Multivariate statistical evaluation of dissolved heavy metals and a water quality assessment in the Lake Aha watershed, Southwest China

Study site

The Lake Aha watershed is located in Guiyang City, Southwest China (26°34′N; 106°43′E), which holds a total area of 180.2 km² with a population of almost 150,000 and provides the vital water source for the local city (Han, Tang & Xu, 2010). The watershed is situated on the Yunnan–Kweichow Plateau, the global largest karst region, where karst aquifers and conduit networks are especially vulnerable to anthropogenic contaminations (Sun et al., 2019). The study site has a subtropical humid and monsoon climate, with a mean annual temperature of about 23.2 °C. The average annual rainfall ranges from 1,140 to 1,200 mm, while the average annual evaporation is 932 mm (Bai et al., 2007; Pan et al., 2019). The average annual rainfall days are about 178 days and most of the precipitation is mainly from May to August, accounting for about 65% of the total, with a relatively short annual average sunshine time over 1,412.6 h.

Lake Aha has three major inflowing tributaries, namely the Youyu River (YR), the Baiyan River (BR) and the Jinzhong River (JR). The bedrock of the YR is Permian limestone and Triassic shale, while that of the BR is Triassic dolomite and shale, and that of JR is Triassic dolomite (Song et al., 2011). The YR covers an area of 61.9 km² and flows eastward for 18.5 km through 11 villages. Agricultural land (53.0%) is the primary land use type of the YR watershed, whereas residential and commercial land is small (3.3%). The BR drains 51.5 km² and is approximately 15 km long. The primary land use type of the BR watershed is forest (40.2%). The JR has a length of 16.5 km and a watershed area of 47.5 km² with around 60,000 inhabitants. The dominant land use type of the JR watershed is residential and commercial land (81.6%), indicating the JR is intensively affected by human activities.

Sampling and analysis

River water samples were collected from 10 cm below the surface water level, at 18 sites in the tributaries, once a month from November 2017 to April 2018 (low flow season). The sampling sites are shown in Fig. 1. Specifically, nine sites were located in the YR, six sites were located in the BR, and three sites were located in the JR.

Ultimately, a total of 103 river water samples were collected, however, five of them were damaged during transportation. The pH, electrical conductivity (EC) and dissolved oxygen (DO) values of water samples were immediately measured in the field using the WTW Multi3430 (WTW Company, Weilheim, Upper Bavaria, Germany). The samples were instantly filtered through 0.22 μm acetate cellulose-acetate membrane after collection. All the water samples were acidified with ultra-purified HNO₃ (pH < 2), then sealed in pre-cleaned polyethylene bottles and kept at 4 °C until determination.

Heavy metals analysis

Nine heavy metals (Mn, Co, Al, Ni, Ba, V, Sb, Fe and Sr) were detected by using ICP-MS (NexION300X; Perkin Elmer, Waltham, MA, USA) at the Institute of Geochemistry, Chinese Academy of Sciences. Quality assurance and quality control were assessed by standard operating procedures, calibration with standards, and analysis of reagent blanks, with each batch of 20 water samples. Relative standard deviations for heavy metals were ∼±5% and recovery percentage ranged from 90% to 110%. Otherwise, the samples were detected again until the data reach the standard.

Statistical analysis

Multivariate statistical approaches, including correlation analysis and PCA, were used to identify the possible origin of nine heavy metals in the study area (Ustaoğlu & Tepe, 2019; Xiao et al., 2019). PCA is the most common approach to explore the sources of heavy metals by reducing the dimensionality of the dataset to several influencing factors (Loska & Wiechuła, 2003; Zeng & Han, 2020). The suitability of the datasets for the PCA method is assessed by Kaiser–Meyer–Olkin (KMO) value and Bartlett’s sphericity test (p < 0.001) (Varol, 2011). Microsoft Office 2019 and the statistical software package SPSS 21.0 were used to perform all of the data processing.

Health risk assessment

The hazard index (HI), the sum of both two pathway HQs, represents total potential non-carcinogenic risks for each heavy metal. If HQ or HI exceeds 1, it indicates a potentially adverse effect on human health and needs for further study (Wang et al., 2017). The HQ and HI are calculated as follows: (1) $$\mathrm{A}\mathrm{D}{\mathrm{D}}_{\mathrm{i}\mathrm{n}\mathrm{g}\mathrm{e}\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}}=\left(C_w\times \mathrm{I}\mathrm{R}\times \mathrm{E}\mathrm{F}\times \mathrm{E}\mathrm{D}\right)/\left(\mathrm{B}\mathrm{W}\times \mathrm{A}\mathrm{T}\right)$$ (2) $$\mathrm{A}\mathrm{D}{\mathrm{D}}_{\mathrm{d}\mathrm{e}\mathrm{r}\mathrm{m}\mathrm{a}\mathrm{l}}=\left(C_w\times \mathrm{S}\mathrm{A}\times K_p\times \mathrm{E}\mathrm{T}\times \mathrm{E}\mathrm{F}\times \mathrm{E}\mathrm{D}\times {10}^{-3}\right)/\left(\mathrm{B}\mathrm{W}\times \mathrm{A}\mathrm{T}\right)$$ (3) $$\mathrm{H}\mathrm{Q}=\mathrm{A}\mathrm{D}\mathrm{D}/\mathrm{R}\mathrm{f}\mathrm{D}$$ (4) $$\mathrm{R}\mathrm{f}{\mathrm{D}}_{\mathrm{d}\mathrm{e}\mathrm{r}\mathrm{m}\mathrm{a}\mathrm{l}}=\mathrm{R}\mathrm{f}\mathrm{D}\times \mathrm{A}\mathrm{B}{\mathrm{S}}_{\mathrm{G}\mathrm{I}}$$ (5) $$\mathrm{H}\mathrm{I}=\sum \mathrm{H}\mathrm{Q}s$$where ADD_ingestion and ADD_dermal are the average daily doses via ingestion or dermal exposure (μg/kg/day), respectively. C_w is the heavy metal concentration of each sample (μg/L); IR is the ingestion rate (L/day); EF is the exposure frequency (day/year); ED is the exposure duration (years); BW is the body weight (kg); AT is the average time (days); SA is the area of exposed skin (cm²); K_p is the dermal permeability coefficient for each heavy metal in water (cm/h); ET is the exposure time (h/day); ABS_GI is the gastrointestinal absorption factor. The above parameters are from the United States Environmental Protection Agency (EPA) (United States Environmental Protection Agency, 2004).

Water quality index

The WQI is calculated as follows: (6) $$\mathrm{W}\mathrm{Q}\mathrm{I}=\sum \left[W_i\times \left(C_i/S_i\right)\right]\times 100$$where $W_i$ is the weight of each element and represents different contributions to the overall water quality, which is calculated by the eigenvalues for each principal component (PC) and factor loading for each heavy metal from the PCA results. $C_i$ is the concentration of each heavy metal tested in this study. $S_i$ represents the limit value of drinking water for each heavy metal. According to the WQI values, water quality can be classified into five categories as excellent water (WQI < 50), good water (50 ≤ WQI < 100), poor water (100 ≤ WQI < 200), very poor water (200 ≤ WQI < 300) and undrinkable water (WQI ≥ 300). V and Sr are excluded from the WQI calculations due to the lack of official drinking water guidelines.

Kolmogorov–Smirnov test of data

Kolmogorov–Smirnov statistics indicate most of the water parameters and heavy metals were in non-normal distribution (Table 1). Therefore, we reported the median concentrations instead of mean concentrations in our study.

Physicochemical characteristics

The physicochemical characteristics of water samples from the Lake Aha watershed are summarized in Table 1. Samples from three tributaries demonstrate slightly alkaline characteristics with a median pH value of 7.88, varying from 6.71 to 8.64. Relatively, the median pH value of the YR (7.64) is lower than the BR (8.05) and the JR (7.96). The DO values range from 0.42 to 18.64 mg/L, with a median of 9.59 mg/L. The EC values range from 199.4 to 1437.0 μS/cm, with a median of 584.8 μS/cm. Overall, the pH, DO and EC values in the Lake Aha watershed show a noticeable variability.

Heavy metals content

The concentrations of dissolved heavy metals are also summarized in Table 1. According to their concentrations, these heavy metals can be divided into three groups: (1) Sr is in the highest level, with a median concentration of >100 μg/L; (2) Fe, Ba, Al, Ni and Sb are in intermediate levels, with median concentrations ranging from 1 to 100 μg/L; (3) Mn, V and Co are in low level, with median concentrations of <1 μg/L.

Table 2 shows the comparison of heavy metals concentrations between this study and the guideline values for drinking water set by China Environmental Protection Administration (2006), United States Environmental Protection Agency (2004) and World Health Organization (WHO) (2006). The guideline values set by China Environmental Protection Administration (2006) are the strictest and most comprehensive of the three. Thus, we used Chinese guideline values as the standard to evaluate water quality in this study. The median concentrations of nine heavy metals in this study are within limited values of drinking water guidelines (except V and Sr, without limited values in three guidelines), indicating the water is not heavily polluted by these heavy metals. However, for specific metals, their concentrations exceed the limit values at some sites (e.g., Mn at YR-5, Sb at BR-14). This indicates that these sites can attribute to a relatively high anthropogenic input. Besides, if the concentration of one sampling site is higher than other sampling sites (even if lower than guideline values), such as the Al concentrations in YR-5, it may also reveal an impact of anthropogenic input.

Comparison and distribution of heavy metals in tributaries

In the low flow season, the temporal distribution of each heavy metals in tributaries is shown in Fig. 2. In general, the Al and Mn concentrations showed an increasing trend with time, whereas the Sb and Ba concentrations displayed a decreasing trend. Meanwhile, the trends of the same heavy metal between three tributaries could be different. For example, the Co concentration of the JR increased with time in contrast to the stable trend in the other tributaries, and Ni concentrations of the YR and the JR was higher in November and December while BR did not change obviously. It is noteworthy that the JR showed more significant variation than the YR and BR, perhaps due to the small number of sampling sites (n = 3) on the JR.

The spatial distribution of heavy metals in three tributaries is shown in Fig. 3. Spatial distributions of heavy metals displayed an extensive variation, and we could conclude three primary patterns based on the spatial distribution features: (1) Al, Mn, Co and Ni, their concentrations increased in the middle reaches of the YR. (2) Fe and Sr, their concentrations were higher in YR but occurred valley values in the upstream. (3) Ba, V and Sb, their concentrations were consistently low in the YR but relatively high value in the other two tributaries.

Different anthropogenic activities may explain the distinction on dissolved heavy metals among three tributaries. The lowest pH value in the YR exhibited an influence of acid mine drainage. The JR was the most typical urban tributary impacted by human activities in our study, and the Jinyang Wastewater Treatment Plant with limited capacity and damaged sewer networks provided an amount of pollutant input (Huang et al., 2009; Song et al., 2011; Zhang et al., 2019). The peak value in BR-14 (a groundwater well) could attribute to the contamination of domestic sewage through karst conduits.

We compare our results with previous results of the upstream of Xijiang River (Beipan River and Nanpan River), Guizhou, China and the world average concentrations (Li, 1982; Liu et al., 2017) (Table 2). The Ni, Sb and Sr concentrations of our study were higher than the worldwide river average concentrations, while Co and Fe concentrations were comparable. Mn, Co, Ba, V and Sb concentrations were comparable to that in the upstream of Xijiang River, but the Sr concentration in both sites was much higher than the worldwide river average. Notably, the geological background of the upstream of Xijiang River was consistent with our study site, indicating that the bedrock lithology, as a way of natural source, could significantly affect dissolved heavy metal contents in water, and the concentrations of Sr exposed to carbonate rocks might be higher than that in other regions (Han & Liu, 2006; Han et al., 2019; Peng et al., 2012; Zeng et al., 2020).

Potential source identification

PC analysis

Principal component analysis is applied for nine heavy metals in three tributaries to identify their possible sources. The KMO value and Bartlett’s sphericity test are respectively 0.69 and zero, indicating that the data in this study is suitable for PCA. Three PCs extracted from the element were shown in Table 4 and Fig. 4, accounting for 74.74% of the total variance.

The PC 1 explained 34.58% of the total variance and had a high loading of Mn (0.96), Co (0.95), Al (0.79) and Ni (0.74). The concentrations of these four heavy metals occurred an increase in the middle reaches of YR, where most of the abandoned coal mining was located (Bai et al., 2007; Bangjiang et al., 2014; Pan et al., 2019). Hence, the PC 1 with high loadings on Mn, Co, Al and Ni might be dominantly influenced by acid coal mine drainage (Abraham, Dowling & Florentine, 2018; Ali et al., 2017; Santelli et al., 2010; Seo et al., 2017; Tiwary, 2001). The PC 2, accounting for 22.61% of the total variance and predominantly contained Ba (0.88), V (0.84) and Sb (0.64), where the concentrations matched well with the upper Xijiang River (Liu et al., 2017). Among three tributaries, Ba, V and Sb show higher concentrations in BR and JR than YR. Previous investigation and the scene observation found that BR and JR are influenced by human activities more than YR, with a higher percentage of residential and commercial land (Huang et al., 2009; Song et al., 2011; Zhang et al., 2019). Considering that Ba, V and Sb are widely used in industrial and medical (Filella, Belzile & Chen, 2002; Hope, 2008; Kravchenko et al., 2014; Wilson et al., 2010), these heavy metals could be attributed to anthropogenic sources. The PC 3 had strong loadings of Fe (0.85) and Sr (0.78), with a variance of 17.55%. Fe is a major element in Earth’s crust (Wang et al., 2017), therefore the high concentrations of Sr (much higher than the world average and consistent with the background value) may be originated from carbonate weathering (Liu et al., 2017). Therefore, PC 3 could be attributed to the sources of geologic origins.

WQI and health risk assessment

According to PCA, the weights of each heavy metal are shown in Table 5. The WQI values of each sampling site were calculated with Eq. (6) and shown in Fig. 5. The WQI values varied from 3.12 to 15.64, with an average of 5.79. YR-5 (WQI = 15.64) was the only site of which the WQI value exceeded 10, regarded as a relatively high site. Therefore, all the water samples could be defined as excellent water quality (WQI < 50), indicating that the natural water in tributaries was suitable for drinking with respect to heavy metals pollution.

According to Eqs. (2)–(6), the HQ and HI values of heavy metals for adults and children by ingestion and dermal pathways are calculated and summarized in Table 6. All HQ_ingestion, HQ_dermal and HI values are far smaller than 1, indicating these dissolved heavy metals in the Lake Aha watershed only make a little hazard for adults and children via ingestion and dermal absorption. Compared to dermal, ingestion is the primary pathway to endanger humans for these dissolved heavy metals. The HI value of Sb for children (0.154) is the only value exceeding 0.1. Consequently, Sb might be a potential non-carcinogenic risk to human health. It is worth noting that the HQ and HI values for children are higher than that for adults, suggesting that children are at higher risk than adults under the exposure of heavy metals.