Italian Legislative Decree No. 28/2016, transposing Directive 2013/51/EURATOM, sets the requirements for the protection of public health regarding radioactive substances in water intended for human consumption. According to this decree, the Regions and Autonomous Provinces design and implement Control Programmes (PdC) for monitoring radioactivity in drinking water. The results are reported annually to the Ministry of Health and to the National Instituto of Health (Istituto Superiore di Sanità, ISS), the national public health institute, which maintains the national archive of both planned schedules and measurements actually performed. The statistical distributions of the activity concentrations of gross alpha, gross beta, radon and tritium in drinking water were analysed from measurements carried out by the regional environmental protection agencies (ARPA) under the first and second PdC during 2018-2022. The main challenge is the treatment of measurements below the Minimum Detectable Activity (MDA), about 55% of the total, with wide variability across parameters: from 28% for radon to 88% for tritium. These left-censored data are handled by two families of methods: substitution with a conventional value, under four assumptions (MDA, MDA/2, 0.7 MDA and MDA_Random), and adaptive methods, which treat censoring explicitly without assigning values to non-detects (maximum likelihood, MLE, and Kaplan-Meier). The Shapiro-Wilk test and the Departure Index identify the log-normal as the most appropriate distribution for all four parameters. The result is confirmed independently: the normal, gamma and Weibull families were fitted to the same measurements by censored likelihood and compared using the Akaike Information Criterion (AIC), requiring no substitution of non-detects. Observed quantiles align with the expected ones fully for gross alpha and radon, only partially for gross beta and tritium. Comparing the standard deviation estimated by MLE with that obtained under the four assumptions shows that substitution methods systematically compress dispersion, increasingly so as censoring grows: the MLE estimate exceeds the substituted one by 20-30% for radon and by almost a factor of three for tritium. The log-normality assumption is therefore verified where the data allow it and extended to Supply Zones (SZ) with few measurements, where a distributional check is not feasible; the intensively monitored sample does not differ appreciably from the overall set of SZs. The distribution thus identified will provide the basis for estimating the probability that future measurements exceed the parametric values. The results are intended for integration into the system used by the ISS to assess the adequacy of current drinking water monitoring, particularly the coverage of supply zones to be monitored.
Descriptive Statistical Distributions of alpha and beta-emitting radionuclides, radon and tritium Activity Concentrations in drinking water: results from data analysis / Teatini, M.F., Caprio, M., Carpentieri, C., Di Carlo, C.. - In: ISTISAN CONGRESSI. - ISSN 2384-857X. - 2:(2026), pp. 119-119.
Descriptive Statistical Distributions of alpha and beta-emitting radionuclides, radon and tritium Activity Concentrations in drinking water: results from data analysis
Teatini Maria Federica
Primo
;Di Carlo Christian
2026
Abstract
Italian Legislative Decree No. 28/2016, transposing Directive 2013/51/EURATOM, sets the requirements for the protection of public health regarding radioactive substances in water intended for human consumption. According to this decree, the Regions and Autonomous Provinces design and implement Control Programmes (PdC) for monitoring radioactivity in drinking water. The results are reported annually to the Ministry of Health and to the National Instituto of Health (Istituto Superiore di Sanità, ISS), the national public health institute, which maintains the national archive of both planned schedules and measurements actually performed. The statistical distributions of the activity concentrations of gross alpha, gross beta, radon and tritium in drinking water were analysed from measurements carried out by the regional environmental protection agencies (ARPA) under the first and second PdC during 2018-2022. The main challenge is the treatment of measurements below the Minimum Detectable Activity (MDA), about 55% of the total, with wide variability across parameters: from 28% for radon to 88% for tritium. These left-censored data are handled by two families of methods: substitution with a conventional value, under four assumptions (MDA, MDA/2, 0.7 MDA and MDA_Random), and adaptive methods, which treat censoring explicitly without assigning values to non-detects (maximum likelihood, MLE, and Kaplan-Meier). The Shapiro-Wilk test and the Departure Index identify the log-normal as the most appropriate distribution for all four parameters. The result is confirmed independently: the normal, gamma and Weibull families were fitted to the same measurements by censored likelihood and compared using the Akaike Information Criterion (AIC), requiring no substitution of non-detects. Observed quantiles align with the expected ones fully for gross alpha and radon, only partially for gross beta and tritium. Comparing the standard deviation estimated by MLE with that obtained under the four assumptions shows that substitution methods systematically compress dispersion, increasingly so as censoring grows: the MLE estimate exceeds the substituted one by 20-30% for radon and by almost a factor of three for tritium. The log-normality assumption is therefore verified where the data allow it and extended to Supply Zones (SZ) with few measurements, where a distributional check is not feasible; the intensively monitored sample does not differ appreciably from the overall set of SZs. The distribution thus identified will provide the basis for estimating the probability that future measurements exceed the parametric values. The results are intended for integration into the system used by the ISS to assess the adequacy of current drinking water monitoring, particularly the coverage of supply zones to be monitored.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.


