Application of airborne radiometric surveys for large-scale geogenic radon potential classification

Background: Indoor radon represents an important health issue to the general population. Therefore, accurate radon risk maps help public authorities to prioritise areas where mitigation actions should be implemented. As the main source of indoor radon is the soil where the building is constructed, maps derived from geogenic factors ([e.g. geogenic radon potential [GRP]) are viewed as valuable tools for radon mapping. Objectives: A novel indirect method for estimating the GRP at national/regional level is presented and evaluated in this article. Design: We calculate the radon risk solely based on the radon concentration in the soil and on the subsoil permeability. The soil gas radon concentration was estimated using airborne gamma-ray spectrometry (i.e. equivalent uranium [eU]), assuming a secular equilibrium between eU and radium (226Ra). The subsoil permeability was estimated based on groundwater subsoil permeability and superficial geology (i.e. quaternary geology) by assigning a permeability category to each soil type (i.e. low, moderate or high). Soil gas predictions were compared with in situ radon measurements for representative areas, and the resulting GRP map was validated with independent indoor radon data. Results: There was good agreement between soil gas radon predictions and in situ measurements, and the resultant GRP map identifies potential radon risk areas. Our model shows that the probability of having an indoor radon concentration higher than the Irish reference level (200 Bq m-3) increases from c. 6% (5.2% – 7.1%) for an area classified as Low risk, to c. 9.7% (9.1% – 10.5%) for Moderate-Low risk areas, c. 14% (13.4% – 15.3%) for Moderate-High risk areas and c. 26% (24.5% – 28.6%) for High risk areas. Conclusions: The method proposed here is a potential alternative approach for radon mapping when airborne radiometric data (i.e. eU) are available.

G lobally, radon is the second most common cause of lung cancer in smokers and the primary cause of lung cancer in non-smokers (1)(2)(3). Although the adverse health effects of radon exposure are a significant public health issue, these effects may be mitigated if appropriate control measures are implemented (4). In this regard, radon maps are used by regional and national authorities to support policies that protect the general population against the harmful effects of ionising radiation (e.g. 5-7). The main goal of radon maps is to delineate areas where high radon concentrations may be expected, in order to 'prioritize' these areas in a National Radon Action Plan (8,9). Radon maps may also be used as a basis for determining if preventive measures are required in new buildings, or if radon measurements are required in workplaces (e.g. 10).
The definition of Radon Priority Areas (RPAs) is ambiguous and several criteria (not mutually exclusive) may be applied, depending on political, public health, economic, and other decisions (8,11). The European Commission (EC), for example, defines RPAs as 'areas where the radon concentration (as an annual average) in a significant number of buildings is expected to exceed the relevant national reference level' (7). However, the EC does not define 'a significant number of buildings' and the reference level is only set as ≤ 300 Bq m −3 , with the specific value dependent on individual national policies. Furthermore, the definition of an 'area' is not specified and may be defined by political boundaries (e.g. municipalities and districts) or regular geographical units (e.g. grid cells of 10 × 10 km or 1 × 1 km). Bossew (8) has discussed the definition and limitations of RPAs in more detail.
Radon maps are normally based on indoor radon measurements, with geogenic factors sometimes used to improve predictions (e.g. [12][13][14][15][16][17][18]. National maps based only on geological information (i.e. geogenic radon maps) are unusual (19). For example, the geogenic radon potential (GRP) map of Germany only takes into account radon in soil gas and soil permeability (20). This latter approach has the advantage that it measures or estimates the amount of radon that the earth is delivering, and therefore it is assumed that the resulting map is independent of building type (21). The approach that uses both indoor radon measurements and geogenic information, on the other hand, has the advantage of taking account of the radon concentration at the point of exposure (12).
A fundamental problem with radon maps based on indoor radon measurements is that they have legal implications in workplaces and public buildings but they are usually developed using indoor radon measurements made in residential dwellings. Since indoor radon behavior in workplaces and domestic dwellings may not be comparable, due to changes in how buildings are used and constructed (e.g. [22][23][24], the applicability of indoor radon maps for legislation in workplaces and public buildings may therefore be questionable (8). Maps that do not depend on building characteristics (e.g. GRP maps) would be preferable in this regard.
Estimation of GRP is a common practice in order to assess the risk of having high indoor radon concentrations in new buildings (e.g. in the Czech Republic), with the Neznal formula (25) being one of the most common methods used for this purpose. At a local scale, this approach may be also valid (26) but in many cases it may be impractical to have sufficient in situ representative data to develop GRP maps at national and/or regional scales.
Here we test a new approach to develop national/ regional GRP maps based on gamma-ray spectrometry radiometric measurements (i.e. eU) and subsoil hydraulic permeability. Such datasets are often collected for geological applications and may thus be available in many countries, but not specifically collected or used for radon mapping. For the analysis in this study, we have used data from the Tellus project, a national mapping programme that is collecting geochemical and geophysical data from Ireland and Northern Ireland (www.gsi.ie/tellus and http://www.bgs.ac.uk/gsni/tellus/overview/). Independent indoor radon measurements and in situ soil gas measurements have been used to validate the results. To date, the Tellus airborne geophysical programme has completed coverage of approximately 70% of the island of Ireland, and data are freely available on the Tellus website. Soil gas radon concentrations were predicted using 1) the measurement of eU at the surface and 2) soil properties (i.e. porosity, density, and radon emanation factor). The Neznal Radon Potential Index (25) was then estimated and compared with indoor radon measurements. Selected test sites were further studied using in situ soil gas radon measurements in order to further validate the radon estimations and to assess the applicability of the methodology proposed in this study.

Geogenic radon potential
We use the Neznal formula (25) for estimating the GRP. It bases the radon risk assessment on two main geogenic factors, namely 1) soil gas radon concentration which is the main indoor radon source and 2) soil permeability which is related to the radon transport in the environment: The equilibrium radon concentration in soil gas (C Rn in kBq m −3 ) and the soil permeability (k in m 2 ) were assigned values according to the categorization of each parameter, as proposed by Elio et al. (26) and summarized in Table 1.
C Rn and k were averaged by grids of 1 × 1 km and thus the GRP represents an average risk of having high indoor radon levels at the same 1 km 2 unit. The grid cells were classified as 1) Low risk (GRP < 10), 2) Moderate-Low risk (10 ≤ GRP < 22.5), 3) Moderate-High risk (22.5 ≤ GRP < 30), and 4) High risk (GRP ≥ 30) ( Table 2). Application of radiometric surveys for Rn mapping

Airborne radiometric survey
The Tellus airborne geophysical survey employs a low-flying aircraft (flying at 60 m in rural areas and 240 m in urban areas) to collect geophysical information on the properties of soils, rocks, and waters. A gamma-ray spectrometer (Exploranium GR820 and Radiation Solution RS-501 from 2018 onward) was employed to measure the concentration of eU (ppm), eTh (ppm), and K (%). The concentrations of uranium (U) and thorium (Th) are expressed as equivalent concentrations since the gamma radiation energy windows used to quantify these elements actually measure 214 Bi and 208 Tl, respectively, and secular equilibrium in each decay series is assumed (27). All surveys have been merged and the results are available on the Tellus website (www.gsi.ie/ tellus). The initial resolution of the data is 50 × 50 m (Fig. 1a), and for the purpose of this study, we aggregated the data into 1 × 1 km grid cells (Fig. 1b), calculating the arithmetic mean (AM) and the standard deviation of eU for each.

Soil gas radon predictions based on equivalent uranium
Radon is a radioactive gas that forms as a decay product of U and Th. However, due to the different half-lives of the isotopes, the radon risk in indoor air is principally caused by 222 Rn (28), which is a decay product of 226 Ra generated in the radioactive decay series of 238 U. 222 Rn is generated in the soil matrix due to the presence of 226 Ra. It has to be recoiled into the pore space (29), and thus its concentration in soil gas depends both on 226 Ra concentration and soil properties, principally soil density, effective porosity, and an emanation coefficient. Therefore, the theoretical 222 Rn concentration in soil gas due to 226 Ra can be estimated by the following equation (e.g. 30-33): where C Rn is the 222 Rn concentration in soil gas (kBq m −3 ), C Ra is the 226 Ra concentration (Bq kg -1 ), ε is the  emanation factor (adimensional), ρ is the soil density (g m −3 ), and n is the effective porosity (adimensional). Water saturation in the pore space (0 ≤ S F ≤ 1) increases the radon concentration in soil gas (31) since radon is concentrated in the air rather than in the water content of the pore space due to its higher affinity to air than water (i.e. K W/air = 0.25, where K W/air is the radon partitioning coefficient between water and air). Tellus airborne gamma-ray spectrometry radiometric data (i.e. eU in ppm) were aggregated by grid cells of 1 × 1 km. Mean values of radium (e 226 Ra; Bq kg -1 ) were then estimated assuming secular equilibrium (i.e. 226 Ra activity equal to 238 U activity), using a conversion factor of 12.35 to transform the eU concentration in ppm to 238 U activity in Bq kg -1 (34). Although a disequilibrium between 214 Bi and 238 U is frequent in the natural environment (i.e. equilibrium occurs after more than 1.2 × 10 6 years), in a closed system, equilibrium between 214 Bi and 226 Ra occurs after approximately 30 days and thus airborne gamma radiometric signals may be related more to the radium activity ( 226 Ra) in soil than to the uranium activity. We maintain, however, the nomenclature of equivalent (eRa) to clarify that we have inferred the concentration using eU. 222 Rn concentration in soil gas was then estimated in each 1 × 1 km grid cell using Equation 2. The radon variability in each grid was assessed by Monte Carlo simulations (n = 1,000), assuming that 1) ε, ρ, and n are normally distributed with typical values of 0.29, 1.35 g m −3 , and 0.30 and standard deviations of 0.03, 0.06, and 0.07, respectively (35), and 2) a uniform distribution of water saturation coefficient (S F ) with a minimum of 0.4 and a maximum of 0.6 ( Fig. 2). The 75th percentile of simulated soil gas radon concentrations in each grid cell was then used for the GRP calculation (i.e. C Rn measured in Table 1). The selection of the 75th percentile (P75) is arbitrary, and other values could be chosen (e.g. P50 and P95). However, changing the percentile cut-off will affect the outcome (GRP). Lower values (e.g. P50) would reduce the size of areas classified as high risk, and vice versa. We therefore consider that P75 is a reasonable value for not being under/over-protective.

Subsoil permeability
Subsoil permeability has been classified into three categories: 1) Low, 2) Moderate, and 3) High, based on the Groundwater Subsoil Permeability (GWSP) map of Ireland (36) and the all-Ireland Quaternary map (37, 38) (Fig. 3). The GWSP divided the country in three categories, depending on how easily water can infiltrate subsoils, that is 'High', 'Moderate', or 'Low'. We have therefore assigned these categories for the model, and where there is no datum (i.e. areas where subsoil is <3 m thick and in Northern Ireland), we inferred it by assigning a soil permeability to each Quaternary deposit type following the classification proposed by Appleton et al. (39,40). Furthermore, in the areas of the Quaternary map where there is no datum, 'Bedrock' was assigned as the soil type. This was then split into two types of permeability depending on whether the aquifer bedrock is karstified or not, as described in the GSI Groundwater Resources (Aquifers) -Aquifer Bedrock map (36) ( Table 3). The resulting-derived all-Ireland soil permeability map can be seen in Fig. 3.

Indoor radon concentrations
Indoor radon measurements were collected by the Environmental Protection Agency (EPA) as part of national surveys (www.radon.ie). Indoor radon was sampled by passive detectors installed in homes for a period of at least 3 months, and the readings were seasonally adjusted to represent an annual average (14). Two different datasets are available for comparing the GRP classification and the risk of having indoor radon concentration higher than the national reference level of 200 Bq m −3 . The first data set corresponds to indoor radon The total numbers of dwellings sampled and georeferenced in each national survey were 31,910 and 6,859, respectively. This corresponds to a total of about 16,700 indoor domestic radon measurements in the area covered by airborne radiometrics (Fig. 4).

Soil gas radon concentration
We have carried out independent soil gas radon concentration measurements in selected areas in Ireland, following the protocol described in Elío et al. (26). These measurements were conducted by first introducing a steel hollow probe in the soil at a typical depth of 75-100 cm, allowing manual sampling of the soil gas with a 150 mL syringe. The soil gas sample was subsequently introduced into an evacuated ionization chamber and after 15 min was measured using an RM-2 detector. The lower detection limit of the instrument is around 3 kBq m −3 and the uncertainty of radon measurements is below 20%. Further considerations for measuring soil gas radon concentration may be found in Neznal and Neznal (42). Three separate campaigns took place in the summer and autumn of 2017 (June, July, and September). We took soil gas radon measurements from 13 grids of 1 × 1 km, selected to cover a wide range of radium concentrations in soil (i.e. mean values of 226 Ra from 6 to 40 Bq kg -1 ; Table 5 and Fig. 8). Points in each grid were randomly selected and a total of 133 soil gas samples were taken (Table 5). Figure 5 shows the soil gas radon estimates based on Tellus airborne radiometrics data (eU). We have divided the region into six radon classes (based on the 75th percentile of the simulated data in Fig. 1 and the radon classification of Table 1). The highest values were found in the Mourne mountains (southeast of Northern Ireland), near Galway (central west part of Ireland), near Roscommon (central part of Ireland), and near Donegal town (northwest Ireland). In the 'Soil gas radon predictions' section, we compare the radon predictions with in situ soil gas radon measurements.

Geogenic radon potential map
The predicted radon values in soil gas (Fig. 5) and the soil permeability map (Fig. 3) were used to produce the GRP map of Fig. 6. The map is in general agreement with previous indoor radon maps both in Ireland and in Northern Ireland (12,14,43,44). Thus, for example, we can identify high-risk areas near Galway in the west, in the Mourne mountains in the northeast, parts of county Roscommon, counties Leitrim and Sligo, and south of Dublin in the  east. Conversely, areas with relatively low risk are also evident and in agreement with previous mapping and modeling (e.g. in the northern part of Northern Ireland and in parts of the west of Ireland). The lack of available geolocated indoor radon for Northern Ireland is not an issue for this map as it does not utilize such measurements.

Indoor radon concentration
Since 1998, Irish building regulations require that new houses built in areas designated as 'high radon areas' (HRA) must have a radon barrier (i.e. areas where the probability of having an indoor radon concentration higher than the national reference level of 200 Bq m −3 is    Application of radiometric surveys for Rn mapping 10% or higher (45)). The proportion of dwellings with radon prevention measures installed in the new indoor radon dataset, which includes radon measurements from 2013 to 2017, may therefore be higher than in the old one, which includes data from 1992 to 2013. In this regard, the use of indoor radon data from remediated dwellings should be avoided and hence in the new dataset, for any given dwelling, only the first indoor radon measurement was selected, reducing the number of data from 7,007 to 6,859 (second measurements in the same dwelling are assumed to be conducted after remediation activities). However, we were unable to further investigate this source of error since neither dataset contains metadata on building characteristics. Comparison between the old and new indoor radon datasets (Table 4) shows slight differences in summary statistics but the respective histogram and probability plots are almost identical (Fig. 7).
A student-t test indicates that there is no significant difference in the AM of both datasets (P-value = 0.042) but the difference between the mean values of the logarithmic transformed data (log 10 ) is statistically significant (P-value = 0.00014), although very small (i.e. 1.79 for the new dataset and 1.82 for the old dataset). On the other hand, if the indoor radon are treated as a binomial variable with respect to the reference level in Ireland (i.e. 1 when InRn > 200 Bq m −3 and 0 otherwise), both datasets have a similar proportion of indoor radon measurements higher than the reference level (Chi-square test; P-value = 0.2082). The probability of having an indoor radon concentration higher than the reference level is 11.9% (CI 95% : 11.14-12.69%) and 12.5% (CI 95% : 12.10-12.82%) for the new and old datasets, respectively. The spatial distribution of the high indoor radon values in both surveys also seems similar (Fig. 4). It is therefore reasonable to merge both datasets to be used for validating the GRP map.

In situ soil gas radon measurements
The results of in situ radon measurements are presented in Table 5. Values range from 13 to 335 kBq m −3 . Soil gas radon concentrations vary considerably within individual 1 × 1 km grid cells, with an average relative mean deviation (RMD) of 39% (range 23-55, Table 5). This indicates the difficulty of calculating an average soil gas radon concentration over large areas and also that by doing so we may lose some of the natural spatial variation of the radon soil gas concentrations. The AM and the geometric mean (GM) are also reported in Table 5. Values range from Moderate (i.e. G70565, G70947, and G69054), to High (i.e. G72081 and G56114), Very High (i.e. G98159), or Extremely High (i.e. G78544, G57626, G90163, G78164, G98929, G64863, and G73606) when classified according to Table 1.

Soil gas radon classification
Predicted radon measurements, at the 50th percentile level, are lower than the GM of the in situ measurements in all cases (Fig. 8), and while in some grids, the predicted values (i.e. 50th percentile) are similar to the in situ GM  (i.e. G70947, G69054, G98159, and G90163; Table 5), in other cases the relative difference may be up to 80% (i.e. G56114) ( Table 5). We acknowledge, however, that soil gas radon concentrations have a large spatial and temporal variability, even at small scale (42), and thus it is difficult to predict a precise radon concentration. At 1 × 1 km scale, it would be even more complicated as suggested by our in situ soil gas radon measurements, with RMD ranging from 23 to 55% (Table 5). For this reason, our goal is not to estimate an exact value but to get an order of magnitude of the soil radon concentration, and then evaluate where we can expect high or low values (Table 1). Finally, detailed surveys could be conducted in selected areas if necessary. Using our approach, a large number of the in situ soil gas radon measurements (black points) are within our predictions (Fig. 8). There are grids where predictions and measurements do not match, and the reason for this is currently unknown. However, our results indicate that predicted radon values based on radiometric data are within the range of in situ measurements, and therefore they may be seen as reasonable estimates of the equilibrium soil gas radon concentration used in Equation 1.
Some discrepancies between in situ and predicted values may arise because estimates only take into account the 222 Rn generated in the soil by the presence of 226 Ra (eRa), but not other possible radon sources (e.g. radon from groundwater or deep sources carried by other gases in fracture areas). Furthermore, we do not take into account the influence of water saturation on the emanation factor, which will also affect radon concentration in soil gas (46,47). Predictions were calculated with the same parameters, and thus we did not make adjustments for different soil types (e.g. use of different values of ε, ρ, and n according to the different Quaternary deposits). Therefore, the initial radon predictions based on airborne radiometrics show the potential of this analysis but it requires modification and improvement before being used to predict radon potential with a higher degree of confidence.
Further analysis of the high variability of radon concentration in soil gas within 1 × 1 km grid cells is also Fig. 8. In situ versus airborne radium (eRa) concentration in soil (color points represent the 2.5th, 50th, 75th, and 97.5th percentiles obtained in the simulation; the black points are the soil gas radon samples; and the black diamonds are the GM of the radon measurements in each grid). necessary ( Fig. 8 and Table 5). In some grids, the range of measured soil gas radon concentrations is relatively low (e.g. grids G70565 and G70947), but in others measured radon concentrations vary by an order of magnitude (e.g. grid G90163). A better understanding of the possible causes such as different Quaternary/Bedrock geology, possible transport of radon in groundwater (e.g. karstified areas), influence of subsoil permeability, or geological faults on radon will help achieve a better interpretation of the airborne radiometric data.

Geogenic radon potential versus indoor radon
The probability of having an indoor radon concentration in excess of the reference level of 200 Bq m −3 in each GRP area rises with the risk classification, from 6% for a Lowrisk area to 26% for a High-risk area ( Table 6). The GM of indoor radon concentration in each GRP area is also increased (i.e. from 47 to 98 Bq m −3 ; Table 6 and Fig. 9). These results suggest that the methodology described here is useful for defining RPAs, and thus when the Tellus project has completed surveying all of the island of Ireland, an all-Ireland GRP map could be developed.

Conclusion
We demonstrate that the methodology described in this study is useful for radon risk assessment at a national scale. The main advantage of this methodology is that indoor radon measurements are not required and therefore the technique may be applied to rural areas with a low-population density or in areas with no available indoor radon data. Furthermore, since the resulting radon potential map is independent of the type of building and occupancy styles, it may be applied to workplaces, and public and residential buildings. The risk of misclassification due to sample design, errors in the geocoding of the dwelling addresses, and/or the misunderstanding of building characteristics and living styles is also reduced with our approach. On the other hand, airborne geophysical surveys are expensive and may not necessarily justify a national survey solely for radon protection. However, airborne geophysical surveys are very useful for geological purposes and National Geological Surveys often carry these out to improve knowledge of the national and regional geology (e.g. [48][49][50], or as a prospecting tool to aid mineral exploration. The approach described here provides an opportunity to add value to existing airborne radiometric datasets.
Regarding possible limitations of using indoor radon data, it is possible that temporal variations of indoor radon and seasonal adjustments for representing annual average radon concentrations may also generate errors that are difficult to quantify. It is also difficult to extract all factors that may affect indoor radon concentration. Indoor radon maps may therefore be biased by the sampling method and by dwellings where radon prevention measures were installed but not reported in the survey. Furthermore, radon maps based on indoor radon measurements require a high number of individual measurements, which are normally taken over several years, meaning indoor radon data may be affected by seasonal and year-to-year variations.   It is anticipated that the average indoor radon concentration, and so the radon risk, in a given area may decrease with the implementation of radon action plans. Given that radon classification for a specific area may then change over time, it may not be best practice to merge radon surveys over different years, especially when metadata on remediation are not available. In this regard, an area classified as High risk may a few years later effectively be classified with a lower radon risk in a subsequent survey. We have attempted to remove this unintended bias in indoor radon datasets by only using the first measurements from a given property. However, we cannot exclude the possibility that we have inadvertently used indoor radon measurements from homes that have been remediated or have a radon barrier. In the case of this study, the indoor radon measurements have been used to validate our model, rather than train it. The fact that we see a good agreement between GRP categories and the GM of indoor radon for these designated areas suggests that the model performs well, despite the possibility that we may have inadvertently included some remediated properties in the model validation exercise.
The inherent spatial variability of soil gas radon concentrations over small distances does not allow a precise relationship between soil gas Rn and airborne Ra to be computed. However, a broad correlation between soil gas Rn and airborne Ra concentrations can be discerned. The predicted capacity of the resulting map may be improved by including both the influence of water saturation on soil permeability (i.e. reducing it) (51) and the possible increase of the emanation factor (ε) due to water saturation (i.e. increasing it) (46,47). It would be also beneficial for the model to differentiate between soil types (i.e. ε, ρ, and n). Finally, an optimization of the RP thresholds (Table 2) may also be applied for taking into account the specific characteristics of the building stock and living styles of a country in the risk assessment.