of PTEs in surface water and sediment from the SOM analysis showed an analogous pattern of PTE distribution in the clusters and PCA in this research. Thus, the integrated tools with the use of GIS-based spatial patterns were robust and important for the source apportionment of PTEs in the analyzed samples. The results of PCA analysis depicted two types of sources in sediment (e.g., geogenic and anthropogenic sources) and, similarly, two types of sources in water (e.g., lithogenic and humaninduced inputs) as the reasons for water resource pollution. In summary, source identification of PTEs and zonation of eco-environmental logical risk attained from the present research will be helpful for decision makers to articulate sustainable-based pollution control measures for the goldmining activities. This study suggested that the geochemical speciation of PTEs in the Wainvesi River ecosystems, Fiji should be monitored and evaluated on a regular basis to ensure environmental sustainability.