Can activated carbon filtration of groundwater cause radiation safety problems?

Granular activated carbon (GAC) filtration is commonly used to remove trace contaminants, organic micropollutants such as pesticides, and their degradation products from groundwater. At the same time, the filter adsorbs radon (Rn-222), which is present in all groundwater. Radioactive radon and its decay products can pose a radiation safety risk to the operating personnel and a waste management problem for spent activated carbon (Lowry et al. 1988; Annanmäki et al. 2000).

Radon is a radioactive noble gas that does not react chemically in the environment. It is continuously generated from ubiquitous crustal uranium (U-238) with a half-life of 4.5 billion years. Gaseous radon dissolves easily in groundwater and moves well with groundwater flows and soil air currents. Its mobility is limited by its short half-life, 3.8 days. The radon concentration in groundwater depends on the uranium concentration in the soil and bedrock and the quality of groundwater. The groundwaters are often soft in Finland, and high bicarbonate concentrations in groundwater correlate with higher uranium concentrations (Lopes et al. 2017). The typical radon concentration in dug wells in Sweden is 10–300 Bq/L (in granite areas 40–400 Bq/L), average 43 Bq/L (Åkerblom et al. 2005). The average concentration in Finland is 50 Bq/L (Vesterbacka et al. 2005), and in 3% of the wells the concentration is over 300 Bq/L (Lopes et al. 2017).

When water is used, radon is released into the air where it forms solid decay products, which are also radioactive. The inhaled decay products attach to the lungs and bronchi and cause radiation exposure. Radon is also absorbed by the body through ingestion (ICRP 2017; Zagà et al. 2021). Due to the carcinogenic risk, the EU Commission has set a parametric value 100 Bq/L for the radon content of tap water, but a Member State may set a maximum level of 100–1,000 Bq/L in national legislation (EURATOM 2013a).

Radon adsorbed on the GAC filter decays through four short-lived radionuclides (Po-218, Pb-214, Bi-214 and Po-214) into the long-lived isotope of lead, Pb-210. This generates gamma radiation that can be detected outside the GAC filter. Gamma radiation is suspected to pose a radiation safety risk near GAC filters installed in homes when the radon content of raw water is very high (Lowry et al. 1988). In contrast, gamma radiation from GAC filters used in the removal of environmental pollutants in groundwater plants, has been less studied.

Another potential problem is the long-lived lead isotope Pb-210 formed in GAC filters. If the concentration of Pb-210 in activated carbon exceeds the clearance level of 1 kBq/kg of Naturally Occurring Radioactive Materials (NORM) laid down in the EU directive, or the national clearance level which may be even stricter, the owner of the waste must notify the competent authority (EURATOM 2013b). In these cases, the competent authority will usually require a more detailed assessment of the radiation exposure. The investigation should cover occupational and public exposure caused by the storage, handling, transport, and disposal of the used GAC. If the clearance level for the occupational or public exposure is exceeded, the handling and disposal of GAC may require licensing. Preparing the investigation and the application for a license is time consuming and requires the consulting of a radiation protection expert. If NORM waste is treated in an incinerator, Pb-210 will be transferred to the fly ash and equipment of the incinerator (IAEA 2003), so incineration is not recommended. Most European states allow NORM waste to be landfilled, provided the procedure is approved by the competent authority. For example, in France, U-238 series NORM waste can be landfilled at a maximum concentration of 20 kBq/kg and in Sweden 10 kBq/kg of dry matter (Garcia-Talavera et al. 2021).

Similarly, waste legislation includes a ban on landfilling of organic carbon, which in some cases has been interpreted as also applying to activated carbon waste (EC 2002). Recycling and other uses are the primary methods of waste treatment compared to landfilling. However, landfill restrictions do not apply if it can be justified that the waste is not suitable for recycling or energy production and if the landfill would be the best solution environmentally. The interpretation of the waste legislation on the landfilling of activated carbon exceeding the NORM clearance level concentration is unclear and vary from country to country (Garcia-Talavera et al. 2021). Exceeding the NORM limit value can cause significant additional costs for the design, approval, and implementation of carbon waste disposal.

In this study, long-term adsorption of radon on activated carbon was measured at two groundwater plants. The aim of the study was to assess the magnitude of the radiation safety risk associated with GAC filtration. The aim was also to draw up recommendations to ensure the radiation safety of workers during the use of the activated carbon and to avoid exceeding the clearance level of Pb-210 and thus waste treatment problems.

The Pekonniemi groundwater plant is located in the city of Jyväskylä, in Central Finland. The groundwater is pumped from Keljonkangas alluvial aquifer that is part of the Central Finland end moraine. The aquifer has formed at the end of the ice age in the ice-marginal process. As the glacier melted, various water-permeable gravel and sand layers formed in the valleys and slopes of the bedrock (SYKE 2020b). The Pekonniemi intake has one dug well. GAC filters have been installed to remove tri- and tetrachloroethylene in 2010. The daily use of Pekonniemi varied during the study period from 300 to 1,200 m³/d and the plant had production shutdowns lasting several months.

The second research site was the Alalampi groundwater plant in the municipality of Keuruu. The alluvial aquifer is located on a northwest-southeast esker. The core part of the esker is sand and sandy gravel, and the soil at the edges of the esker is silty (SYKE 2020a). The groundwater intake has two screen wells that are used alternately. The distance between the wells is only 10 meters. Pesticides and their decomposition products were detected in the groundwater in 2009 and the following year GAC filters were installed at the plant. The major compounds removed by the GAC filter are atrazine, BAM (2,6-dichlorobenzamide), hexazinone, lenacil, and terbuthylazine. Unlike Pekonniemi, the Alalampi intake stops for a few hours every day.

Pumping histories, water qualities, design values and utilization rates of groundwater plants during the service life of activated carbon (GAC) are presented in Appendixes A-C. The quality of raw water in both intakes is otherwise good, but slightly acidic.

The activated carbon grade used at the Pekonniemi water plant is Silcarbon S835 (Silcarbon Aktivkohle GmbH, Kirchhundem, Germany) and Alalampi AquaSorb^® AS2000 12 × 40 US MESH (Jacobi Carbons AB, Kalmar, Sweden). Both activated carbons are manufactured by a steam activation process from coal. In the spring of 2021, Filtrasorb^® F300 8 × 30 US MESH (Calgon Carbon Corporation, Pittsburgh, USA) carbon was installed as a reference carbon in one of the three filters at Pekonniemi. F300 is made of bituminous coal by a process known as reagglomeration (Kaya et al. 1997). The specific weights and weighing results for activated carbon are presented in Appendix D.

Initially, the radon adsorption efficiency of old activated carbons was studied by analysing the radon concentration in the influent and the effluent and by measuring the ambient dose rate of gamma radiation outside the filters. In addition, material samples were taken from the old carbons for the determination of the activity concentration of naturally occurring radionuclides. After the initial test period, the carbons in both waterworks were replaced with new ones of similar quality and the radon adsorption efficiency and dose rate of the new carbons were also determined. In Pekonniemi, different flow rates were set for parallel filters to obtain samples at different contact times.

Water samples for the determination of radon concentration were taken by passing water from the sample tap with a hose to the bottom of a 500 ml beaker. The water was allowed to drain over the edge until the water in the beaker changed several times and all air bubbles were removed. A 10 ml water sample was taken from the beaker with a pipette and injected into a scintillation vial pre-filled with scintillation cocktail (12 ml of Ultima Gold XR; PerkinElmer Inc, Boston, USA). The water sample and liquid scintillation cocktail were immediately mixed after closing the vial (Kitto 1994).

Ambient dose equivalent rates due to gamma radiation from GAC filters were measured using a calibrated DGM-Turva (Kata Safety Oy, Joensuu, Finland) dose rate meter. Measurements were made on the outer surface of the filter jacket and in series of measurements at 0.1–0.8 m from the filter. Each measurement was made as the arithmetic mean of at least 10 consecutive readings so that the interval between individual measurements was at least 10 s. The background dose rate inside the service building was measured at a distance of at least 4 meters from the GAC filters. The background dose rate was subtracted from the measured mean value and the result was corrected using the calibration coefficient provided by Finnish Radiation and Nuclear Safety Authority (STUK), which is the national standard laboratory for radiation quantities.

Activated carbon samples were taken from the surface layer of each GAC filter through the service hatch to determine the Pb-210 concentration. In addition, GAC filters from Pekonniemi were sampled from the bottom layer through the filter drain valve.

The flow data for Pekonniemi were process data retrieved from the automation system's historical database, and for Alalampi, data from operating logs and reports were used. The process flow meters were magnetic flow meters connected to the automation systems, whose models were Proline Promag 10W (Endress + Hauser AG, Reinach, Swizerland) in Pekonniemi and Sitrans FM MAG3100/5000 (Siemens, Munich, Germany) in Alalampi. The accuracy of the flow measurements in the calculations was 1%, which was estimated to include both the internal error of the flow meter and the error caused by the process installation.

The amount of carbon installed, and the technical drawings of the filters were used to calculate the volume of the carbon bed and contact time. The sorting of the granules during rinsing can cause changes in the volume of activated carbon.

The concentration of radon in water and Pb-210 in activated carbon samples were analysed in the laboratory of the Finnish Radiation and Nuclear Safety Authority (STUK) in Helsinki. The laboratory is accredited according to the international standard EN ISO/IEC 17025:2017 (FINAS 2021).

Radon concentrations were analysed using the Guardian 1414 Liquid Scintillation Counter (EG&G Wallac, Turku, Finland) (Salonen & Hukkanen 1997). The uncertainties in the pulse rate of the sample and the calibration of the device have been taken into account in the results.

The activity concentration of Pb-210 in activated carbon was determined from the gamma energy of 46.5 keV measured by broad energy germanium detectors (Canberra Industies, Meridien, USA) and analysed by UniSAMPO Gamma Spectrum Analysis Software (Aarnio et al. 2001). The uncertainties include both statistical uncertainty and uncertainty due to the efficiency calibration. The method, which is modified from IEC 1452 (IEC 1995) standard, is documented in the quality manual of the laboratory.

In the Pekonniemi groundwater plant, the average radon concentration in raw water was 62.8 Bq/L (n = 14) and the range of individual measurements was 60.2–67.0 Bq/L. The adsorption efficiency of the new S835 activated carbon was 29–60% of radon and 54–75% with the new F300 carbon, but only 10–28% with the old S835 carbon (EBCT 20–60 min). In Alalampi, the radon concentration in raw water was slightly lower, 55.7 Bq/L (n = 9). Radon concentration in well #1 ranged 58.3–61.0 Bq/L and in well #2 49.2–53.4 Bq/L. The radon adsorption efficiency of the new AS2000 carbon was 37–52% and that of the old AS2000 only 4–12% (EBCT 11–12 min). The analyses of carbon grades S835 and AS2000 covered the entire service life of the activated carbon, but the analyses of the F300 only the beginning. The adsorption efficiencies measured on different days are shown in Appendix E.

The observed temporal deterioration of radon adsorption efficiency (Figure 1) was compared to the temporal trends observed in the adsorption efficiencies of other substances. The mean adsorption of trichloroethylene calculated from co-sampled pairs was 3.0 μg/L (reduction 58%) and non purgeable organic carbon (NPOC) 0.2 mg/L (18%). At the beginning of the carbon life 0.6–0.9 mg NPOC/L of organic matter was adsorbed, but after 1.5 years only <0.1 mg NPOC/L. Adsorption of other substances such as iron and manganese could not be shown statistically in water samples.

Trichloroethylene, iron, and manganese concentrations were analysed from samples of S835 carbon material taken during carbon exchange. Trichloroethylene was found only at the top of the filters 1.0 mg/kg, iron at the top 1,700 mg/kg and bottom 3,800 mg/kg, manganese at the top 14 mg/kg and bottom 91 mg/kg (Appendix B).

When modelling the accumulation of Pb-210, the specific weight of the carbon is an important variable which varies according to the water content of sample. Based on the weighing of the material samples in Pekonniemi, the activated carbon had sorted in the filters so that the specific weight of the surface samples is clearly lower than the weight of the samples taken from the bottom. Activated carbon is washed efficiently, and pumping is very rarely stopped, which explains the sorting of the material in the washing direction. In Alalampi the material samples were taken only from the surface of the carbon and their specific weight was clearly higher than specified by the manufacturer. The results show the sorting of the filter material in the flow direction. In Alalampi, the pumps are started and stopped daily. From this, pressure shocks are applied to the filter, which could explain the sorting of the filter material according to the flow direction. The sorted specific weights of the surface and bottom samples are taken into account in Figure 7.

The typical EBCT of activated carbon filters designed to remove solvents and pesticides is short, typically 10–30 minutes (Lykins et al. 1990; Crittenden et al. 2012), and the reduction of radon is only 40–60% (Lowry & Brandow 1985; Dixon & Lee 1988; Hiltebrand et al. 1988). Instead, several previous radon studies have focused on the elimination of very high radon concentrations from raw water with activated carbon. The reductions are usually required to be 90–99%, in which case the EBCT must be significantly longer than in this study, 70–120 min (Lowry et al. 1987).

The concentration of radon adsorbed from water to carbon is determined by Equation (3) (Section 2.4), where the EBCT is inversely proportional to flow rate. The influent flow of radon to the filter increases linearly with increasing flow rate or radon concentration. In this study, we found that the increase in flow rate produces more radon to the filter, than a decrease in adsorption rate reduces the rate of accumulation. Correspondingly, as EBCT increases, incoming radon flow decreases faster than the adsorption rate increases. Although the kinetics of the adsorption rate shown in Equations (1)–(2) (Section 2.4) have long been known (Lowry & Lowry 1987), the importance of this phenomenon is emphasized when calculating the accumulation of radon in activated carbon at short EBCT <60 min, which is especially evident in the radon accumulation rate in Figure 3.

In the case of new carbon (Figure 5), the adsorption takes place mainly at the point where the water enters the filter, which is also reflected in the measurements of the vertical gamma activity profiles of several other studies (Lowry & Brandow 1985; Turtiainen et al. 2000). The uneven distribution of radon in the activated carbon filter can be explained by the Half Lengths effect. If the first period of the filter binds 50% of the impurity (radon), the second period of equal length binds 25% and the third 12.5% of the initial concentration (WQA 2013). When the adsorption capacity of the carbon is impaired, detachment and re-attachment of nuclides smooth out the vertical distribution.

The K_SS constant describes the rate of radon adsorption on carbon. The K_SS of each measured carbon type began to decline immediately after the deployment, and the K_SS was halved at about 8000 BV (4–10 months, Figure 2). This has a significant effect on the magnitude of total Pb-210 accumulation. Adsorption of organic compounds, iron, manganese or other particles in activated carbon or the growth of biofilms have been suspected of causing to impair adsorption of the activated carbon (Kinner et al. 1990). Differences in hydraulic properties of the filter or influent water quality have been estimated in the literature to cause variations in K_SS values (Alabdula'aly & Maghrawy 2011). The temperatures of the investigated groundwater were +5– + 7 °C. Guo et al. (2017), in cold conditions the adsorption capacity of activated carbon is better than in warm conditions. In studies with long contact times a decrease in K_SS is more difficult to detect than with short (EBCT <60 min) contact times of this study.

With the exception for compounds to be removed, at both water plants, the raw water was of good physico-chemical quality and the amount of organic matter was low (Appendix B). The microbiological quality of activated carbon was not studied separately, but some estimates can be made based on the change in organic matter concentration. Velten et al. (2011) has shown that the growth of biofilm consumes a lot of dissolved organic carbon (DOC) at the beginning of the activated carbon use, but after the inflection point, DOC is consumed only for the maintenance and vital functions of the biofilm. NPOC results are available for the organic matter content at Pekonniemi. In the months after the changing of activated carbon, 0.6–0.9 mg NPOC/L of organic matter was retained in the filter, but <0.1 mg NPOC/L after the inflection point (1.5 years). This was much less than the 0.3–0.5 mg DOC/L reported by Velten et al. (2011). In Finland, phosphorus has been shown in many cases to be the limiting nutrient for biofilms of the water supply network (Miettinen et al. 1997), so phosphorus scarcity may explain the difference in results. In Middle-Finland, groundwater is usually low in phosphorus (Lahermo et al. 2002), at Pekonniemi 9.0 μg/L.

Radon adsorbed on the GAC filter decays into long-lived Pb-210 through four short-lived decay products. Two of these, Pb-214 and Bi-214 emit relatively intense gamma radiation outside the filter. The magnitude of the dose rate is determined by the concentration of radon in carbon, the measuring distance, and the properties of the media.

As a result of the self-attenuation of gamma radiation by the water-carbon mixture, about 90% of the radiation measured from outside the filter is generated at distance of less than 30 cm from the filter jacket. If the flow rate slows down near the jacket compared to the middle area of the filter, the concentration of radon in carbon near the jacket is also lower than in the middle of the filter. In case of uneven distribution, the radon concentration of the filter cannot be assessed by external dose measurement.

Outside the filter, the dose rate decreases rapidly, at working distance, the dose rate is only half of its maximum value (Figure 6). In Pekonniemi, the maximum dose rate of 0.75–1.25 μSv/h with F300 carbon was measured from the working distance (C₀ = 62.5 Bq/L, EBCT 23 min). If the radon concentration in the raw water is high (400 Bq/L) and the contact time is short (EBCT 10 min), the dose rate could increase to 30 μSv/h on the outer surface of the filter jacket and at a working distance of 0.5 m 5–10 μSv/h. As the adsorption rate of radon decreases during the use of activated carbon, the external dose rates also decrease. In Pekonniemi, both the measured radon accumulation rate and the measured dose rate both decreased to approximately one-fifth of the original after three years of use.

Water plant maintenance personnel do not work long periods of time near the GAC filter. Typical tasks are carbon change, backwash (an automated process), possible repairs, and water sampling. The dose received by the worker can be estimated based on the dose rate, working distance, and time. If the working distance is 0.5 m and the working time is 8 h/a, the occupational dose from gamma radiation in Pekonniemi would be about 8 μSv/a. At the maximum dose rate of 5–10 μSv/h, the annual dose would be 40–80 μSv/a. This dose does not exceed the usual reference value for occupational exposure to natural radioactivity (1 mSv/a excluding radon in indoor air) and can be compared, for example, to the annual dose of 50 μSv/a received from air travel (STUK 2020). Based on the results, working of several days near GAC filters should be avoided if the radon concentration in the raw water exceeds 200 Bq/L and the activated carbon is less than one year old. When designing new water plants, the radiation risk must be taken into account in the design of work areas and maintenance bridges.

Activated carbon accumulates Pb-210, which is produced by the radioactive decay of radon adsorbed on carbon. In activated carbon samples taken from Pekonniemi, the concentrations of Pb-210 in the surface samples were more than twice as high as in the samples taken from the bottom (Appendix F). A similar uneven accumulation of radon can be observed in the water samples taken from different parts of the filter, and in the radiation dose rate profiles in Figure 5.

During the long period of use, the accumulated radon in the activated carbon is practically completely decayed into Pb-210, and some of the formed Pb-210 has had enough time to decay further into the stable isotope Pb-206. In Pekonniemi, the average accumulation of Pb-210 calculated based on water samples was practically equal as the average of the material samples. Although the log-logistic model used for the K_SS deterioration had to interpolate a large BV range, the result was in good agreement with the measured one.

In the GAC filter, detachment and reattachment of both radon and Pb-210 atoms occur, which increase as the filter material ages. The phenomenon is described by the K_SS constant. For radon, this can be seen in Figure 5 as a maximum point shift and smoothing as the K_SS decreases. Unlike radon, Pb-210 is chemically highly reactive in the aquatic environment. The main explanation for the solubility of lead is pH, and the most significant change in solubility occurs between pH 6.0 and 7.0 (Sigworth & Smith 1972). Lead precipitates at pH 8–11 into lead oxides and carbonates. At pH <8, part of the lead dissolves into lead ions, but in the pH range 5–8.5 lead can also be present as carbonates if there is enough carbon dioxide in the water (Naylor & Dague 1975). In Finland, groundwater is typically slightly acidic. The pH values of the raw water in Pekonniemi and Alalampi are 6.4 and 6.2, so the state of lead depends on the pH-carbon dioxide balance. The small difference between the calculated and measured Pb-210 in the samples (Figure 7) shows that, at least in Pekonniemi, there is no significant release due to the dissolution of lead. Some studies have reported that activated carbon removes more than 90% of Pb-210. The result suggests very good retention of Pb-210, but measurements have been reported in the higher pH range of 7.1–8.1 (Turtiainen et al. 2000) than in this study (pH 6.2–6.4).

When activated carbon is used for several years, the concentration of Pb-210 in the filter may exceed the NORM clearance level of 1 kBq/kg. As stated at the outset, the limit values for the disposal of Pb-210 in landfills differ from one country to another (Garcia-Talavera et al. 2021). Similarly, in some countries the dry weight of the sample is used in the clearing level comparison (SSMFS 2018) and in others, such as Finland, the weight of the sample in the moisture content of the treatment. Predictive models of Pb-210 reported in the literature have generally been calculated for dried activated carbon (Martins 1992; Lewis & Houle 2008) which should be considered when comparing the models.

Figure 8 shows two forecasts of Pb-210 accumulation. The specific weight of the drained dried carbon (880 kg/m³) has been used as the weight of the material in the forecasts, and the forecasts can be scaled to different water contents. With the normal 4–5-year service life of activated carbon, the risk of exceeding the clearance level increases if the radon concentration in the raw water exceeds 100 Bq/L. The accumulation of Pb-210 in the Pekonniemi GAC was faster than in the Alalampi GAC, based on both water and material samples, so the difference in accumulation rates is real. The differences are explained by the quality of raw water, the quality of the carbon (K_SS), the contact time (EBCT) and the possible release of Pb-210 over the long period of use. Attention should be paid to the radon concentration and pH of raw water and the changes in the K_SS value of activated carbon (Figure 2). The data of the Pekonniemi covered the entire life cycle of carbon better than the data of Alalampi, which may also explain the difference in forecasts.

The main challenge in preparing a Pb-210 accumulation forecast is related to the uncertainty of the radon measurement results. Analyses of water samples were performed in STUK's laboratory, which is the standard laboratory for radiation measurements in Finland. The results take into account statistical uncertainties in the count rate of ionizing radiation and the calibration of the measuring instrument. Uncertainties were increased by the delivery time of the samples to the laboratory, which ranged from 2 to 5 days. Especially with older activated carbons for which a low (<5%) radon adsorption efficiency is calculated, the uncertainty of the result may be significant (Appendix E).

Another noteworthy source of uncertainty is the accuracy of flow measurements in water utilities. The measuring equipment is factory calibrated, but the best measuring accuracy is only achieved within a certain flow rate range specified by the manufacturer. As the groundwater pumps in both plants are operated on/off, it can be estimated that there are no slow flows that impair the measurement accuracy. The measurement accuracy can also be impaired by the flow conditions at the installation site and the processing of the measurement signal in the data transmission and reporting system.

The technical dimensional accuracy of the filter equipment, and in particular the actual amount of carbon in the filter and any channelling of the flow, may impair the accuracy of the derived results. The volume of carbon is not an exact measurand and may vary during use, for example due to carbon sorting.

The accumulation model for radon and Pb-210 is particularly affected by the adsorption rate of carbon and its deterioration during use. K_SS is a variable that depends on the quality of carbon and water at a particular point in time. Several measurements of radon concentration (and thereby K_SS value) in water are needed for the accumulation estimates for the entire service life.

The concentration of radon or Pb-210 in the carbon is expressed in relation to the specific weight, which depends on the water content of the sample. It is important to note whether the results are reported in completely dry, drained, or wet weight of the carbon.

Modelling of Pb-210 accumulation leaves uncertainty about the long-term dissolution of Pb-210, which may affect the final Pb-210 concentration.

In waterworks using groundwater from alluvial aquifer, gamma radiation outside activated carbon filters is unlikely to pose a significant radiation protection problem. However, if the raw water contains more than 200 Bq/L of radon, it is recommended to measure the dose rate approximately 3 weeks after the addition of activated carbon and to assess the occupational radiation safety risk and to avoid prolonged work near filter. After about a year of use, the dose rate decreases significantly.

The concentration of Pb-210 retained in activated carbon may exceed the clearance level of 1 kBq/kg for NORM materials. Pb-210 was found to be unevenly distributed in the filter material. When typical usage times of carbon batches are considered, we can conclude that the risk of exceeding the NORM clearance level increases significantly when the concentration of Pb-210 in raw water exceeds 100 Bq/L. If the clearance level is exceeded, the owner of the waste is responsible for notifying the competent authority, and for correct treatment and disposal of the carbon waste.

The national radiation authority and the waste management authority should jointly provide guidance on the final disposal of carbon waste used for groundwater treatment, so that proper treatment of the waste is possible, and the procedure is known to the waste owners.

The authors wish to thank Dr Niina Leikoski for her help in sampling the activated carbons and Dr Antti Kallio for analyzing the gamma-ray spectra of these samples. The authors also want to acknowledge Mrs Tarja Heikkinen's great work on radon measurements in water samples, and Keuruu waterworks and water treatment team at Alva-yhtiöt Ltd for making this study possible.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.