Variation in Radon Concentration Between Apartments in Housing Cooperatives

Abstract

Housing cooperatives are a common form of housing in Nordic countries, being tasked with responsibilities such as maintenance, renovation, and, when needed, radon mitigation. This study analyzed the radon level variation in nearly 16,000 apartments across 3552 housing cooperatives. The analysis explored how radon levels varied based on the number of measurements conducted within each cooperative, assuming that apartments sharing the same plot address belong to the same cooperative. The radon concentrations in the apartments of the cooperative typically followed a log-normal distribution. The geometric standard deviation (GSD) specific to each housing cooperative varied considerably. The median GSD ranged between 1.5 and 2.0, depending on the number of apartments measured. A predictive model was developed to estimate the likelihood of apartments exceeding the radon reference level based on the housing cooperative’s geometric mean radon concentration. The results highlight the importance of measuring radon levels in all apartments within housing cooperatives to ensure radon safety. Additionally, the model offers support for housing cooperative decision-makers and epidemiological studies, helping to address uncertainties and to account for spatial variations in radon exposure.

1. Introduction

Finland is one of the European countries where indoor radon levels often exceed the EU reference level of 300 Bq/m³ in many regions [1]. This is primarily due to the presence of granitic rocks with high uranium content and permeable glacial soils like gravel and sand, which allow radon to migrate into buildings. The common ground-supported floor structure further facilitates radon infiltration without proper sealing [2]. Additionally, long, cold winters create a pressure gradient that draws and accumulates radon indoors, as airtight construction on walls and ceilings and minimal ventilation are used to reduce heating costs [3]. However, a survey of Finnish new ground-contact dwellings (n = 1332) found that the airtightness (q₅₀-value) of a building does not correlate with its indoor radon levels. Instead, radon levels are most influenced by the type of foundation and the implementation of radon prevention measures during construction [4].

The concentration of radon in buildings varies both temporally and spatially. Temporal variation is influenced by factors such as building automation systems, the activities of building occupants, and weather conditions [5,6,7]. Local variations in radon concentration within a building are not only primarily affected by the same factors but also by potential unsealed areas in the building’s foundations, which can allow radon to seep into specific spaces [8].

The spatial variation in radon concentrations in dwellings has been widely studied. Typically, this variation is examined within the same dwelling using ratios, such as by comparing the second floor to the ground floor or bedrooms to basements, and so on [9,10,11]. A Finnish survey found that ground-floor flats in multi-story buildings have radon levels similar to those in detached and terraced houses, while upper floors have significantly lower levels, averaging 44 Bq/m³. According to the authors, the radon in upper floors mainly originates from building materials, with less variation than radon entering from the ground [12]. Additionally, variations in radon concentrations have been studied across dwellings or workplaces within the same area or region [13,14,15]. However, fewer studies have been published on radon concentration variations between apartments located on the same plot of land. The variation in radon concentration between apartments has, however, been recognized and is also taken into account; for example, in the radon measurement protocol for apartments published by the Swedish authorities [16].

In Nordic countries, housing cooperatives (asunto-osakeyhtiö in Finnish; bostadsrättsförening in Swedish) are common. The purpose of a housing cooperative is to own and manage at least one building or part of a building, typically located on a single plot of land. Residents purchase shares in the cooperative, which grant them the right to occupy a specific dwelling. The dwelling itself can be a standalone house, a semi-detached house, or a flat in a terraced house or a multi-story building. From now on, all dwellings sharing the same plot address will be referred to as apartments. The apartments are technically owned by the cooperative, which retains decision-making authority over matters such as building maintenance and renovations. The shareholders are generally responsible only for maintaining the interior surfaces of their homes.

If a radon measurement approved by the competent authority in an apartment within a housing cooperative reveals a radon concentration exceeding the reference level (300 Bq/m³ in Finland), the shareholder is not permitted to independently carry out radon mitigation. Instead, the responsibility for mitigation lies with the cooperative. If the cooperative’s board is unwilling to address the issue, health authorities in Finland can mandate the cooperative to investigate and eliminate the health hazard, i.e., to perform the necessary measurements and undertake radon mitigation.

The Radiation and Nuclear Safety Authority (STUK) in Finland maintains a national radon registry that contains radon measurement results for approximately 179,000 dwellings and 30,000 workplaces. The results in the register for dwellings have been conducted using STUK’s own integrating radon detector, while for workplaces, some of the results have been obtained using detectors from other laboratories. Examining the registry data reveals that there are housing cooperatives where only part of the apartments has been measured, identifiable; for example, by terraced house apartment identifiers (e.g., 3A, 3C). The registry also shows that the ordering and payment for measurements were often carried out by individual residents rather than by the housing cooperative. In such cases, there is a clear risk that individual shareholders have measured the radon concentration in their own apartments and that other shareholders have estimated the radon levels in their apartments based on an extrapolation of their neighbor’s results.

The study aimed to investigate the variations among ground-floor apartments of housing cooperatives. A direct benefit of this is in the planning of measurement campaigns and in motivating residents to conduct appropriate measurements. The results also provide additional information on uncertainties related to, for example, the radon concentration modeling used in epidemiological studies [17]. In radon concentration modeling, the radon level in an apartment is typically assessed using parameters such as regional radon levels, soil type, building type, year of construction, building frame materials, foundation type, and ventilation type. In this study, apartments within the same housing cooperative share nearly identical model parameters, allowing the observed variation in their radon concentrations to illustrate uncertainties in modeling.

2. Materials and Methods

2.1. Selection and Processing of Measurement Results

All household measurements included in the national radon register and, therefore, also the results used in this study have been obtained using STUK’s own integrating alpha track detector, Radonpurkki [18]. The measurements are primarily ordered by residents or housing cooperatives, but the register also includes free measurements conducted by STUK as part of its radon surveys. Since most of the measurements are ordered by residents or housing cooperatives, they are regionally concentrated in radon priority areas of Finland. Often, the name of the housing cooperative has not been recorded in the registry—only the name of the apartment’s occupant—so apartments with the same plot address (e.g., Lake Street 3 A 10 and Lake Street 3 F 4) are assumed to belong to the same housing cooperative, as they are at least located on the same plot. In rare cases, the plot may have been divided between two or more privately owned detached houses under a joint ownership agreement. Since one homeowner might estimate the radon concentration of their house based on their neighbor’s measurement, such cases were included in the analysis. From now on, the building or buildings at the plot address will be referred to as a “housing cooperative”, even though this may not be accurate in all cases.

To quantify spatial variations based on the radon measurement results, the dataset was filtered using the following criteria:

• Measurements in the same apartment had to start within 30 days of each other.
• The measurement duration had to be at least 40 days.
• Measurements had to be conducted during the measurement season, i.e., between September 1 and May 31.
• Measurements in the same housing cooperative had to start within 100 days.
• If an apartment had been measured with two or more detectors, the geometric mean of the results was used for analysis.

The first four filtering criteria aimed to minimize the impact of weather conditions on the measurement results, while the fifth criterion ensured that all apartments were given equal weight in the analysis. The third criterion was established because the difference between two- or three-month measurements and the annual mean radon concentration is significantly smaller during the measurement season than in the summer months. In other words, the variability of the results is smaller during the measurement season [19]. The fourth criterion may have been too lenient, as in 91% of cases, measurements in different apartments were started within 30 days of each other. The final analysis included 15,893 apartments from 3552 housing cooperatives. The measurements cover the period from 1983 to 2020, with most of the measurements carried out after 2005.

Unfortunately, it was impossible to determine on which floor the apartment was located in a multi-story building. In Finland, radon measurements have traditionally focused on apartments on ground-level floors, as the radon emission from building materials has not been known to cause exceedances of reference levels in Finland (cf. Sweden’s blue concrete houses). The data may include a few individual apartments from the upper floors of multi-story buildings that were measured as part of national radon surveys, but these are not relevant for this analysis due to their sparsity. All the results of this study can therefore be considered to pertain solely to the variation in radon levels in ground-contact apartments.

2.2. Statistical Key Figures of the Data

Statistical key figures and the distribution of the results were calculated from the data. Subsequently, housing cooperatives with >20 simultaneous measurements were selected for separate analysis. In these housing cooperatives, the Shapiro–Wilk test was used to examine the distribution of radon concentrations or their geometric means in individual apartments. Based on the distribution testing, it was concluded that the housing-cooperative-specific data follows a log-normal distribution. Therefore, the geometric standard deviation (GSD) was chosen as the key figure to describe the variability in radon concentrations within the housing cooperatives, and the geometric mean (GM) was selected as the key figure to represent the radon concentration in a housing cooperative.

2.3. Factors Affecting Concentration Variability

The next step was to examine whether the housing-cooperative-specific radon concentration affects the variability of radon concentrations, i.e., the GSD, which is, by definition, a relative indicator. First, housing cooperatives with a GM of 20 Bq/m³ or less were excluded from the dataset. The detection limit of the measurement method is approximately 13 Bq/m³ for a two-month measurement period, meaning that the lowest measurement results in these housing cooperatives are likely to be below the detection limit and may have been recorded, for example, as half the detection limit value. Therefore, there is a high degree of uncertainty associated with the housing-cooperative-specific GSD when analyzing low radon concentrations.

Random uncertainty associated with the measurement method introduces variability into the results. This random uncertainty is caused by factors such as the uneven quality of the alpha-track detector material and its housing, the number of observed alpha tracks, variations in etching solutions and conditions, etc. All of these factors contribute to an increase in the GSD of the data set. To determine the actual GSD of radon concentrations (GSD_C), the GSD caused by the measurement method (GSD_D) must be subtracted from the observed GSD (GSD_O) using Equation (1). The simulation of variability caused by the measurement method is described in Appendix A.

$${GSD}_C=e^{\left[\ln \left({GSD}_O\right)-\mathrm{l}\mathrm{n}\left({GSD}_D\right)\right]}$$

(1)

The lower the measured radon concentration, the more the measurement technique increases the observed GSD value. The number of measured apartments also increases the observed GSD value, but the effect is clearly smaller than that for the concentration. For this reason, the measurements in the data were categorized not only by radon concentration but also by the number of measurements.

After it was demonstrated that the average radon concentration in the housing cooperative seems not to affect the value of GSD caused by the radon concentration variability, the data could be categorized based on the number of measured apartments. The categories were formed in such a way that each category would include at least one hundred housing cooperatives. In this analysis, it was not possible to subtract the variability in concentrations caused by the measurement technique, as the housing cooperatives grouped by the number of apartments measured exhibited significant differences in their radon concentration. Within each category, the GM of the category’s radon concentration was calculated, and the distribution of the observed GSD values within the category was examined by calculating the median and the 5th, 16th, 25th, 75th, 84th, and 95th percentiles.

2.4. Modeling of Variability

When examining the distributions of category-specific GSD values, it was observed that they did not follow any commonly used distribution, such as the normal or log-normal distribution. Therefore, the distribution of GSD values in each category was modeled using a modified Burr distribution’s quantile function.

$$GSD=1+h·{\left({\displaystyle \frac{1}{p}}-1\right)}^{{\displaystyle \frac{1}{-b}}}$$

(2)

where h and b are the shape parameters of the function, and p is the probability. The fitting was performed for the 5th, 16th, 25th, 50th, 75th, 84th, and 95th GSD percentile values (cumulative probability values) using the least-squares method and the MS Excel GRG Nonlinear engine.

The shape parameter values of each category were further modeled with a logarithmic function based on the number of measured apartments. As a result, two equations were obtained that could predict the values of h and b when the number of measurements is known and, thereby, the probability distribution of GSD values for each category.

2.5. Utilizing the Model

A total of 68,000 random numbers between 0 and 1 (probability p in Equation (2)) were simulated, from which the same number of GSD values were calculated for each category of 2–20 measured apartments, using the h and b values from the model. Then, a test concentration—representing the geometric mean of the housing cooperative—was input into the simulation. Using this geometric mean, the simulated GSD, and the random probability (number 0–1), the radon concentration of a randomly selected apartment within the category was calculated using the inverse of the log-normal distribution function using MS Excel. The number of simulated radon concentrations exceeding 300 Bq/m³ were counted, and the probability of an apartment exceeding the reference level was calculated based on the input geometric mean of the radon concentration and the category of measured apartments.

Each simulation was repeated six times, ensuring that each simulated probability was based on more than 400,000 calculations, reducing variability to the fourth significant figure. Simulations were repeated by inputting test concentrations (the geometric mean of radon concentrations in the housing cooperative’s apartments) ranging from 20 to 300 Bq/m³.

After this, the probability that one or more apartments in the housing cooperative has a radon concentration exceeding the reference value (P_exc) was calculated. This was most easily computed as the complement of the event in which none of the measurement results exceed the reference value. Specifically:

$$P_{exc}=1-{\left(1-p_s\right)}^N$$

(3)

where p_s represents the apartment-specific probability that the reference level is exceeded, and N is the number of apartments measured.

3. Results

3.1. Statistical Key Figures of the Data

The dataset analyzed here was heavily weighted toward high radon concentrations, as expected. The radon concentrations in individual apartments or the geometric means of apartment-specific radon concentrations (n = 15,893) followed a log-normal distribution. The geometric mean was 169 Bq/m³, and the geometric standard deviation was 3.1.

The distribution of the number of apartments per housing cooperative was strongly right-skewed, with housing cooperatives consisting of two or three apartments accounting for half of the apartments studied.

The distribution of apartment-specific radon concentrations (C) could be examined in greater detail in those housing cooperatives where >20 measurements had been conducted. The Shapiro–Wilk test for the logarithms of the apartment-specific radon concentrations in a given cooperative did not deviate statistically significantly from a normal distribution. Therefore, apartment-specific radon concentrations or their geometric means almost invariably followed a log-normal distribution. Based on this, it could be concluded that the GSD is a suitable variable to describe the variability in concentrations between apartments (Figure 1).

3.2. Factors Affecting Concentration Variability

To determine whether there is a correlation between the housing cooperative’s general radon concentration and the GSD of the concentration, the housing cooperatives were divided into categories based on the average radon concentration and the number of measured apartments. Only in cases where measurements were conducted simultaneously in 2–6 apartments were there sufficient numbers of observations to study the correlation.

The results left room for interpretation (Appendix B,

Table A2. Spearman correlation analysis between the radon concentration and its variation within housing cooperatives. The GM range refers to the range of geometric means of radon concentrations for housing cooperatives, with C indicating the median GM in each category. N represents the number of housing cooperatives analyzed in each category, and GSD denotes the median geometric standard deviation within these categories.

GM Range	No. of Measured Apartments
	2			3			4			5			6
	N	C	GSD	N	C	GSD	N	C	GSD	N	C	GSD	N	C	GSD
20–40 Bq/m3	93	31	1.41	49	30	1.38	26	32	1.45	18	31	1.33	7	31	1.47
40–60 Bq/m3	133	50	1.51	55	49	1.55	34	49	1.72	9	44	2.12	9	54	1.59
60–100 Bq/m3	239	80	1.48	79	80	1.63	64	81	1.64	33	79	2.00	20	79	1.88
100–150 Bq/m3	234	123	1.50	96	125	1.62	49	114	1.64	39	128	1.69	25	123	1.89
150–200 Bq/m3	191	171	1.48	70	173	1.72	49	176	1.73	36	171	1.59	25	172	1.68
200–300 Bq/m3	258	243	1.49	93	236	1.58	62	243	1.77	53	244	1.69	34	250	1.63
300–500 Bq/m3	176	374	1.49	81	370	1.63	65	372	1.66	51	382	1.52	31	385	1.87
500– Bq/m3	148	796	1.66	68	698	1.73	32	755	1.75	18	731	1.77	22	851	1.67
r(6)	0.429			0.786			0.683			–0.0479			0.310
P	0.315			0.034			0.062			0.910			0.501

). Spearman correlation analyses showed no statistically significant correlations in cases with 2 or 4–6 measured apartments. A statistically significant correlation (p = 0.034) was observed only when three apartments were measured, but the p-value was close to the 0.05 significance threshold. Additionally, the case where five apartments were measured showed a negative, non-significant correlation between the GM and GSD, while the others showed a positive, mostly non-significant correlation. These findings suggest that the observed correlation in the case of three measurements may be due to chance.

The dataset was subsequently categorized solely by the number of measured apartments. Housing cooperatives with more than seven measured apartments were grouped into aggregate categories to ensure an adequate number of observations per category. Despite this grouping, the radon concentration remained consistent across all the apartment number categories, with the median GMs ranging from 153 to 232 Bq/m³, enabling meaningful comparison of the measurement results.

A Spearman correlation analysis revealed a statistically significant strong positive correlation between the number of measured apartments and the median GSD of the housing cooperative (p < 0.001).

3.3. Modeling of Variability

The 5th, 16th, 25th, 50th, 75th, 84th, and 95th percentiles of GSDs in each category are presented in

Table 1. Distribution of housing-cooperative-specific GSDs for different numbers of measured apartments (N). M(GM) indicates the median geometric mean radon concentration in Bq/m³ in each category. Parameters h and b are associated with the model. The GSD distributions are provided in percentiles; for example, P5(GSD) is the 5th percentile of GSD values in the category. The figures in parentheses are based on the fitted model.

Category	2	3	4	5	6	7	8–9	10–14	15–
GM(N)	2	3	4.0	5	6	7	8.4	11.5	19.4
M(GM)	153	157	161	195	196	232	189	193	184
GM(GSD)	1.60	1.73	1.80	1.79	1.86	1.86	1.86	1.92	2.04
h	0.42	0.59	0.69	0.70	0.74	0.77	0.80	0.86	0.99
b	1.55	1.97	2.31	2.33	2.42	2.69	3.16	3.14	3.64
P5(GSD)	1.03 (1.06)	1.11 (1.13)	1.19 (1.19)	1.21 (1.20)	1.25 (1.22)	1.34 (1.26)	1.30 (1.31)	1.39 (1.34)	1.49 (1.44)
P16(GSD)	1.11 (1.14)	1.23 (1.26)	1.32 (1.34)	1.35 (1.34)	1.37 (1.37)	1.45 (1.41)	1.44 (1.47)	1.52 (1.51)	1.62 (1.63)
P25(GSD)	1.17 (1.21)	1.31 (1.34)	1.40 (1.43)	1.42 (1.44)	1.53 (1.47)	1.54 (1.51)	1.53 (1.56)	1.59 (1.60)	1.73 (1.73)
P50(GSD)	1.41 (1.42)	1.58 (1.59)	1.68 (1.69)	1.70 (1.70)	1.77 (1.74)	1.75 (1.77)	1.80 (1.80)	1.78 (1.86)	1.96 (1.99)
P75(GSD)	1.87 (1.85)	2.04 (2.04)	2.14 (2.11)	2.17 (2.13)	2.15 (2.16)	2.07 (2.15)	2.20 (2.13)	2.26 (2.22)	2.35 (2.34)
P84(GSD)	2.26 (2.22)	2.44 (2.38)	2.44 (2.42)	2.40 (2.43)	2.36 (2.47)	2.40 (2.42)	2.36 (2.35)	2.45 (2.46)	2.52 (2.56)
P95(GSD)	3.78 (3.79)	3.63 (3.65)	3.46 (3.48)	3.50 (3.49)	3.54 (3.49)	3.33 (3.29)	2.99 (3.02)	3.20 (3.20)	3.24 (3.22)

. The parameters h and b of Equation (2) were easy to obtain by fitting Equation (2) to the observed values. The fit corresponded well to the measured values (Appendix C, Figure A1).

The fitted values of h and b (Table 1) can be further described using a logarithmic equation as a function of measured apartments (Figure 2). This allows their values to be predicted using the following equations:

h = 0.2273 × ln(N) + 0.3229

(4)

b = 0.9209 × ln(N) + 0.9335

(5)

where N is the number of apartments.

Figure 3 shows the prediction intervals provided by the model, as well as the intervals based on the observations. The model is indeed able to predict the variation in radon concentrations of apartments in housing cooperatives quite reliably.

3.4. Utilizing the Model

The apartment-specific probability of exceeding the radon reference level is quite small (<7%) when the housing cooperative’s radon concentration GM is 100 Bq/m³ or lower (

Table 2. Apartment-specific probabilities that the radon concentration exceeds the reference level of 300 Bq/m³ when the geometric mean (GM) of the measurements is known and different numbers of apartments have been measured.

GM(C)	No. of Measured Apartments
	2	3	5	8	12	16	20
20 Bq/m3	0.5%	0.3%	0.3%	0.2%	0.2%	0.2%	0.2%
40 Bq/m3	1.1%	0.9%	0.8%	0.8%	0.8%	0.8%	0.9%
60 Bq/m3	2.0%	1.7%	1.6%	1.7%	1.9%	2.1%	2.2%
100 Bq/m3	4.2%	4.2%	4.6%	5.2%	5.8%	6.4%	6.8%
140 Bq/m3	7.2%	7.6%	8.6%	10%	11%	12%	13%
200 Bq/m3	14%	16%	18%	20%	22%	23%	24%
240 Bq/m3	22%	24%	27%	29%	30%	31%	32%
280 Bq/m3	39%	42%	43%	44%	45%	45%	46%

). However, the situation changes when viewed from the perspective of the housing cooperative rather than the individual apartment dweller (

Table 3. Probabilities (%) that there are one or more apartments in the housing cooperative where the reference level of 300 Bq/m³ is exceeded. The probabilities are categorized based on the housing cooperative’s average radon concentration (GM(C)) and the number of measurements taken in the cooperative.

GM(C)	No. of Measured Apartments
	2	3	5	8	12	16	20
20 Bq/m3	1.0%	1.0%	1.2%	1.7%	2.4%	3.2%	4.0%
40 Bq/m3	2.2%	2.6%	3.8%	5.9%	9.1%	13%	16%
60 Bq/m3	3.9%	5.1%	8.0%	13%	21%	28%	36%
100 Bq/m3	8.3%	12%	21%	35%	51%	65%	75%
140 Bq/m3	14%	21%	36%	57%	76%	87%	93%
200 Bq/m3	27%	41%	63%	84%	95%	98%	100%
240 Bq/m3	39%	57%	79%	93%	99%	100%	100%
280 Bq/m3	63%	80%	94%	99%	100%	100%	100%

). At a concentration level of 100 Bq/m³, housing cooperatives already face a high risk that one or more apartments will have excessive radon levels, especially in larger cooperatives. For example, in a housing cooperative with 16 apartments and a radon concentration GM of 100 Bq/m³, there is already a two-thirds probability that one or more apartments will exceed the reference value.

4. Discussion

As stated in the introduction, the variation in radon concentration between apartments in housing cooperatives has scarcely been studied. However, there are numerous studies examining the variation in indoor radon concentrations within a single apartment. These studies typically calculate a ratio; for example, between the living room and the basement, the upper and lower floors, or two rooms on the same floor [20,21,22]. From one dataset, one distribution of ratios and one GSD value can be obtained. In those studies, the GSD values have ranged between 1.38 and 1.78. This study examined separate housing cooperatives and the GSD values of the concentration distributions within them rather than ratios, making comparisons with apartment-specific studies challenging.

In contrast, the distribution of regional radon concentrations has been studied; for example, using 10 × 10 km map grids or regional sampling surveys investigating radon concentrations. According to a summary by Yarmoshenko et al., the median GSD for ground-floor residential rooms within 10 × 10 km map grids in Europe is 1.85 [21]. ICRU Report 88 provides a comprehensive summary of the spatial variation within 10 × 10 km map grids across 15 European countries [23]. The reported country-specific GSD values typically ranged from 1.7 to 2.2, with a calculated median of 1.91 across all countries. In this study, the median GSD for housing cooperatives, ranging from 1.41 to 1.96, depended on the number of measured apartments. In large housing cooperatives, where eight or more apartments were measured, the variation is therefore of the same order of magnitude as the average within the 10 × 10 km map grids.

It was easy to create a model for the radon data from housing cooperatives, based on which the variations in radon concentration between apartments in a cooperative can be predicted. However, the model was created using data that were clearly biased towards Finland’s radon priority areas, as the GM of all the measurements was 167 Bq/m³, whereas the arithmetic mean concentration in all Finnish dwellings is estimated at 94 Bq/m³ [24]. However, the GSD does not seem to depend on the radon concentration in the housing cooperative, so it could potentially be applied elsewhere in Finland. However, the results in this regard were not entirely clear.

Housing cooperatives with a GM of radon concentration below 20 Bq/m³ were excluded from the dataset. This was because the detection limit for radon concentration with STUK’s own detector and a two-month measurement is approximately 13 Bq/m³, meaning that a significant portion of the measurements in these cooperatives fell below the detection limit, and the GSD value could not be reliably calculated. As noted in Section 3.2, the GSD of radon concentrations in the apartments of a housing cooperative does not appear to depend on the cooperative’s average radon level. Therefore, the model should be applicable to housing cooperatives with radon concentrations below 20 Bq/m³, but this could not be confirmed during the study. On the other hand, the risk of exceeding the reference level in one or more apartments in these cooperatives is small.

Regulatory requirements and construction practices related to buildings vary by country. In Finland, for example, the commonly used geomembrane for radon prevention is not utilized; instead, sealing is typically carried out using bitumen felt. Furthermore, the base-floor construction of a building is typically constructed with a foundation wall (either cast concrete or block structure) and a separately cast ground-supported slab. This construction method can create a leakage point between the slab and the foundation wall if sealing with the bitumen felt fails or is omitted. Therefore, the results of this study may not be directly applicable in countries where foundation structures differ significantly from the Finnish construction approach.

An interesting observation is how the risk of exceeding the reference level is very different when considering the risk from the perspective of an individual resident or the entire housing cooperative. Especially in large cooperatives, even at fairly low radon levels (<100 Bq/m³), there is a considerable risk that the reference value for the radon concentration will be exceeded in some apartments. The housing cooperative is obligated to carry out radon remediation in apartments, as the shareholders are primarily responsible for the renovation and maintenance of the apartment’s surfaces. For this reason, it is important that every apartment in a housing cooperative is measured and that results are not extrapolated from the measurements of every second or third apartment.

A shortcoming of the model is, naturally, that a housing cooperative’s decision may rely on one or two radon measurements, potentially representing minimum values. The cooperative’s GM value remains unknown until all apartments are measured, and the model cannot predict how many apartments in the housing cooperative have been left unmeasured. That would have been valuable baseline information and would have refined the focus so that the results would not need to be interpreted based on the number of measured apartments but rather on the actual number of apartments in the housing cooperative.

As mentioned earlier, the model developed in this study can mainly be used to assess the variation in radon levels in ground-contact apartments. Based on previous studies, it can be concluded that, for example, in a multi-story residential building, radon levels are highest in the ground-floor apartments in contact with the soil, while radon levels are lower and more consistent in upper floors. However, there are few results on radon levels in upper floors in Finland. It is possible that there are variations, and that radon levels may even exceed reference levels if, for instance, radon can be transported to upper floors through an elevator shaft or if there is a particularly poorly ventilated room in the building from which radon can leak into an adjacent apartment.

Many epidemiological studies are based on modeled radon exposure according to factors like the year of construction, building type, and location. Based on this analysis, it can be concluded that radon concentrations vary considerably even between apartments built on the same plot of land. This uncertainty can now be better accounted for in modeling.

5. Conclusions

The radon concentrations in ground-contact apartments of housing cooperatives follow a log-normal distribution. Therefore, the best descriptor of the variation in concentrations is the GSD, which is, by definition, multiplicative and, thus, relative. In the dataset analyzed, the GSD did not appear to depend on the measured radon concentration once the variation caused by the measurement technique was accounted for. However, the variation did depend on the number of apartments measured, with greater variation observed when more apartments were included. For 2 measured apartments, the median GSD and its interquartile range were 1.41 and 1.11–1.87, respectively, whereas for measurements involving more than 15 apartments, the corresponding values were 1.96 and 1.62–2.35.

According to the study, each ground-contact apartment in a housing cooperative must be measured, as the radon concentrations vary greatly from apartment to apartment. The results will be used in the measurement protocol for Finnish homes to be developed in 2025.