Daily Fluctuations in the Isotope and Elemental Composition of Tap Water in Ljubljana, Slovenia

: The isotope and elemental composition of tap water reﬂects its multiple distinct inputs and provides a link between infrastructure and the environment over a range of scales. For example, on a local scale, they can be helpful in understanding the geological, hydrogeological, and hydrological conditions and monitor the proper functioning of the water supply system (WSS). However, despite this, studies examining the urban water system remain limited. This study sought to address this knowledge gap by performing a 24 h multiparameter analysis of tap water extracted from a region where the mixing of groundwater between two recharge areas occurs. This work included measurements of temperature and electrical conductivity, as well as pH, δ 2 H, δ 18 O, d , δ 13 C DIC , and 87 Sr/ 86 Sr ratios and major and trace elements at hourly intervals over a 24 h period. Although the data show only slight variations in the measured parameters, four groups were distinguishable using visual grouping, and multivariate analysis (Spearman correlation coefﬁcient analysis, hierarchical cluster analysis, and principal components analysis). Finally, changes in the mixing ratios of the two sources were estimated using a linear mixing model. The results conﬁrm that the relative contribution from each source varied considerably over 24 h.


Introduction
Increasing demand for drinking water and the complexity and heterogeneity of the urban water supply system (WSS), often with its fragmented and ageing infrastructure [1,2], represent a significant challenge for water supply managers as the system is susceptible to contamination or physical interruption [3]. Usually, the physical structure and the necessary information about the WSS are known but in large systems or in developing countries, where this information is either missing or difficult to obtain, methods are needed to study complex WSS without having an in-depth knowledge of the physical infrastructure [4].
Stable water isotope analysis has proven to be a useful tool in understanding hydrological processes [4][5][6]. This is because stable isotope ratios of hydrogen ( 2 H/ 1 H) and oxygen ( 18 O/ 16 O) in water provide a characteristic signature that can be used to investigate the origin of different water sources that contribute to the stream, natural, and artificial mixing of waters, sources of the groundwater recharge and to quantify the variability of climate change [4,6,7].
The isotopic composition of dissolved inorganic carbon (δ 13 C DIC ), the main species in water draining carbonate aquifers, is also helpful for assessing the origin of water. The primary processes that affect the carbon isotope composition are the dissolution of carbonates, the microbial decomposition of organic matter and its removal via carbonate precipitation [8]. The δ 13 C DIC is indicative of the biogeochemical processes within the aquifer, especially in carbonate-rock aquifers [9][10][11]. Other than these three studies, little information was available at the outset of this investigation relating to the δ 13 C DIC of tap water. Similarly, the 87 Sr/ 86 Sr can connect local geology to a specific recharge area and provide a better understanding of the processes that impact waters affected by human management [12,13]. Knowing the concentration of the minor and major elements in water can also provide complementary information to the isotope composition [14].
More recently, water isotopes have been used to understand the origin of tap water, [6,15,16]. Unlike surface water with an isotopic composition similar to regional precipitation [6,15], the isotope information provided by tap water is not so easy to interpret, as it can represent contributions from different water sources and different static and dynamic regions, e.g., regions supplied predominantly by one source compared with regions experiencing active mixing between various sources [4,16]. Differences in water isotopes can be used to investigate different inputs into the system (i.e., lakes, precipitation, surface water) as different processes affect the isotope composition (seasonal changes in recharge, contribution of meteoric waters, temperature, and humidity) [15]. Elemental signatures related to the geological (lithology), hydrogeological and hydrological background can also be used to distinguish specific aquifers. Although it is essential to know the elemental composition of the different water sources, it is also necessary to consider the contribution made by elements leaching from pipe scale, e.g., Ca, S, Mn, Zn, P, Mg, Al, and Fe. Human activities can also alter the water composition (i.e., accidental spills, supply contamination) [4]. However, care is needed in any interpretation of elemental composition since urban water is subject to contamination by natural water supplies, water losses from the supply system, prolong retention times, and dead-end areas [17].
In urban areas, conducting field studies remains challenging as constraining the water fluxes requires the monitoring of vast areas. Therefore, simultaneous field studies are also needed covering the heterogenic and complex WSS where the researchers or the water system managers perform monitoring and sampling over spatiotemporal scales. In addition, there is a need to understand the water transport dynamics in the WSS and establish a spatially distributed multiparameter data set for a defined period.
In Ljubljana, Slovenia's capital city (population approximately 330,000), the primary source of drinking water is groundwater. The contribution that surface water and precipitation makes to groundwater changes through the hydrological year [18]. Within the area, some contaminations were defined: hexavalent chromium plumes, nitrate and new emerging pollutants, and desethyl-atrazine plumes [19]. In the city, water consumption varies daily; however, the time it takes for water from the wellfield to reach the city centre is no more than few hours (personal communication, VOKA SNAGA manager). Regular monitoring of drinking water quality in the WSS, based on demands of national regulations [20] that are harmonized with of European Drinking Water Directive [21], does require analysis of elements related to the pollution of drinking water (i.e., Cd, Cr, Fe, Pb, Mn) as high concentrations of toxic metals in water pose a risk to health [22]. The management of the groundwater quality at the functional urban area and the feasible measures for decreasing of the concentration of relevant contaminants were discussed recently [23,24].
Until 2018, no studies had looked at the Ljubljana WSS or attempted to explain possible isotopic and elemental compositional changes in water during its journey from "source to tap" [25]. Then, in 2018, the first multi-tracer investigation of urban water from different sources was performed by Vreča et al. [26,27], revealing that certain elements, e.g., As, B, Li, and Sr, are characteristic of a specific source or recharge area. Surprisingly, only Du et al. [16] report the hourly fluctuations in water isotopes, but no one has investigated changes in δ 13 C DIC or changes in the elemental composition coupled with stable isotopes (δ 13 C DIC , δ 18 O, δ 2 H) in tap water.
This study's overarching goal was to use a multi-tracer approach to investigate the daily variability in the isotope and elemental composition of tap water. It is the first study of its kind in Slovenia looking specifically at water isotopes and building on our 2018 study.
The study also gave the opportunity to collect samples of tap water from the mixing of two aquifers with different recharge areas and geochemical facies. The mixing of water of different origin at this location is known to the water managers, who are interested in information on drinking water origins, for various reasons. The more information they have, the safer the water supply they are able to implement. This is important for when they have to control planned changes in the water distribution net, review the consequences of unintended changes, and take measures to mitigate any adverse effects. The main aims of this study were to characterize the daily geochemical variability of tap water over time and identify geochemical tracers to estimate the mixing ratio of water sources at the selected location.

Site Description
Hourly samples of tap water were collected from the main building's basement at Jožef Stefan Institute (JSI in Figure 1), Ljubljana, Slovenia (lat: 46.04207, long: 14.487400). The water originates from two different wellfields: Kleče and Brest (Figure 1), located in aquifers with different hydrogeological characteristics. Kleče is located in the Ljubljansko polje aquifer and Brest on the Ljubljansko barje aquifer. Two rivers bind the Ljubljansko polje: the River Ljubljanica to the south and the River Sava to the north in the eastern part of the Ljubljana basin. The basin was formed by tectonic subsidence in the early Pleistocene and is composed of Permian and Carboniferous slate claystone and sandstone [28]. The Pleistocene and Holocene fluvial sediments, accumulated by the River Sava, form highly permeable partially cemented sand and gravel with lenses of conglomerate [29]. The aquifer is recharged by both precipitation and the River Sava, mainly in the north-western part. It is also recharged via lateral inflow from the Ljubljansko barje multi-aquifer system in the south [30,31]. The depression of the Ljubljansko barje is located on the southern part of the Ljubljansko polje; it formed within the permeable limestone and dolomite basement and was filled by alluvial, marshy, and lacustrine sediments during the Pleistocene and Holocene. The upper Holocene aquifers are recharged directly from precipitation and surface streams, while the lower aquifer is recharged from the karst recharge area [32]. Sediments in this area are heterogeneous, and the hydrogeological conditions are more complex than on the Ljubljansko polje [33].
The Ljubljana WSS consists of water supply facilities, 44 wells, eight minor pumping stations, and more than 1100 km of supply network [34]. Groundwater is exploited at the Ljubljansko polje from the Kleče, Hrastje, Jarški prod and Šentvid wellfields and at Ljubljansko barje from the Brest wellfield ( Figure 1). In the central system, some settlements are continuously supplied with drinking water from a single wellfield (A, C, D, and E), while others are supplied from two or more wellfields (F, G, H, and I2) [33].

In-Situ Measurements and Water Sampling
Monitoring of tap water was performed from 09:00 on 24 April 2019 until 09:00 on 25 April 2019. In-situ temperature (T), electrical conductivity (EC), and pH were measured every full hour, using a Hanna HI 9829 Multiparameter instrument (Woonsocket, RI, USA), with an accuracy of ±0.15 • C, ±1 µS/cm and ±0.02, respectively. Quick calibration of the instrument was performed after 4 h, at 13:10.
Before collecting the first sample, the tap water was allowed to run for 60 s. Samples were then collected every hour. In total, 25 water samples were collected. The samples for δ 2 H and δ 18 O analysis were stored in prewashed 30 mL high-density polyethylene (HDPE) bottles and stored at room temperature. Samples for δ 13 C DIC were filtered on-site through a 0.45 µm nylon filter into 12 mL glass exetainers. In contrast, samples for determining the multi-elemental composition and 87 Sr/ 86 Sr ratios were collected in prewashed 50 mL polypropylene (PP) centrifuge tubes and acidified using HNO 3 (68% v/v, suprapur, Carlo Erba Reagents, Val de Reuil, France). All samples were stored at 4-6 • C.

Analytical Procedures
The δ 2 H, δ 18  Jaklitsch). The water bath was set to 18 • C. The water vapor trap was cooled to -55 • C. H 2 (IAEA) and CO 2 (Messer 4.5) gases were used as working standards. Samples (3 mL) were allowed to equilibrate for 2 (H 2 -H 2 O) and 6 (CO 2 -H 2 O) hours before analysis.
All measurements were performed together with laboratory reference materials (LRM) calibrated periodically against primary IAEA calibration standards to the VSMOW/SLAP scale. The defined isotope values and measurement uncertainty of LRMs were used to normalize the data, and independent quality control was calculated using the Kragten method [39][40][41]. All samples were measured in duplicate. The results were normalized to VSMOW/SLAP using LIMS (Laboratory Information Management System for Light Stable Isotopes) program and expressed in the standard δ notation (in ‰) using the conventional delta notation:  [43] and [44] methods. Saturated phosphoric acid (100%) was added (100-200 µL) to a septum-sealed tube and purged with pure He. A water sample (1 mL) was then injected into the tube, and the isotope composition of CO 2 measured directly from the headspace using a continuous flow IsoPrime100 stable isotope mass spectrometer (CF IRMS) coupled with the MultiFlow Bio equilibration unit. The results were normalized to VPDB and expressed in the standard δ notation in ‰ (see Section 2.3.1). A standard solution of Na 2 CO 3 (Carlo Erba reagents, Val de Reuil, France) with a known δ 13 C DIC of -10.8 ± 0.2‰ was used to determine the optimal extraction procedure for tap water samples. The average sample repeatability was 0.1‰.

Determination of Major and Trace Elements
Four major (Ca, Na, K, Mg) and 23 trace elements concentrations (Ag, Al, As, B, Ba, Cd, Co, Cr, Cu, Fe, K, Li, Mn, Mo, Ni, Pb, Rb, Sb, Se, Sr, U, V, and Zn) were determined using an Agilent 7900x inductively coupled plasma mass spectrometer (ICP-MS, Agilent Technologies, Tokyo, Japan). To measure accuracy, two surface water reference materials: SLRS-5 (National Research Council Canada, Ottawa, ON, Canada) and SPS-SW1 (Spectrapure Standards, Manglerud, Norway), were analysed at the beginning, in the middle and at the end of the sequence. Recovery ranged from 97% to 102% for all elements, and the repeatability was better than 5%.

Determination of 87 Sr/ 86 Sr Isotope Ratios
Five samples were selected for 87 Sr/ 86 Sr isotope ratio analysis according to the Sr and Rb concentrations and their ratio (10:00, 17:00, 00:00, 03:00 and 08:00). The 87 Sr/ 86 Sr isotope ratio was determined using the method described in Zuliani et al. [12]. Briefly, water samples (from 0.1 to 1 mL) were evaporated to dryness and redissolved in 1 ml of 8M HNO 3 . For Rb/Sr separation, Sr specific resin (Eichrom ® , Triskem International, Bruz, France) was used. The 87 Sr/ 86 Sr isotope ratio was determined using a Nu plasma II multi-collector ICP-MS (Nu Instruments Ltd., Wrexham, UK) fitted with an Aridus II™ Desolvating Nebulizer System (Teledyne Cetac, Omaha, NE, USA). Measurements were performed following the standard-sample-standard bracketing method using a NIST SRM 987 SrCO 3 (0.71034 ± 0.00026, National Institute of Standards and Technology, Gaithersburg, MD, USA) as the standard. All samples were prepared in triplicate.

Data Evaluation
Metadata information is explained in Supplementary Table S1, and all data obtained are presented in Tables S2 and S3. Since the Ag content in most samples and Se in all samples (Table S3) were below the LOD, they were excluded from further data analysis. The data were analyzed using Microsoft ® office Excel 2019 for basic descriptive statistics and OriginPro 2021 for multivariate analysis: Spearman correlation coefficient analysis (SCA), hierarchical cluster analysis (HCA), and principal components analysis (PCA). The HCA method was used to order data and create groups that share common properties. Euclidean distances were chosen as the distance between the different sampling times, and Ward's method was used to form the clusters.
Finally, the simple linear mixing model (SLMM) was used to quantify the relative contribution from each well (Kleče and Brest), using the long-term average concentrations of sodium (Na), chromium (Cr), and arsenic (As) in the two end-members as follows: Here, the subscript JSI, K and B represent tap water mixture at the JSI and the two water sources Kleče and Brest. For the end-member concentration, long-term data collected at private households at the Kleče and Brest wellfield were used as no simultaneous sampling of source water was provided [45].

Results and discussion
The results are presented as electronic supplementary material (Tables S2 and S3) and summarized graphically in Figures 2 and 3. The descriptive statistics of all hydrogeochemical parameters are presented in Table 1.  Differences observed over the 24 h sampling period were relatively small (Table 1), but a more detailed inspection reveals variations in particular parameters. For instance, after stabilization, the temperature varied between 13.5 • C to 14.5 • C, however the initial reading after one minute was 18.9 • C. The pH ranged from 7.10 to 7.61 (Table 1) and is in the range required for drinking water in Slovenia, i.e., between ≥6.5 and ≤9.5 [20], which is similar to the European standards [21,46] (Table 1). The EC values were between 325 µS/cm and 355 µS/cm and were below the official standard for drinking water of 2500 µS/cm (Table 1). Based on EC classification [48], tap water at the JSI sampled has a low mineral concentration and is homogenous.

δ 2 H and δ 18 O in the Tap Water
The diurnal variations of stable isotope ratios in tap water are presented in Figure 4 and, as a whole, show only slight statistically nonsignificant variations. The δ 2 H and δ 18 O vary from −62.0‰ to −61.1‰ and from −9.30‰ to −9.26‰ (Table 1), respectively, standard deviations for hourly δ 2 H and δ 18 O are 0.2‰ and 0.02‰, respectively. The highest and the lowest δ 2 H appeared at 05:00 and 21:00 (Figure 4a). The highest δ 18 O appeared at 09:01 and 08:00, and the lowest at 15:00, 23:00, and 05:00 (Figure 4b). Also, there was no significant difference between the average values in tap water during the daytime (09:01-20:00) and night-time (21:00-09:00) [49]. However, there is a slight difference between the average isotope values during morning hours (09:01-11:00 and 07:00-09:00) and the rest of the day/night time (12:00-06:00). The difference for δ 2 H and δ 18 O is 0.13‰ and 0.04‰. However, the variability is still small and is in the range of measurement uncertainty. It also shows that the sources of tap water from Kleče and Brest have similar isotope compositions, which prevents the differentiation of tap water origin solely using stable isotope ratios (Figure 4). No correlation (r = 0.3) was observed between δ 2 H and δ 18 O since the source water does not change in such a short period (24 h). Moreover, as the water moves through the soil, the groundwater signal is attenuated resulting in small differences between aquifers. The values obtained are typical for the source water, i.e., groundwater from shallow aquifers of Ljubljansko polje and Ljubljansko barje [25].
All values fall close to the Global Meteoric Water Line (GMWL) and the Local Meteoric Water Line for Ljubljana [50], confirming that groundwater from the Ljubljansko polje aquifer originates primarily from infiltration of local precipitation and water from the River Sava [18] and at the Ljubljansko barje aquifer from precipitation and surface streams [33]. Deuterium excess (d) was also relatively constant, with an average of 12.5‰, and ranging between 12.2‰ and 13.3‰. Therefore, it could not be used as an indicator of different water sources.

δ 13 C DIC in the Tap Water
The 24 h δ 13 C variability of tap water ranged from −12.1‰ to −11.4‰ (Table 1). The lowest δ 13 C DIC of −12.1‰ was observed at 14:00, 15:00, 17:00, and 3:00, while the highest value of −11.4‰ occurred at 08:00. Samples with lower δ 13 C DIC are more characteristic for Brest, while samples with higher δ 13 C DIC are characteristic for Kleče [26]. All values in tap water indicate both biogeochemical processes: CO 2 produced during organic matter decomposition and carbonate dissolution.

Concentrations of Major and Trace Elements in the Tap Water
The largest variation in major elements was observed for Na (CV = 28.1%; Figure 5a) followed by K (CV = 14.2%) and Mg (15.0%; Figure 5b) (Table 1). Generally, higher values were observed for Ca and Na at the beginning and at the end of the experiment and opposite for Mg and K that had lower values at the start and end of the experiment (Figure 5a,b). However, data for K show much higher fluctuations during the 24 h experiment. There is no numerical Slovenian drinking water quality guideline for Ca, Mg, and K (Table 1). Trace elements were detected in all water samples, but were below the limits set by the Slovenian regulation and the EU Drinking Water Directive [20][21][22]46,47].

Sr/ 86 Sr Isotope Ratio
The 87 Sr/ 86 Sr values are presented in Supplementary Table S3. The Rb concentration in the samples was 0.558 to 0.614 µg/L, with the lowest values at 10:00 and 08:00. The Sr concentration was in the range of 65.2 µg/L to 147 µg/L. The water collected during morning hours (at 10:00 and 08:00) had a lower Rb/Sr ratio than the samples collected at 17:00, 00:00, and 03:00. The hours when the Rb/Sr ratio corresponds to the low Mg 2+ /Ca 2+ ratios (chapter 3.4) allow us to conclude that the samples collected in the morning belong to the water from Kleče, while higher ratios indicate the prevailing water from Brest [27].

Multivariate Statistical Analysis
The multivariate statistical analysis was performed using SCA, HCA, and PCA. The SCA results are summarized in Supplementary Figure S1, HCA is presented in Figure 7 and Supplementary Table S4, while PCA results are presented in Supplementary Table S5. The final dataset used for multivariate statistical analysis is a data matrix of 25 samples (observations) by 32 parameters (variables) for SCA, while for the PCA and HCA, only 25 parameters, i.e., major and trace elements, were used ( Table 1). The distribution of most chemical parameters is positively (Al, B, Ba, Cr, Fe, Li, Mo, Ni, Pb, Sb, Sr, Zn) or negatively (Mg, Mn, As, Rb, V) skewed and only a few are close to a normal distribution (Ca, K, Na, Cd, Co, Cu, Tl, U). Chemical parameters that were positively or negatively skewed were log-transformed. Finally, standardization was applied to 16 lognormal data and 12 normal distributions to ensure that each variable is weighted equally. δ 2 H, δ 18 O, δ 13 C DIC showed a significant correlation with 12, 15, and 18 other parameters, respectively, and can therefore be used as possible indicators for determining water origin and changes in the WSS. A significant positive Spearman correlation (r s = 0.9) exists between Ca and Na, B, Ba, Cr, Li, and Sr, suggesting a common water origin. It also shows a high negative correlation (r s ≥ −0.7) with As, Mg, Mn, and V, resulting in a possible different origin of the water.
HCA was first performed on the whole set of data (Figure 7a; N = 25), and then on the results after removing the first observation (sample: 09:01; Figure 7b; N = 24). In this study, a horizontal line is drawn across both dendrograms at a linkage distance of about 12. Three distinct clusters were identified: A2 (N = 6), A3 (N = 11), and A4 (N = 7), while cluster A1 (Figure 7a) represents only the first sample. Observation of the dendrogram reveals some similarities between clusters; however, the clusters A1 (only in the left dendrogram) and A2 are less similar as they have high linking similarity clusters A3 and A4. To describe the characteristics of each cluster, Table S4 presents the median values of geochemical data. To the group A1 belongs sample collected at 09:01, to group A2 belong samples collected from 10:00 to 12:00 and from 07:00 to 09:00 and can be attributed to water from Kleče. On contrary, to A3 belong samples collected during the hours when majority of water was coming from Brest (from 13:00 to 20:00 and 01:00 to 03:00). To the group A4 belong samples collected from 21:00 to 00:00 and from 04:00 to 06:00 and represents hours in between the origin change. For cluster A1, we can see that the elevated median values of Fe, Ni, Pb, and Zn indicate the leaching of elements from the WSS collected at the beginning of the experiment and can be linked to the visual classified Group 3. The second dendrogram shows that the clusters include the same tap water samples as in the first dendrogram; however, the representative median concentrations changed (Table S2). The most representative elements for the A2 cluster of the second dendrogram are Ca, Na, B, Ba, Cr, Fe, Li, Mo, Sb, and Sr, while for the A3 cluster, higher values of K, Mg, As, Cd, Co, Ni, Rb, Tl, V, and Zn are characteristic. Most elements (Ca, Na, B, Ba, Cr, Li, and Sr) representative of A2 belongs to Group 1 (Fe, Mo, and Sb belong to Group 3), while elements K, Mg, As, V and Cd, Co, Ni, Tl, Zn belong to visually divided Group 2 and Group 3, respectively (Figure 5b,c). For A4, the most representative elements are Al, Cu, and U that belong to Group 4 ( Figure 5d) and confirm the mixing of water, presented in Figure 6.
In 2018, the multi-elemental analysis was performed on water samples from one well in Brest and one well in Kleče. The results showed higher values of B, Ba, Li, and Sr in Kleče, whereas in Brest, higher Mg, As, and V values were detected [27]. Based on all results, we can conclude that the tap water samples in the cluster A2 represent water from Kleče and those in A3 from Brest.
The first two principal components account for 72.7% of the total variance in the dataset; the principal component loadings are presented in Table S5 (left). Loadings that represent the most important variables for the components are bolded for values greater than 0.25. PC1 explains the greatest amount of the variance and is characterized by positive loadings in Ca, Na, B, Ba, Cr, Li, and Sr (Table S5), and belongs to Group 1 ( Figure 5a) and can be attributed to the wellfield Kleče. Component 2 is characterized by positive loadings in Al, Cd, Co, Fe, Ni, Pb, and Zn, were all but Al represent Group 3 ( Figure 5c).
Further, the first sample was excluded, and the analysis was performed for 24 samples. The results are presented in Table S5 (right). The first two components account for 70.2% of the total variance in the dataset. Component 1 is characterized by positive loadings in Mg, Mn, and V, representing Group 2 ( Figure 5b, Table S5), while component 2 is characterized by positive loadings in Mo, Sb, and Tl. The latter elements belong to subgroup 3. The most significant loadings belong to elements related to the leaching from the pipes within the WSS.

Mixing of Water
Tap water can involve two or more discrete end-members and is easy to observe and calculate. It is important to demonstrate that the tracers of mixing behave conservatively and not react with other solutes, solids, or gases. For the source partitioning, Na was selected for SLMM. Also, Cr and As were used since they represent common contaminants in groundwater [51]. It is known that Ljubljansko polje is higher in Cr, while higher As values are characteristic of Ljubljansko barje. Also, Cr and As are highly correlated with Na and can be used for SLMM. Data of Na concentrations were gathered from 2016 to 2019 in periodical sampling of drinking water [45] with an average value of 4.9 mg/L (min = 3.3 mg/L, max = 7.7 mg/L) for Kleče and 1.6 mg/L (min = 0.99 mg/L, max = 4.4 mg/L) for Brest. By comparing these results with this study's data, all tap water samples fall in the mixing area between Kleče and Brest. Using equation 2.3.5 and average Na concentration data for end members, the proportion of water from Kleče was calculated. Results show that over 24 h the mixing ratio changed from 7% to 90%. When using Cr and As as end-members and long-term data from private users [45], it gave estimates of 23% to 74% and 0.1% to 100%, respectively. Moreover, based on the high positive correlation between Na and, e.g., Mg, Li, Mn, Sr, and V these elements could also be used for the source determination if the end-member concentration would be known.

Conclusions
This study is, to the best of our knowledge, the first to look at the variability of stable isotopes and elemental composition during a 24 h analysis of tap water. It involved in-situ monitoring of T, EC, and pH. Samples were collected for multi-elemental and stable isotope analyses (δ 2 H, δ 18 O and δ 13 C DIC ) every hour over 24 h in tap water at a location where water from Kleče and Brest wellfields is mixed.
Using the isotope composition (δ 2 H, δ 18 O, and δ 13 C DIC ) to determine the mixing ratios remains inconclusive given the isotope similarity between the two sources waters. Concentrations of elements, although low, did carry more information. Based on the visual observation of temporal differences, four groups were identified. The characteristics of these were: higher values in the beginning and at the end and lower between; lower values in the beginning and at the end and higher in between; higher values at the beginning of the experiment; and a group with no specific pattern. While the isotopes can help us to understand the sources and dynamics of flows in urban areas, the use of additional hydrogeochemical parameter in the SCA, HCA, and PCA analyses helps with testing and identifying hourly variability and the origin of tap water at given time. Based on SCA we divided the investigated parameters into three groups based on the number of significant correlations: parameters that have 10 or less significant correlations (i.e., d), parameters with 11 to 19 significant correlations (i.e., δ 18 O) and parameters with 20 or 21 significant correlations (i.e., Ca). Stable isotopes are weakly but significantly correlated with different parameters. Considering HCA for all samples collected, samples were linked together based on the time of sampling: the tap water was divided into four clusters (A1, A2, A3, and A4), however when removing the first sample, samples were grouped into three groups (A2, A3, and A4). The cluster A1 (also part of the third visual observed pattern), indicates the influence of the leaching of specific elements (e.g., Fe, Ni, Pb, Zn), probably due to leaching from the water pipe of the internal WSS at the beginning of the experiment, when drinking water starts flowing. The latter was also confirmed with elevated PCA loadings.
Altogether, the results indicate that, as expected, the proportion of water from Kleče and Brest changed throughout the day. However, since no simultaneous data from the sources were provided, the long-term average concentration of Na was considered for SLMM. The proportion of water from Kleče changed from 7% to 90% over the 24 h experiment. Moreover, Cr and As show similar mixing ratios. The results show that at the beginning and the end of the experiment, a higher proportion of the water was from Kleče, whereas between 12:00 and 18:00, most water was from Brest. It should be emphasized that the water managers know the WSS, namely the exact information of the capacity at which wells operate, measured in 15 min intervals. However, by performing this experiment we confirmed their assumptions about the mixing of water in the investigated area with application of elemental composition of tap water. The results, which positively reflect assumptions, show that during the day, when water consumption is high, Brest wellfield contributes a larger share of water. During the night, the Brest wellfield contribution is low and the Kleče wellfield contribution pushes the watershed between the wellfields to the south of the city.
Within this study we can conclude that elemental composition of some elements (i.e., N and As) could be used to provide good proxies of water mixing from two different reservoirs. We must acknowledge that shallow aquifers are characterized by the hydrometeorological seasonal variability that affects the water chemistry. At the time of the experiment, data on the chemistry of water from wells had not been acquired, and long-term data of regular monitoring was used. Therefore, long-term multi-parameter monitoring should be established to determine the monthly and seasonal variations. Finally, we planned to repeat the experiment under different conditions in WSS (i.e., during different seasons and the COVID-19 pandemic when water consumption has significantly changed due to lockdowns) to deduce parameters that can help in long-term evaluation of mixing water at the tap.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/w13111451/s1, Figure S1: Pearson correlation matrix for all the analyzed parameters. * p ≤ 0.005. H, O, and C represents δ 2 H, δ 18 O and δ 13 C DIC , respectively; Table S1: Metadata information regarding data attributes; Table S2: Basic sampling attributes and results of temperature (T), electrical conductivity (EC) and pH for all measurements; Table S3: Elemental compositions of the analyzed tap water samples collected every full hour; Table S4: Geochemical characteristics of median concentrations: left for four clusters and right for three clusters; Table S5: Principal component loadings and explained variance for the two components.