Radon potential mapping in Piemonte ( North-West Italy ) : An experimental approach

Early radon studies in Piemonte, an administrative district in North-West Italy (25200 km−2, around 4300000 inhabitants) have been done since 19901991, when a general radon survey of the dwellings of Piemonte was performed in order to assess the average radon exposure of the whole population. The survey, executed in the framework of the National Radon Survey by the National Environmental Protection Agency (former ANPA, now ISPRA) and ISS (National Health Institute), involved about 430 dwellings, chosen randomly with a stratified sampling technique. After this first step, radon researches were continued in different areas of Piemonte and involved schools as well as dwellings. In particular, radon surveys were conducted in areas where the geological conditions (i.e., the occurrence of rocks with Uranium content well above the typical average concentration found in the Earth crust) appear to favour a stronger radon emanation. Besides this kind of studies, other surveys were performed in order to assess the radon exposure in schools, where children and young students, the most radio-sensitive part of the population could be exposed to high radon concentrations. These extensive radon monitoring programs led to the implementation of


Introduction
Early radon studies in Piemonte, an administrative district in North-West Italy (25200 km −2 , around 4300000 inhabitants) have been done since 1990-1991, when a general radon survey of the dwellings of Piemonte was performed in order to assess the average radon exposure of the whole population.The survey, executed in the framework of the National Radon Survey by the National Environmental Protection Agency (former ANPA, now IS-PRA) and ISS (National Health Institute), involved about 430 dwellings, chosen randomly with a stratified sampling technique.
After this first step, radon researches were continued in different areas of Piemonte and involved schools as well as dwellings.
In particular, radon surveys were conducted in areas where the geological conditions (i.e., the occurrence of rocks with Uranium content well above the typical average concentration found in the Earth crust) appear to favour a stronger radon emanation.Besides this kind of studies, other surveys were performed in order to assess the radon exposure in schools, where children and young students, the most radio-sensitive part of the population could be exposed to high radon concentrations.
These extensive radon monitoring programs led to the implementation of a large radon database of more than 3500 radon measurements distributed all over the Piemonte Region.The whole radon database, before being used as a tool for the definition of the radon prone areas of Piemonte, was subject to a careful analysis and selection, in order to eliminate not representative measurements.In particular, to minimize a possible bias due to the wellknown radon fluctuations both on daily and seasonal basis, we considered only long term measurements (annual), performed using the nuclear track etch detectors technique (LR 115 or CR-39).
The radon potential mapping of the whole Piemonte was then achieved developing a "geolithological correlation model", based on a statistical analysis of the radon experimental data and the underlying geological, lithological and radiometric characteristics of soils and rocks.

Material and methods
The well-known phenomenon of the fluctuation of indoor radon concentrations both on daily and seasonal basis is probably the most important factor to be taken into account in order to harmonize a radon database.In fact, grab sampling measurements and short-term measurements (i.e., lasting a few days) often give results very different from long-term measurements, that are considered much more reliable, especially for radon mapping purposes.Therefore, for each sampling site, we decided to consider only those measurements able to give the annual average radon concentration.
Moreover, in order to minimize possible calibration and measurement procedure bias, we decided to include only the measurements performed with the same technique, based on a dosimeter equipped with standard nuclear track etch detectors .In this way, the original database was resized to about 2400 measurements.
In order to reduce the heterogeneity of the sample, due in particular to the floor where the dosimeters were installed (fig.1), a ground floor normalization of the data referred to higher floors was performed and validated (fig.2).
Assuming that the distribution of the indoor radon concentration at ground floor is approximately lognormal, the normalization was done as follows:     the geometric mean and geometric standard deviation.If we suppose that, in any given dwelling, a linear relationship holds between the radon concentration at ground floor (C GF ) and the radon concentration C F at a generic floor F , i.e.: where k is a constant to be determined, the radon distribution at generic floor F can be written as follows: where μ F = ln k + μ GF , σ F = σ GF and k is given by k = e μ F e μ GF .In order to include also the measurements performed in schools that represent the 42% of the whole database (fig.3), another normalization is needed.In fact, because of the different constructive characteristics, the radon concentration in typical school buildings is generally lower than those in dwellings.Therefore, the "school concentrations" were normalized to "dwelling concentrations": C dwellings = C schools + ΔC, where ΔC = GM dwellings − GM schools (fig.4).
Once obtained a global ground floor normalized database, the following step was the definition of the basic criteria of the radon potential mapping.First of all it was decided to consider a subdivision of the Region in 1206 administrative units, corresponding to the municipality of Piemonte.It was then defined, as radon potential indicator, the mean of the radon concentration measurements performed at ground floor and the related log-normal distribution.Unfortunately, being the number of municipality of Piemonte very large (1206), the actual database cannot give a representative sample for each administrative district.In fact, only in the municipality where the number of valid measurements were greater than 4, the mean and the related log-normal distribution was experimentally obtained.Therefore, in order to attribute an appropriate mean radon concentration value and a reliable log-normal distribution to all the administrative districts, we had to estimate these quantities for the municipalities where the experimental data are lacking.So, a "geolithological correlation model" was developed, based on a statistical analysis of the radon experimental data (fig.5) and the underlying geological, lithological (fig.6) and radiometric characteristics of the soils and rocks.
An ad hoc subdivision of the Region in lithological-radiometric units was performed taking into account also for the results of the analysis of a wide measurement campaign (γ spectrometry) of the natural radioactivity (mainly due to uranium) in the soils and the rocks of Piemonte.
In table I are reported the data of 214 Pb and 214 Bi, the natural radioisotopes belonging to the uranium series that, being the short lived radon daughters, can be regarded as good indicators of the potential radon emanation.
The rocks were then classified in four different categories, accordingly with their radioactivity content (average values of 214 Pb and 214 Bi): very low radioactivity (< 14 Bq/kg) low radioactivity (14 Bq/kg -30 Bq/kg) high radioactivity (30 Bq/kg -60 Bq/kg) very high radioactivity (> 60 Bq/kg).Taking into account for this information, it was possible to establish a more simple subdivision of the Region in only 26 lithological-radiometric units, instead of the original 44 lithological units (fig.6).Radon was then evaluated, using the experimental radon concentrations, in each of the new 26 lithological-radiometric units, thus obtaining, for each unit, a "Lithologic Mean" (LM)(table II).
It was then possible to compute the radon concentration mean AM j for the generic j -th municipality in whose area p different lithologicalradiometric units were present: where: LM k : Rn concentration mean (normalized to ground floor) of the k -th lithological unit AC j : surface of the j -th municipality area AL k : surface of the k -th lithological unit The geometric standard deviation (GSD), necessary for the definition of the log-normal distributions, was evaluated considering the asymptotic value of all the experimental GSDs (fig.7).
This approach was then validated comparing the values predicted by the model with the means experimentally calculated in those municipalities where the data available (fig.8).In this analysis, in order to avoid auto-correlation effects, the lithological means used for the model prediction were calculated excluding the data of the municipality where the mean was evaluated from experimental data.

EPJ Web of Conferences
The result obtained in this validation analysis can be considered quite good (R 2 = 0.71), thus allowing the use of the model for prediction purposes.

Results and discussion
It is well-known that the experimental data of indoor radon can be considered roughly distributed accordingly to a log-normal function: where and c i are the experimental values of the radon concentration.
The radon mapping of Piemonte was thus performed defining, for each municipality, the function f (c) reported in eq. ( 2), whose parameters were calculated in two different ways.For the municipality where experimental data were considered representative, f (c) was defined simply evaluating μ and σ from eq. ( 3).In the other cases, σ was calculated from the extrapolation of the experimental GSDs (see fig. 7), while μ was calculated from the AM j , evaluated by means of eq. ( 1), taking into account that the arithmetic mean AM of a variable log-normally distributed can be expressed as follows: (4) From the log-normal distributions, can also be calculated, in each sampling unit, the percentage of dwellings that exceed a given reference level R L .
In fig. 9 the results of the calculation of eq. ( 5) for the whole Region are reported.It can be seen that about 2% of the dwellings of Piemonte exceed the European Reference level of 200 Bq m −3 .In fig. 10 the map displaying the radon concentration AM for each municipality of Piemonte is reported.

Conclusions
With the present work it was possible to estimate the average radon levels in each of the 1206 municipalities of Piemonte (fig.9), more interestingly, to assess the percentage of the population exposed above a given radon concentration (fig.10), and to define the radon prone areas of the Region, an important achievement in order to evaluate the possible health effects for the population.The overall results (Regional arithmetic mean = 71 Bq m −3 ) were also in good agreement (fig.11) with those obtained in the first radon survey (National survey: 69 Bq m −3 ), performed in 1991 with a limited sampling program (430 dwellings).

Figure 1 :
Figure 1: Radon measurements available in our database: 56% were performed at ground floor.

Figure 3 :
Figure 3: Schools represent 42% of the whole radon database.

Figure 6 :
Figure 6: Lithological and radiometric characteristics of soils and rocks.

Figure 8 :
Figure 8: Validation of the model.

Figure 9 :
Figure 9: Estimation, by means of eq.(5) of the percentage of the dwellings of Piemonte exceeding a given radon concentration value (ground floor).

Figure 10 :
Figure 10: Average values in the municipalities of Piemonte -ground floor concentration (Bq m −3 ).

Figure 11 :
Figure 11: Distribution of the radon concentration in dwellings.
the GM GF and GSD GF being

Table I :
γ-spectrometry measurements (HPGe) in various types of soils and rocks ( 214 Pb e 214 Bi).

Table I :
Continued.

Table I :
Continued.