Distribution of Natural Trace Elements in Drinking Water Sources of Hungary

1. Introduction

Risk based approach to drinking water safety is taken up world-wide to complement endpoint monitoring (World Health Organization 2022). Monitoring at the tap is still seen as an indispensable element of drinking water control but shifts from being the primary means of safeguarding drinking water quality towards becoming a tool for validating water safety plans (or similar frameworks). The recent recast of the European Union (EU) 2020/2184 Drinking Water Directive (DWD) introduced for the first time the obligation of risk assessment for the entire water supply chain [1]. As drinking water quality is strongly influenced by the water source, better understanding of source water quality and the associated hazards is a cornerstone of the risk-based approach [2]. Hazard assessment of the catchments of drinking water abstraction points is also a requirement under the DWD [1].

Drinking water quality regulations addressing parameters derived from the source water usually focus on potential anthropogenic pollutants and a few geological compounds of known health impact, such as arsenic. Monitoring requirements under the former EU regulation in the European Union followed the same approach, focusing mainly on heavy metals associated primarily with human pollution (e.g. mercury, lead, cadmium, chromium, nickel) [3]. The number of geogenic pollutants for which substantial monitoring data is available in the EU is limited, comprising generally of arsenic, selenium, boron, and fluoride, though national legislations may include additional parameters. The recast DWD extended monitoring requirements (starting from 2026) to new components, including metals of health relevance. Uranium was added with a parametric value of 30 µg/L following the recommendation of the World Health Organisation (WHO) [4]. Calcium, potassium and magnesium were also included in the list of parameters to be monitored, but only for consumer information, and no parametric values have been assigned [1].

The newly introduced natural elements have a confirmed or presumptive impact on human health. Uranium is a nephrotoxic pollutant of primarily geological origin [4,5]. Its concentration in natural waters is usually low (<1.0 µg/L), but depending on the geological framework (e.g. granite rich in uranium) and the mobilizing capacity of groundwater flow systems high concentrations can also occur in certain areas (up to 57 µg/L in Mongolia, 85 µg/L in Italy, 288 µg/L in China and 750 µg/L in Norway) [6,7,8,9,10].

Potassium, calcium and magnesium are essential elements. Potassium concentrations also vary widely in natural waters, but the concentrations present in drinking water are generally not a risk to human health and may be of significance only for sensitive individuals (e.g. those suffering from renal failure) [4].

The inverse association between magnesium and calcium in drinking water and cardiovascular mortality is well established. The protective effect is clearer for magnesium, while further studies are needed for calcium [11,12,13,14]. Low magnesium and calcium intake via drinking water have also been associated with high blood pressure, metabolic syndrome, preeclampsia and amyotrophic sclerosis [15]. For these elements, prior data is scarce, though many jurisdictions, including the Hungarian, require monitoring of total hardness, which primarily consists of magnesium and calcium salts.

Vanadium is not regulated by the DWD, but some EU Member States include it in national legislation, e.g. Italy, with a parametric value of 140 µg/L [10]. The health effects of vanadium are unclear, but therapeutic and toxic effects on human health are possible (e.g. bone, cardiovascular, neurological, immune system, and body weight effects) [16,17]. Its concentration varies in a wide range in surface waters (0.010 µg/L to 68 µg/L), reaching even higher levels in groundwaters, especially in volcanic areas (up to 350 µg/l) [10,18].

Lithium is also an unregulated element of health relevance. The concentration of lithium in drinking water in European countries ranged from <1 µg/L to 20–60 µg/L in northern England, Lithuania and Italy [19,20,21], over 100 µg/L in Greece [22], and more than 1000 µg/L in Austria [23]. Several studies have found a protective association between lithium in drinking water and suicide mortality [20,23,24,25,26,27].

Molybdenum and selenium are essential elements. The concentration of Mo in drinking water is typically low (<10 µg/L), although near mining areas, this value can be much higher (>200 µg/L). The recommended health-based guideline value of WHO is 70 µg/L. The optimal range of Se intake is narrow. Selenium deficiency has been associated with cardiomyopathy and joint disease, while symptoms of selenosis include brittle hair and hair loss, nail malformation and diarrhea [28]. The parametric value of Se has been increased from 10 to 20 µg/L during the recast, based on recent scientific evidence [1,29]. WHO guideline value is 40 µg/L. Environmental concentrations are generally well below 10 µg/L, except in a few seleniferous areas.

The health impact of barium is unclear. There is no evidence of its carcinogenic and mutagenic properties, but there are contradictory results for cardiovascular toxicity (hypertension) and it has been shown to cause nephropathy in animal studies. The WHO recommends a guideline value of 1300 µg/L in drinking water [4,17].

In Hungary, the primary source of drinking water is groundwater (93% of total production volume), including 31% bank filtered water and a smaller proportion of karst water (12%). Surface water accounts for only 7% of the water supply [30]. There are 1403 public utility water supply systems registered in the National Drinking Water Database, operated by 38 water suppliers. Of these systems, 36 supplies use mainly bank filtration, 116 karst water and 1201 other groundwater sources, while 22 rely on surface water abstraction.

Drinking water quality is regularly monitored by the water suppliers and the local public health authorities. Monitoring does not cover elements of natural origin beyond the EU DWD requirements, although total hardness – comprising primarily of calcium and magnesium – is regulated as an indicator parameter (acceptable range: 50–350 mg/L CaO equivalent). Drinking water in Hungary is typically moderately hard (between 100 and 200 mg CaO/L) or hard (>200 mg CaO/L), with total hardness ranging from <1.0 to 393 mg CaO/L [30]. For other unregulated elements, only local studies are available. Previous targeted investigations detected uranium in concentrations up to 25 µg/L in some drinking water sources in Hungary, though not exceeding the new parametric value [31,32,33,34].

The aim of the present study was to carry out a nationwide survey mapping the presence and distribution of (previously) unregulated elements of natural origin in Hungarian drinking water sources. We aimed to identify if the observed distribution of the elements was associated with the water source type or the geographical location of the water supply, reflecting differences in (hydro)geological factors. It was evaluated whether any of these elements poses a potential hazard for drinking water production and requires further attention in risk assessment (water safety plans) on a national or local scale. The survey outcomes also provide representative data for epidemiological studies assessing potential health risks or gains associated with these compounds in drinking water.

2. Materials and Methods

2.1. Study Design

The survey was initiated by the National Center for Public and Pharmacy (NCPHP) as the national regulatory authority. All public utility drinking water suppliers were invited to participate. Participation was voluntary. Water suppliers responding positively received sample containers, instructions for sampling, and a data sheet to be completed during sampling. The sampling data sheet recorded the date and exact location of sampling, the type of raw water (surface, deep groundwater (second or deeper aquifer or depth >50 m), shallow groundwater (from the first aquifer or depth <50 m), karst, bank filtered, other/mixed) and information on water treatment and disinfection.

2.2. Sampling

Water samples were collected between March 2018 and June 2022 by the drinking water supply operators. At least one source water sample was taken for each water supply. Water suppliers were instructed to take a sample representative of the source water composition before water treatment. From water supplies with multiple abstraction points, either a representative subset of the wells or mixed well water representing normal operation were sampled. Samples from water supplies using surface water abstraction were collected after abstraction and mechanical filtration. Samples from water supplies using bank filtration were taken from selected bank filtration wells or their mixed water. Most water supplies were sampled twice in the sampling period. Samples were collected in 50 ml PP centrifuge tubes, and the pH of the samples was reduced below 2 with high-purity nitric acid for preservation (ISO 5667-11:2009). Samples were transported to the laboratory of the NCPHP and stored at room temperature until analysis within 30 days.

2.3. Chemical Analysis

The analytical tests were performed at the NCPHP’s laboratory. Elements were analyzed by ICP-MS ((iCAP RQ, Thermo Fisher Scientific, Germany) (ISO 17294-2:2016). The concentrations of the following elements were measured: boron (B), barium (Ba), beryllium (Be), calcium (Ca), cobalt (Co), potassium (K), lithium (Li), magnesium (Mg), molybdenum (Mo), sodium (Na), selenium (Se), strontium (Sr), titanium (Ti), uranium (U) and vanadium (V).

2.4. Statistical and Geospatial Methods

Statistical evaluation of the results was carried out using the Statistica program (TIBCO Software Inc. (2020). Data Science Workbench, version 14. http://tibco.com.). Values under the lower limit of quantification (LOQ) were replaced by LOQ/√2 in the statistical calculations [35].

Hierarchical cluster analysis (HCA) was used to group the samples based on their similar characteristics [36]. In HCA, each sampling point is initially a separate cluster, and then in iterative steps, the two closest clusters are merged into a single cluster. The number of clusters is thus reduced in each step, eventually leaving a single cluster. Ward’s method and quadratic Euclidean distance were used for merging the clusters. Discriminant analysis (DA) was used to test the appropriateness of the clusters formed by HCA.

Principal component analysis (PCA) was used to reveal the basic characteristics and hidden relationships of the large amount of data and for data reduction [37].

For multivariate statistical analyses (both PCA and HCA), the results were standardised by z-score transformation to reduce the bias resulting from the often orders of magnitude differences in the concentration of the elements.

Elemental concentrations of the different clusters were compared using the non-parametric Kruskal-Wallis H-test (p-value: 0.05), and the correlation between elements was tested using Pearson's correlation test (p-value: 0.05).

Maps were created using ArcGIS Desktop 10.8. program (Ersi Inc., version 10.8.0.12790, https://www.esri.com/en-us/home).

3. Results

3.1. Prevalence and Distribution of Natural Elements

Eighty-seven percent of public utility drinking water suppliers (33 of 38) participated in the survey. Altogether, 1256 samples were collected from 1155 water supply systems, representing 82% of Hungarian drinking water supplies (Figure 1). Of the samples, 84% were deep groundwater, 8.8% karst water, 2.5% bank filtered water, 2.3% shallow groundwater, 1.4% surface water, and <1% mixed water.

The results of the elemental analysis are summarized in Table 1. The concentration of Be did not exceed LOQ in any of the samples, and Co concentrations were above LOQ only in 1% (13 samples). Therefore, these two elements were not included in further analyses. Se and V were rarely present in Hungarian drinking water sources, with more than 85% of samples below the LOQ. Median concentrations of U and Mo were below LOQ for both elements, and even the upper quartile values were only 126 and 164% of LOQ, respectively. The concentration of U exceeded the EU parametric value of 30 µg/L in a single sample with a maximum concentration of 41 µg/L. Ti, B, Li and Ba were detected in the majority of the samples with concentrations ranging over several orders of magnitude from below LOQ to 83, 2570, 265, and 551 µg/L, respectively). K and Na were present in almost all samples (21 and 1 <LOQ samples, max 32 and 252 mg/L, respectively), and Mg, Ca and Sr were above LOQ in every measurement (min-max Mg: 0.55-100 mg/L; Ca: 2.5-174 mg/L; Sr: 21-3310 µg/L). A moderately strong positive correlation (r=0.402-0.556) was observed between Li and K; B and Na; Mg and Ca; Mg and Sr; Se and U; and a moderately strong negative correlation (r=–0.579) between Na and Ca (Table 2).

3.2. Water Supply System Clusters by Water Composition

Based on the HCA, 6 clusters were identified, also confirmed by DA (Wilks' lambda: 0.00363). The dendrogram is shown in Figure S1 in the Electronic Supplementary Material (ESM). Clusters 3 and 4 were the largest, together comprising two-thirds of the water resources (488 and 343 water resources, respectively). Cluster 2 and 5 accounted for 12 and 16% (151 and 200 water resources), respectively. Cluster 1 and 5 were small, with 31 and 43 water sources (2.5 and 3.4%, respectively). In Clusters 1, 2, 4 and 5, 85-99% of the samples are from deep groundwater aquifers (Table 3). Cluster 3 was the most heterogeneous, with deep groundwater comprising 68%, karst water 21%, shallow groundwater and bank filtration 4-4%. This means that cluster 3 includes the vast majority of all karst, shallow groundwater, bank filtered and surface water samples (91, 69, 63 and 67%, respectively). In Cluster 6, deep groundwater samples accounted for 63% of the samples, with shallow groundwater, surface and bank filtered samples comprising 14%, 12% and 9.3%, respectively. All surface water samples in cluster 6 are from Lake Balaton.

Li concentrations were highest in Cluster 1 (median 139 µg/L). Concentrations were typically an order of magnitude lower in the other clusters, with the lowest values in Clusters 2 and 3 (median 9.6 and 7.5 µg/L, respectively) (Figure 2a). Clusters 1 and 5 were characterized by high B, Na and Mo concentrations (median 198 and 154 µg/L; 73 and 102 mg/L; 1.9 and 2.4 µg/L, respectively), while significantly lower levels were measured in the other clusters, with the lowest in Cluster 3 (median B 15 µg/L; Na 15 mg/L; Mo <LOQ) (Figure 2b-d). Mg concentrations were the highest in Clusters 6 and 2 (median 36 and 35 mg/L), moderately high in Clusters 3 and 4 (median 28 and 22 mg/L), and low in Clusters 1 and 5 (median 12 and 8.9 mg/L). K was most abundant in Clusters 1 and 6 (median 3.9 and 2.5 mg/L) and typically low in the other clusters (median between 1.2 and 1.6 mg/L) but more variable in Clusters 2 and 3. Ca concentrations were the highest in Clusters 2 and 3 (median 81 and 79 mg/L), medium-high in Clusters 1, 4 and 6 (median 40; 54 and 63 mg/L), and low in Cluster 5 (median 21 mg/L). Cluster 2 had elevated levels of Ti (median 41 µg/L) compared to the other clusters (median 1.0-3.0 µg/L). Median concentrations of V and Se exceeded the LOQ only in Cluster 6 (median 2.0 and 1.1 µg/L, respectively). All 6 clusters had significant Sr concentrations (median 239-477 µg/L). The order of the clusters in decreasing concentration from the highest to the lowest are Cluster 4, Cluster 6, Cluster 2, Cluster 3, Cluster 1, Cluster 5. The concentration of Ba was the highest in Cluster 4 (median 195 µg/L) and significantly lower in the others, being lowest in clusters 3 and 6 (median 61 and 65 µg/L). Median U concentrations were below LOQ in most clusters, except in cluster 6. (median 6.4 µg/L).

3.3. Spatial Distribution

The geographical distribution of the 6 clusters is shown in Figure 3.

Geographically, Cluster 1 is a distinct group located in the northern part of Hungary. Drinking water sources belong to Cluster 2 form recognisable clusters in the southwest, south and central parts of the country. Cluster 3 is dispersed across the country, although it is more prevalent in the western and central parts of the country than in the eastern region. This cluster includes the vast majority of the non-deep groundwater sources (karst, bank filtered, shallow groundwater and surface water). Drinking water sources classified as Cluster 4 are scattered throughout the country. Cluster 5 drinking water sources are localised in the east-southeast part of the country and a specific area in the western half of the country. Cluster 6 is mainly associated with the southern foothills of Transdanubian and Northern Ranges. Characteristics of clusters are summarized in Table 3.

3.4. Principal Component Analysis

PCA analysis identified 5 principal components (PCs) with eigenvalues greater than 1. These 5 PCs explain 65% of the total variance. Characteristic elements (loading values <-0,4 or >0,4) of PC1 are B, Ca, Mg, Mo and Na (total variance 21.7%) and K, Li and Sr in case of PC2 (total variance 14.5%). Loading and eigenvalues of PCs are presented in Table S1 and S2 of the ESM. Samples belonging to the 6 clusters were also grouped together in the PCA, and the clusters were separated by the first two PCs (Figure 3a). Comparing water types, only karst waters formed a distinct group according to the first two PCs (Figure 3b). The overwhelming dominance of groundwater samples masked any other separation by water type.

5. Conclusions

The study confirmed that none of the investigated parameters pose a risk to drinking water quality and consequently to the health of the consumers on a national scale. However, local hydrogeological conditions may lead to emergence of natural elements in higher concentration in some water supplies. Such local exceedances e.g. of U or V require further investigation and if needed, risk mitigation measures.

In hydrogeological studies, usually the major elements (Ca, Mg, Na, K, Cl, SO₄, HCO₃) are used for characterization of composition of waters. The presented study highlighted the importance of trace elements and the multivariate data analysis tools, which can reveal links and explain geographical distributions. Nevertheless, involving more data into the analysis (well depth, lithology, flow system evaluation results) would result in more detailed picture about the origin of these elements.