The study area and sampling points were identified using online geological maps, which contains data under the Norwegian license for public data (NLOD), made available by the Geological Survey of Norway (NGU). These maps show major bedrock types (see Figure 1) and the estimated indoor radon potential across Norway (45, 46). However, due to the low density of registered wells in some areas, and to avoid identification of participants in this study, the exact sampling locations are not marked.

The study area itself is a sparsely populated area, part of the Caledonian nappes in western Norway. High indoor ²²²Rn has been recorded previously and five anonymous volunteers plus one waterwork agreed to participate in water quality monitoring of the current study. The sampling points are located in the same general area, within 10 km of each other in steep terrain. The catchment area likely extends to the top of the hillside (800–1,000 meters above sea level) and there is no industrial activity and limited human activity upstream of the collection points. According to the bedrock maps of NGU (Figure 1), geology of the area consists of granite and rhyolite, which are chemically similar rock types known for containing ²³⁸U (47, 48). The cover material of the study area is dominated by glaciofluvial deposits and avalanche material (49). Information about the water wells was collected from the national groundwater database GRANADA (50). In order to investigate how water quality in different water sources vary over time, six sampling points were selected:

Following the initial sampling conducted in June of 2023, the owners of sampling points 1 through 5 agreed to participate in monitoring of water quality over time. Before the sampling in September of the same year, sampling point 6 the granite well, was identified and was also included in monitoring over time. Sampling frequency was largely dictated by planning and coordination with the volunteers. In order to assess the evolution of drinking water quality between winter and summer, samples were collected monthly from January through July of 2024. The last sampling, conducted in November of 2024, was performed shortly after a heavy precipitation event in order to investigate the effects of large volumes of rainwater. The final sample number ended up being 8 for the granite well and stream source, and 9 for other sampling points.

Meteorological data from the study area was collected from a publicly database available through the Norwegian Centre For Climate Services (51). Values for total precipitation and average air temperature from the month prior to each sampling were used to assess seasonal dependency of water quality. Measurements of pH, electric conductivity (EC) and water temperature were performed on site using a WTW Multi 3,401 combined pH and EC meter with a Sentix 41 and Tetracon 325 electrode. In order to determine total organic carbon (TOC) untreated samples were collected in 10 mL tubes. The samples were analyzed using a Shimadzu TOC-analyzer.

Stable elements, ²²⁶Ra, ²³²Th and ²³⁸U were measured by ICP-MS. Before ICP-MS analysis, samples were acidified to 5% v/v HNO₃ The analysis of ²²⁶Ra was done on Perkin Elmer NexION 5,000, and ²³⁸U, ²³²Th, as well as the content of 52 major-and trace elements in water samples was determined using the Agilent 8,900 QQQ ICP-MS. Anion chromatography was performed on 1.5 mL of untreated sample. Quantification of fluoride (F⁻), chloride (Cl⁻), sulfate (SO₄²⁻), and nitrate (NO₃⁻) concentrations were done on a Dionex ICS-6000 HPIC.

The activity concentration of polonium-210 (²¹⁰Po) was determined by α-spectrometry from 10 L water samples. Polonium was concentrated using iron hydroxide co-precipitation and spontaneous deposition on nickel disks following a method adapted from Chen, Aarkrog (52) and Skipperud, Jørgensen (53). Another isotope of polonium (²⁰⁹Po) was used as yield monitor. Analyses were done on Ortec Model 7,401 spectrometers. The samples were then left for approximately 6 months, allowing the ingrowth of ²¹⁰Po before the process was repeated. This allowed for the determination of lead-210 (²¹⁰Pb) using Equation 1 (54):

where A_Pb, is the estimated activity concentration of ²¹⁰Pb, A_Po is the measured activity concentration of ²¹⁰Po, T_Pb and T_Po are the half-lives of ²¹⁰Pb and ²¹⁰Po respectively, and t is the time elapsed between the first and second separation.

Due to the volatility of ²²²Rn-gas, loss of analyte during and after sampling is a common problem (55) Samples used for determining ²²²Rn activity concentrations were collected by a method adapted from Strand and Lind (56). Ensuring minimal air contact, water samples were taken into scintillation vials prefilled with Maxilight water immiscible scintillation cocktail. The samples were sealed using hinge tape and aluminum (Al) lined caps and stored in the dark at room temperature for no more than 3 days. Sample stability was tested by storing three samples taken from the granite well in room temperature for just under 9 days and estimating activity concentration of ²²²Rn (Figure 2). No meaningful loss of analyte could be identified in the relevant time frame due to storage alone. Analysis of ²²²Rn was done using a Hidex 600SLe liquid scintillation counter. The samples were counted at 17°C for 30 min. The estimated counting efficiency was 2.688 as ²²²Rn, ²¹⁸Po and ²¹⁴Po are all counted.

The radiological risk from NOR in each sampling point was assessed by calculating ID according to Equation 2 using effective dose coefficients found in (57) and assuming a consumption rate of 730 L/y (20):

where E is the ID in Sv/y, a is the activity concentration of a certain NOR in water, e is the effective dose coefficient for adults.

To investigate covariance between different parameters over time, correlation coefficients were calculated in R (v.4.4.1). Due to challenges in planning and coordination of sample collection, the final dataset is limited. To avoid making inaccurate assumptions about data distribution, all statistical analyses were done using non-parametric statistical tests ranked data. Spearman’s rho does not require data to be normally distributed, and was used for measuring the monotonic association between pairs of variables. Since analyses were performed continuously throughout the monitoring period, most of the studied variables had multiple overlapping reporting limits. For radiometric analyses decision thresholds (DT) and detection limits (DL) are calculated according to De Felice, Jerome (58). For mass-based measurement techniques, limits of detection (LOD) and quantifications (LOQ) are calculated as three and ten times the standard deviation of method blanks, respectively. To enable statistical analysis on variables with one or more censored observations, Kendall’s rank test on U-scores was used instead (59). Correlation was considered significant when p < 0.05.

The vast majority of the 65 parameters investigated in this study area were within the norm values according to Norwegian drinking water regulations (18, 19). The main constituents of these waters are the alkaline and alkaline-earth metals, as well as Cl⁻, NO₃⁻, SO₄²⁻ (Table 1). The concentrations of these vary between the sampling points as well as over time. The measured concentration of F-in the granite well was relatively high compared to other sampling points, but still no values above the limit of 1.5 mg/L were measured. The concentration of SO₄²⁻ was also higher in this sampling point than in any other measured, although it remains unclear whether this is due to a difference in kinetic factors, or due to the different mineral compositions of the aquifers. Water from both the granite and rhyolite rock well displayed high Ca concentration at times when compared to the other sampling points. While this may cause issues in terms of usage, it is not associated with negative health effects (60). Water from the rhyolite displayed a larger range in observed EC and concentration of major constituents, although the variations in pH and temperature were comparable to that found in the granite well. In the rhyolite well, the concentration of arsenic (As) varied between 5.6 and 7.5 μg/L, but was never observed above the 10 μg/L limit. Concentrations of Al up to 110 μg/L was also observed, which is higher than most other sample points, but below the drinking water limit of 200 μg/L. Although water quality from the two rock wells were similar at times, it seems the rhyolite well is, to a greater degree, affected by outside influence.

Between the two sample points in the unconsolidated well, there is little change in the parameters listed in Table 1. However, certain elements appear to differ slightly in concentration: apart from the last measurement, the concentration of iron (Fe) varied between 3.8 and 6.4 μg/L in sampling point A, and varied between 2.0 and 3.0 μg/L in sampling point B. Similarly, manganese (Mn) varied between 0.27 and 0.92 μg/L in sampling point A, and 0.19 and 0.38 μg/L in sampling point B. Both metals are redox sensitive and are likely to precipitate during storage in contact with air, a mechanism that is used for removal of the metals (31, 61). This could also simply be caused by random variation and the distance between the sampling points, as in the last samples collected from these points, concentration of Fe was 10 and 22 μg/L, and the concentration of Mn was 0.72 and 1.2 μg/L in sampling point A and B, respectively. This coincides with the lowest measured pH value of 6.0 and may be caused by the extreme meteorological precipitation leading up to the sampling day. In comparison, the spring source is located on similar bedrock, in steeper terrain, but lacking the recharge from nearby surface waters. It would, therefore, be expected that the water from the spring is slightly more mature than water from the unconsolidated well, something that is reflected in a higher EC and higher concentrations of SO₄²⁻, Na, K and Ca. Water from the stream source had the lowest measured EC, pH, and concentrations of SO₄²⁻, Mg, K and Ca. This is also where the highest TOC was found, an indication of high biological input. Compared to the other sources, the low ion concentration indicates limited interaction between this water and the mineral phase, and the water is exposed to atmospheric and biological input comparable to the rhyolite well and sampling point B in the unconsolidated well. Also similar to the rhyolite well, high concentrations of Al up to 170 μg/L were measured in the stream water. The two sampling points are located less than 1 kilometer apart and may have similar mineral phases even though the waters undergo different processes before being collected.

Differences in water quality between the sampling points are likely explained by the different source types. The lowest observed pH and EC were found in the unconsolidated well and the stream. Water from the surface and near-surface is composed in a high degree of rainwater, i.e., slightly acidic and low in electrolytes (31). Meanwhile, groundwater from the spring, and especially the rock wells have higher pH and EC due to higher dissolution of minerals during contact time. This is comparable to a study by Banks, Frengstad (11), where the median pH in a high number of groundwater samples from different types of crystalline aquifers was found to be 8.07, and slightly lower pH-values in quaternary groundwater and surface waters. The difference in pH was attributed to carbonate mineral dissolution in the crystalline wells. Consequently, mineral dissolution is associated with an increased electrolyte concentration, and thus, higher EC (62, 63). The range of measured EC in the rhyolite well was much greater than in the granite well, indicating less stable conditions over time.

The highest observed concentration of ²³⁸U was 13 μg/L, the equivalent of 0.16 Bq/L, and was found in the granite well. This is higher than the 3.26 μg/L median found in groundwater drinking water in Sweden, but well below the recommended limit of 30 μg/L, based on its chemical toxicity (21, 43). According to Knutsson (44) the highest concentrations of ²³⁸U and ²²²Rn measured in Scandinavian groundwaters are 12,400 μg/L and 77,500 Bq/L respectively, both found in Finland.

Measured concentrations of ²³²Th varied between <LOD and 0.11 μg/L. The highest concentration is equivalent to 0.45 mBq/L and was recorded in water from the rhyolite well. Thorium has very low solubility under most conditions and the highest reported concentrations in Norwegian drinking waters are in the range 3.1 and 4.76 μg/L (38, 64). Concentrations of ²²⁶Ra were never observed above the LOQ of 73 mBq/L in any of the investigated sampling points, which is below the maximum observed activity concentrations in groundwater from Sweden and Finland, which are 2.08 Bq/L and 7.5 Bq/L, respectively (41, 43). Activity concentration of ²¹⁰Pb in the granite well varied between 7.19 and 41.7 mBq/L over time. This is comparable to that found in a study of Finnish groundwaters, where out of 288 samples the median activity concentration of ²¹⁰Pb was 14 mBq/L and the maximum 540 mBq/L (42). In the Finnish ground water study, activity concentrations up to 2.0 Bq/L of ²¹⁰Po where found, while the median value was 9 mBq/L. Meanwhile in the present study, observed activity concentrations of ²¹⁰Po varied by two orders of magnitude, between 2.37 mBq/L and 312 mBq/L.

The radiological risk of naturally occurring radionuclides apart from ²²²Rn was evaluated by calculating the ID in each sampling point (Table 2). A conservative estimate was made using the highest measured activity concentration of each NOR during the sampling period, dose conversion factors were obtained from ICRP publication 119, and assuming a consumption rate of 730 L of water per year (20, 57).

Large variations in measured activity concentration of ²²²Rn were observed across the sampling points. As expected, the highest activity concentrations of ²²²Rn were found in the rock wells, as waters here typically have longer residence time and are enclosed in the aquifer. Although measured activity concentrations in the granite well greatly exceeded the limit of 100 Bq/L as shown in Table 2, values were in line with observations in other studies of similar areas. Mean values for waterborne ²²²Rn in previous studies of Norwegian wells, primarily drilled in rock, have found a mean value of 400 Bq/L, while Finnish private rock wells tend to be slightly higher, with a mean value of 930 Bq/L (41, 65). A decrease in measured activity concentration of ²²²Rn was observed between sampling point A to B in the unconsolidated well. During aeration and storage of the water ²²²Rn is likely removed through emanation (3, 14).

The National Research Council (14) suggested an effective dose coefficient of 3.5*10⁻⁹ Sv/Bq for ingestion of ²²²Rn by adults. Using this coefficient, the estimated dose from the highest observed activity concentration is 3.1 mSv/y, making ²²²Rn the greatest contributor to ingested dose, although it is not typically included in calculation of ID (20). Otherwise, the greatest contribution to ID was in most cases from ²¹⁰Po, while those of ²³⁸U and ²³²Th are negligible compared to the 0.1 mSv/y ID limit. Even though the concentration of ²²⁶Ra and ²¹⁰Pb is below the derived concentration defined by European Council (20), the dose contribution could be significant depending on the concentration of other NORs not included in this study. Komperød, Rudjord (66) estimated that an average dose of 0.054 mSv/y from all types of drinking water to the Norwegian population, where the estimated dose from ingested ²²²Rn alone was 0.049 mSv/y. In the present study of groundwater in a ²²²Rn-prone area in western Norway, estimated doses from NORs other than ²²²Rn were in the range of 6.7⨯10⁻³ to 0.31 mSv/y exceed the national average of 0.05 mSv/y, while ²²²Rn alone was the greatest contribution to ingested dose.

Over the course of the monitoring period, water parameters in the different sampling points displayed varying degrees of stability (Table 1). Meteorological data show that from winter to summer the air temperature increases from just below zero to almost 20°C during summer. The monthly precipitation varies from above 150 mm in January to just above 50 mm in June. Leading up to the final sampling in early November, there was a high rainfall event with 116 mm registered in 1 day. In general, precipitation was higher during the cold season and a significant (p < 0.05) negative correlation was found between air temperature and precipitation. However, due to the large catchment area, potential frost during winter, and different types of water source, there is not necessarily a direct connection between precipitation and recharge, although precipitation in the area rarely comes as snow.

The water quality in the granite well was relatively stable throughout the monitoring period. The activity concentration of ²¹⁰Po was at its highest during winter and lowest in summer, although activity concentration of ²¹⁰Pb and ²¹⁰Po show very limited correlation with other variables. A statistically significant (p < 0.05) positive correlation was observed between precipitation, SO₄²⁻, Mg, K, REEs, ²²²Rn and ²³⁸U. There was also a positive correlation between Na, Rb, Sr. and Cs which correlate negatively with the previous group of variables. Additionally, a positive correlation was found between TOC and Mg, K, Cu, Zn Ba, Sm and Er using Kendall’s rank correlation (p < 0.05). The lack of coherence between different group 1 and 2 elements may be caused by anion exchange processes or the presence of minerals with different weathering rates (31). A significant (p < 0.05) positive correlation was observed between activity concentration of ²²²Rn and precipitation, possibly indicating that the gas is dissolved in recharge water and transported to the aquifer. A similar observation was made in a karst aquifer in Switzerland (67). The study found that ²²⁶Ra-content was higher in regolith compared to the underlying bedrock. This led to higher activity concentrations of ²²²Rn in pore-waters and thus higher activity concentrations in the aquifer following recharge. This process could explain the observations made in the granite well in the present study, although a better understanding of the role of catchment geology is needed to be certain.

Comparably, the conditions in the rhyolite well were less stable as shown in Figure 3. A positive correlation was found between precipitation, TOC, and REEs, as well as between REEs and Al, Mn, Fe, Co and ²³²Th (p < 0.05). These metals are typically associated with complexes, colloids or particles, both organic and inorganic (68–70). A significant (p < 0.05) negative correlation was found between pH, and REE and ²³²Th, meaning the metals may desorb due to chemical changes. Alternatively, particles may be suspended during influx of rainwater. Meanwhile, there was a negative correlation between precipitation, and EC, Na, Mg, S, K, Ca and ²³⁸U. Apart from in aquifers composed of highly insoluble minerals, the amount of dissolved minerals in groundwater is typically dependent on residence time alongside surface area, i.e., grain size or fracture geometry (31, 38, 71–73). In periods of high precipitation, the groundwater is possibly diluted by less mature recharge water, leading to lower concentrations of certain metals. High variability in ²²²Rn activity concentration and EC was observed in the rhyolite rock well, as well as the concentrations of several other metals. It has been found that episodic recharge of surface water may lead to rapid changes in water composition (74). During spring this source is likely affected by intrusion from rainwater, as well as meltwater coming from higher parts of the terrain. As ²²²Rn-content might also be affected by such events, it is possible that the lack of correlation with other parameters is due to the time it takes for ²²²Rn to be produced from ²²⁶Ra-decay.

In sampling point A of the unconsolidated well, temporal variation in ²²²Rn concentration correlated positively with EC, Na, Mg, K and Ca (p < 0.05), which may be explained by mixing of surface water and more mature groundwater. A significant negative correlation was observed between ²¹²²Rn and air- and water temperature (p < 0.05), and during winter there is considerably less water in the nearby river. This leads to less rapid recharge from surface water, meaning more mature water is collected in the well, and thus a higher activity concentration of ²²²Rn and metals associated with mineral dissolution. In sampling point B in the unconsolidated well, seasonal variation in major constituents resembled the patterns observed in sampling point A. The ²²²Rn concentration in sampling point B was, however more stable over time as much of the gas has already emanated during storage, highlighting the usefulness of aeration as a method for ²²²Rn removal (3, 28).

In the spring source there was a negative correlation between air temperature and EC, Na, Mg, K and Ba, indicating a seasonal variation in mixing. It is unknown how this source is recharged, as there are no bodies of water at surface level near the water work. The activity concentration of ²²²Rn remains relatively stable compared to other sampling points, except for the last sampling in November 2024. At this point in time, activity concentration increased two-fold from previous measurements, which is unexpected considering the dilution observed in the unconsolidated well and the rhyolite well. This coincides with the highest measured concentrations of all REEs. Similarly to the granite well, the high activity concentration of ²²²Rn may be caused by recharge water percolating through material with higher ²²²Rn-emanation rate and with longer residence time.

The stream had the lowest measured EC, during the entire monitoring period and comparably low concentrations of most elements. Similarly to the spring and granite well a positive correlation was observed between REEs and air and water temperature, TOC, Al, Ti, V, Cr, As, Zr, Ag, Sb, Pb and U. Although very little ²²²Rn was found in this sampling point, this pattern is similar to that observed in the other sampling points, and these metals are likely controlled by similar processes as in the other sampling points, such as complexation and or adsorption (75).