Unveiling water quality and health risks from groundwater chemicals in Poyang Lake basin of China: a sophisticated analysis

EWQI assessment

Entropically Weighted Water Quality Index (EWQI) method is a technique for analyzing the distribution of groundwater quality in a study area. The approach involves determining the weight of water quality indicators through information entropy and condensing a large amount of water quality data into a representative numerical value that reflects the water quality condition (Nisar et al., 2024). Based on the calculated EWQI values, the groundwater quality in the study area can be divided into five levels from “very poor” to “excellent” as shown in Table 1. The calculation process for the Entropically Weighted Water Quality Index (EWQI) is intricate and involves five key steps: (1) constructing the initial water quality matrix (Eq. (1)); (2) standardizing the data (Eqs. (2) and (3)); (3) determining the weights (Eqs.(4) and (5)); (4) setting the quantification criteria for classification (Eq. (6)); and (5) calculating the EWQI itself (Eq. (7)).

where, X is the initial data matrix with m groundwater samples and n water–quality indicators, and x_ij is the original value of the jth indicator in the ith sample. max(x_j) and min(x_j) are the max and min values of the jth indicator. e_j isthe information entropy of the jth indicator, w_jthe entropy weight, with lower e_j meaning greater indicator impact. s_j is the standard allowable value for the jth indicator (except for pH) in the Chinese groundwater quality standard, in mg/L. c_ij is the standard allowable value for the jth indicator (except pH) from China’s groundwater quality standard, in mg/L. The weight of chemical characteristics for computing EWQI is shown as Table S1.

Irrigation risk assessment

Groundwater serves as a crucial source of irrigation water in the Poyang Lake basin during dry seasons or periods of extreme drought. Given the significant impact of irrigation water quality on crops, it is essential to evaluate the water quality from the perspective of irrigation feasibility. In this study, the quality of irrigation water is assessed using indicators such as the sodium adsorption ratio (SAR), soluble sodium percentage (SP), Kelley’s ratio (KR), permeability index (PI), and magnesium hazard ratio (MHR).

Sodium adsorption ratio: this index evaluates the propensity of sodium ions to accumulate in the soil and plants due to irrigation, which can affect soil structure and plant growth. The formula for calculating SAR is given by: (8)${\displaystyle \mathrm{SAR}=\frac{{\mathrm{Na}}^{+}}{\sqrt{\left({\mathrm{Ca}}^{2+}+{\mathrm{Mg}}^{2+}\right)/2}}.}$

Soluble sodium percentage (SSP): This percentage indicates the level of sodium ions that are soluble and potentially harmful to the soil. The calculation for SSP is as follows: (9)${\displaystyle \mathrm{SSP}=\left(\frac{{\mathrm{Na}}^{+}+{\mathrm{K}}^{+}}{{\mathrm{Na}}^{+}+{\mathrm{K}}^{+}+{\mathrm{Ca}}^{2+}+{\mathrm{Mg}}^{2+}}\right)\times 100.}$

Magnesium hazard ratio: MHR indicates the potential risk that magnesium ions pose to crops, which can be detrimental in high concentrations. The MHR is determined by: (10)${\displaystyle \mathrm{MHR}=\frac{{\mathrm{Mg}}^{2+}}{{\mathrm{Ca}}^{2+}+{\mathrm{Mg}}^{2+}}.}$

The permeability index (PI): an indicator that measures the proportion of sodium ions relative to calcium and magnesium ions in irrigation water to assess the potential impact of the water on soil structure and permeability. (11)${\displaystyle PI=\frac{{\mathrm{Na}}^{+}+\sqrt{{\mathrm{HCO}}_3^{-}}}{\left({\mathrm{Na}}^{+}+{\mathrm{Ca}}^{2+}{\mathrm{Mg}}^{2+}\right)}\times 100.}$

The Kelley ratio (KR): an indicator that measures the proportion of sodium ions relative to calcium and magnesium ions in irrigation water to assess the potential impact of the water on soil structure and permeability. (12)${\displaystyle \mathrm{KR}=\frac{{\mathrm{Na}}^{+}}{{\mathrm{Ca}}^{2+}+{\mathrm{Mg}}^{2+}}.}$

The assessment criteria of these indicators are shown in Table 1.

Deterministic model

Following the 2011 guidelines of the United States Environmental Protection Agency (USEPA), deterministic models are commonly used to assess potential health risks. These models assign specific values to parameters in the formulas, reflecting group characteristics, to evaluate the health impacts of individual groundwater samples. GIS technology, such as Kriging interpolation, is utilized to map the spatial distribution of these risks. Health risk assessments typically include evaluations of cancer risks (CR) and non-cancer hazard indices (NCHQ), with exposure mainly occurring through ingestion via drinking water and dermal contact during bathing. In this study, risks were assessed by calculating the chronic daily intake (CDI) for oral exposure and the dermal absorbed dose (DAD) for skin exposure, aggregating CR and NCHQ values from different exposure pathways for contaminants in the groundwater. The acceptable threshold for CR is set at 1 ×10⁻⁴, indicating an acceptable lifetime cancer risk below this level. Groundwater is considered a health risk and unsafe if the NCHQ exceeds 1. Detailed information for the computing process is shown as following: (13)${\displaystyle CDI=\frac{CW\times IR\times EF\times ED}{BW\times AT}}$ (14)${\displaystyle DAD=\frac{D{A}_{event}\times EV\times ED\times EF\times SA}{BW\times AT}}$ (15)${\displaystyle D{A}_{event}=K_P\times CW\times t_{event}\times 10^{-3}}$ (16)${\displaystyle NCH{Q}_{Oral-i}=CD{I}_i/Rf{D}_i}$ (17)${\displaystyle NCH{Q}_{Dermal-i}=DA{D}_i/Rf{D}_i}$ (18)${\displaystyle NCH{Q}_i=NCH{Q}_{Oral-i}+NCH{Q}_{Dermal-i}}$ (19)${\displaystyle NCH{Q}_{total}=\sum _{i=1}^{4}NCH{Q}_i=NCH{Q}_{F^{-}}+NCH{Q}_{N{O}_3^{-}}}$

where, CW represents the contaminant concentration in groundwater, expressed in mg/L. IR refers to the ingestion rate, in L/d, while EF signifies exposure frequency, in d/a. ED indicates the exposure duration in years to a specific chemical through drinking water consumption. BW is the average body weight in kg, and AT stands for the average exposure time in days. For dermal exposure, DAevent is the absorbed dose per event, in mg/cm². SA represents the skin area exposed to contaminants, measured in cm², and EV denotes the event frequency. K_p is the permeability coefficient for a substance through skin in water, given in cm/h. t_event refers to the duration of a single contact event in hours. RfD is the benchmark dose, in mg/kg d, with specific values for F⁻ and NO₃⁻ being 0.04 and 1.6 mg/kg d, respectively. The parameters for the calculations are listed in Table S2.

Uncertainty analysis

The Monte Carlo model is a computer simulation technique that utilizes random sampling and statistical methods to mimic the behavior and outcomes of complex systems. In health risk assessments, it accounts for uncertainties and variability in input parameters such as pollutant concentrations, ingestion rates, body weight, and exposure frequency. The process involves establishing probability distributions for these inputs, randomly selecting values that reflect the diversity of the population’s characteristics, and then conducting numerous iterations—like 10,000—to generate a distribution of risk outcomes. A key advantage of the Monte Carlo model is its ability to reflect the inherent variability and uncertainty in input parameters, offering a range of risk estimates rather than a single value. This makes it particularly effective for managing uncertainties stemming from sample distribution, medium heterogeneity, measurement errors, and human factors, thus providing a more comprehensive and realistic risk assessment. The fitted distribution of NO₃⁻-N, F⁻ and exposure variables are shown in Table S3.

Groundwater quality for drinking

The water quality of the study area is represented by calculating the groundwater EWQI (Fig. 3). The results show that the EWQI values of groundwater range from 15.11 to 945.26, with a mean value of 60.58, indicating that the overall water quality is acceptable, but there are areas where water quality has deteriorated. Within the study area, 26 sites have EWQI values between 100–150, and 30 sites have EWQI values greater than 150, indicating that the water quality at these sites has reached poor and very poor levels, respectively, with pollution being quite severe and not suitable for drinking water use. Additionally, polluted sites are widely distributed across various tributaries within the Poyang Lake basin (Fig. 4). Extensive human activities within the Poyang Lake basin, such as iron and coal mining and agricultural activities in the Ganjiang River basin, copper mining in the Rao River basin, large-scale agricultural activities in the upper reaches of the Xinjiang River, and widespread industrial production activities throughout the basin, may generate pollutants that, under the influence of abundant precipitation in the basin, can lead to erosion and infiltration (Li et al., 2016; Li et al., 2021; Pan, Cao & Zhang, 2020), thereby causing groundwater contamination.

We conducted random forest and sensitivity analyses on the EWQI and hydrochemical characteristics to determine which factors are more sensitive to EWQI (Fig. 4 and Fig. S2), thereby further identifying the sources of water pollution in the Poyang Lake basin. The random forest analysis results indicated that Mn and NH₄⁺ are the primary influencing factors of EWQI, followed by Fe, Na, and Al. This is highly consistent with existing results regarding the surface water quality of the Poyang Lake basin, which considers NH₄⁺ as one of the most significant influencing factors of water pollution in the region (Xu et al., 2022). From the aforementioned analysis, it is evident that erosion processes, industrial wastewater, and non-point source pollution are the main causes of Mn and other exceedances in the basin, potentially being the primary controlling factors for water pollution within the basin. However, the sensitivity analysis revealed that EWQI is most sensitive to changes in Al and COD, suggesting that although they are not the primary contributors to water quality changes, even minor fluctuations can significantly impact water quality in the study area and warrant attention.

Irrigation risk

Considering the extensive distribution of irrigation areas in the Poyang Lake basin, various irrigation risk indices such as SAR, KR, PS, PI, and MHR were employed to assess the quality of irrigation water, with the spatial distribution of these indices shown in Fig. 5. SAR serves as an indicator to determine the potential Na⁺ accumulation (or sodium damage) in soil and plants due to irrigation water. Except for one site in the upper reaches of the Xinjiang River, SAR values in other areas are less than 18, indicating minimal sodium damage in the study area. In terms of KR, MHR, and PI, “unsuitable” irrigation sites are primarily distributed in the Ganjiang River and its estuary (KR > 1, MHR > 50, PI > 120), followed by individual sites in the Fu River, Xinjiang River, and Rao River (MHR > 50, KR > 1), suggesting that groundwater irrigation at these sites may cause alkaline damage. Furthermore, the soluble sodium percentage (SSP) evaluation results reveal widespread areas unsuitable for irrigation within the Poyang Lake basin (SSP > 60). Integrating these evaluation indices indicates that there may be extensive areas unsuitable for irrigation in the study area, with the Ganjiang River basin being more pronounced, which is consistent with the significantly higher concentrations of Ca²⁺, K⁺, Mg²⁺, and Na⁺ in the Ganjiang River basin (Fig. S1). Therefore, measures should be taken to prevent alkaline damage that may be caused by groundwater irrigation in the Ganjiang River Basin, especially in the context of frequent extreme dry conditions where groundwater resources become more critical for agricultural development.

Deterministic health risk assessment

Numerous studies have analyzed the significant health risks posed by F⁻ and NO₃⁻ in water bodies to human health (Gao et al., 2022; Kaur, Rishi & Siddiqui, 2020; Padilla-Reyes et al., 2024; Shen et al., 2022). This study employs the deterministic non-carcinogenic health risk assessment model recommended by the United States Environmental Protection Agency (USEPA) to evaluate the risks of F⁻ and NO₃⁻ in groundwater (USEPA, 2011). The model takes into account direct ingestion and dermal contact as exposure pathways, with relevant model parameters presented in Table S1. Figure S3 displays the results of the deterministic risk assessment. We found that different populations face varying risks. For children, the NCH_total ranges from 0.64 to 3.37, with a mean of 1.04; for adults, the NCH_total ranges from 0.27 to 1.45, with a mean of 0.44. Additionally, different contaminants have varying impacts on non-carcinogenic health risks to humans. For children, the NCH_F− ranges from 0.58 to 1.37, with a mean of 0.78, which is significantly higher than for adults (NCH_F−, 0.24–0.59; mean: 0.34). Similarly, for children, the NCH_NO3− ranges from 0.008 to 2.62, with a mean of 0.25, also higher than for adults (NCH_NO3−: 0.004–1.13; mean, 0.11). This indicates that groundwater in the Poyang Lake basin poses a significant non-carcinogenic health risk to humans, with a higher risk to children compared to adults, and F⁻ poses a greater threat than NO₃⁻. This is consistent with existing research, which suggests that children are more threatened by contact with groundwater, and F⁻ poses a greater health risk than NO₃⁻ to human health (Kaur, Rishi & Siddiqui, 2020; Qiu et al., 2023b; Shen et al., 2022; Yang et al., 2022).

Furthermore, the impact of various contaminants on human health varies. As shown in Fig. S3, 34 groundwater samples exceed the non-carcinogenic risk threshold for children’s fluoride exposure (HI = 1), and no samples exceed the non-carcinogenic risk threshold for adults. For nitrate, there are 17 and 2 groundwater sites exceeding the acceptable risk levels for children and adults, respectively. This indicates that F⁻ and NO₃⁻ in groundwater pose non-carcinogenic health risks to humans only within certain regional scopes, with F⁻ presenting a higher risk and broader impact, and the health risks posed by F⁻ in the future should not be overlooked. Additionally, different exposure pathways have varying impacts on health risks. Figure S4 shows that the non-carcinogenic health risk from ingestion is much higher than that from dermal contact, a result consistent with the study by Qiu et al. (2023a). Therefore, greater attention should be paid to the quality of drinking water, and ensuring the safety of drinking water by controlling the main exceedance contaminants in it is essential.

Uncertainty analysis

Numerous studies have utilized Monte Carlo models to simulate health risks (Gao et al., 2022; Padilla-Reyes et al., 2024; Qiu et al., 2023b). Probabilistic modeling can clearly display the probabilities of various simulation outcomes and provide more comprehensive analysis results under conditions of limited data resources. This study employed the Monte Carlo method for probabilistic health risk assessment, with results shown in Fig. 6 and Fig. S4. The results indicate that the health risk probabilities for both populations generally follow a log-normal distribution. At the 95% confidence level, the range of NCH_total for children is 0.27–1.10, with a mean of 0.51; for adults, the range is 0.08–0.29, with a mean of 0.15. The probability of children’s NCH_total exceeding the threshold (NCH = 1) is approximately 2%, while for adults, it is 0%. This suggests that the groundwater in the study area poses no health risk to adults, and although there is a certain health risk to children, the probability of risk occurrence is not high.

The health risk probabilities for different contaminants vary. At the 95% confidence level, the range of NCH_F− for children is 0.23–0.62, with an average of 0.38; for adults, it is 0.06–0.18, with an average of 0.11. Concurrently, the range of NCH_NO3− for children is 0.01–0.54, with an average of 0.13; for adults, it is 0–0.15, with an average of 0.04. Additionally, the probabilities of NCH_F− and NCH_NO3− exceeding the threshold (NCH = 1) for both children and adults are 0%. This indicates that the probabilities of groundwater contaminants F⁻ and NO₃⁻ posing non-carcinogenic health risks to humans are minimal.

The risk probability simulations for different exposure pathways of groundwater contaminants show that, at the 95% confidence level, the health risk range for children and adults through oral intake is 0.27-1 and 0.08–0.29, respectively, with average values of 0.51 and 0.15, respectively. The probabilities of health risks through oral intake exceeding the threshold (NCH = 1) for children and adults are approximately 2% and 1%, respectively. The health risks through dermal exposure are nearly 0. This demonstrates that, although the probabilities of non-carcinogenic health risks from different exposure pathways are relatively small, oral intake is the primary pathway through which groundwater in the Poyang Lake basin affects human health, with a greater impact on children’s non-carcinogenic health risks. Previous studies also indicated the the higher health risk of water pollutants through oral intake, especially for children (Yang et al., 2022).

By comparison, it can be observed that the health risk assessment results for different contaminants obtained from deterministic analysis are to some extent confirmed in probabilistic simulation results. However, it is also evident that deterministic analysis overestimates the potential health risk impact of F⁻ and NO₃⁻ in groundwater to some extent. For instance, probabilistic simulation results show that the probabilities of F⁻ and NO₃⁻ posing non-carcinogenic health risks to humans are virtually 0, while deterministic analysis indicates that F- poses non-carcinogenic health risks at multiple sites across the basin. The overestimation by deterministic analysis may be attributed to its inability to account for the inherent variability and uncertainty in input parameters, which can lead to either over conservative or under conservative risk estimates (Gao et al., 2022; Padilla-Reyes et al., 2024). In contrast, probabilistic methods like Monte Carlo simulation provide a more nuanced understanding by incorporating the uncertainty distributions of contaminants’ concentrations and exposure factors (Asefa et al., 2024). This nuanced approach is particularly valuable in regions like the Poyang Lake basin, where groundwater quality can be influenced by a multitude of variable factors (Garg, Yadav & Aswal, 2019), including agricultural practices, natural geological conditions, and hydrological dynamics. Therefore, our findings underscore the importance of integrating probabilistic methods into health risk assessment frameworks for groundwater contaminants. They offer decision-makers a more reliable basis for prioritizing risk mitigation strategies and allocating resources effectively. Moreover, the minimal risk probabilities for F⁻ and NO₃⁻ suggest that while these contaminants may not be the primary focus for stringent regulatory actions, continued monitoring and targeted interventions in high-risk areas (particularly those affecting children) are still warranted to ensure long-term public health protection.

Sensitivity analysis

A sensitivity analysis was undertaken to evaluate the role of predictor variables in assessing non-carcinogenic health risks. The analysis was designed to pinpoint the key predictor variables that have a significant impact on the health risk assessment outcomes (Fig. S5 and Fig. 7). The sensitivity analysis indicates that for children and adults, the non-carcinogenic health risks from dermal and oral exposure to groundwater are sensitive to the factors in the order: NO₃⁻ > EF > F- > BW > IR. NCH_NO3− is most sensitive to NO₃⁻ concentration. For NCH_F−, the sensitivity order is EF/F- > BW > IR. NCH correlates positively with EF and pollutant concentration, and negatively with BW. High pollutant concentrations increase risk, and lighter individuals drinking contaminated groundwater frequently face greater non-carcinogenic health risks. These results align with Gao et al. (2022), confirming that groundwater poses a higher non-carcinogenic health risk to children than adults. Therefore, controlling groundwater non-carcinogenic health risks should be tailored to local pollutant concentrations, with additional preventive measures for children, who are more vulnerable.