Radon production from rock and soil and its migration toward the surface along geological faults are natural processes that can favor radon entry in buildings, thus constituting a health risk. The analysis of these two processes and the construction of spatial models, as the contribution of different proxies of the geological radon source (GRS) (e.g., geology, soil properties, radionuclide content), and of the geological radon migration (GRM) pathways (e.g., faults, karst) in the subsoil, can be used to construct a geogenic radon hazard index (GRHI) map as a tool to predicting the susceptibility of an area (Radon Priority Areas, RPA) to increased indoor radon concentration for geogenic reasons. Many direct and indirect models (e.g., deterministic and probabilistic) have been used to create GRHI maps of a certain region. Here, we propose a bottom- up analysis including the integration of different factors (predictors and/or proxies) to construct a GRHI map of the whole Italian territory using a GIS-based (spatial) regression and by weighs their importance. In particular, we fitted a model by using Forest-based classification and Regression tool in ArcGIS based on about 35000 measured soil gas radon concentrations and known values of 8 explanatory variables (i.e., proxies) as a part of training dataset. The tool creates the model and generates predictions using an adaptation of Leo Breiman's random forest algorithm, which is a supervised machine learning method. The model can then be used to predict unknown values in a prediction dataset according to a 2x2km regular grid that has the same associated explanatory variables. The following explanatory variables are included in the model: U content (Bq/kg) of bedrock (Nogarotto et al., 2017), U content of the soil available from GEMAS (Reimann et al., 2014, http://gemas.geolba.ac.at/) and FOREGS database (Salminen et al., 2005; www.gtk.fi/publ/foregatlas), Fine Fraction (FF%) and Available Water Content (AWC %)(Ballabio et al., 2016), Fault Density (FD, Number of Faults/km2) from Italian national and regional database, Heat Flow (HF mW/m2) (Cataldi et al., 1995) and Karst Areas (KA) from the map of the world karst areas (Chen et al., 2017). All these predictors were transformed in 2x2km grid maps and then standardised by using fuzzy classification to transform input data to a 0/1 scale. The final map will be used by national and regional authorities to identify the Radon Prone Areas (RPA) as required by the European Directive 2013/59/EURATOM (art. 103).
Mapping the Geogenic Radon Hazard Index of Italy
Giustini F;Ciotoli G
2021
Abstract
Radon production from rock and soil and its migration toward the surface along geological faults are natural processes that can favor radon entry in buildings, thus constituting a health risk. The analysis of these two processes and the construction of spatial models, as the contribution of different proxies of the geological radon source (GRS) (e.g., geology, soil properties, radionuclide content), and of the geological radon migration (GRM) pathways (e.g., faults, karst) in the subsoil, can be used to construct a geogenic radon hazard index (GRHI) map as a tool to predicting the susceptibility of an area (Radon Priority Areas, RPA) to increased indoor radon concentration for geogenic reasons. Many direct and indirect models (e.g., deterministic and probabilistic) have been used to create GRHI maps of a certain region. Here, we propose a bottom- up analysis including the integration of different factors (predictors and/or proxies) to construct a GRHI map of the whole Italian territory using a GIS-based (spatial) regression and by weighs their importance. In particular, we fitted a model by using Forest-based classification and Regression tool in ArcGIS based on about 35000 measured soil gas radon concentrations and known values of 8 explanatory variables (i.e., proxies) as a part of training dataset. The tool creates the model and generates predictions using an adaptation of Leo Breiman's random forest algorithm, which is a supervised machine learning method. The model can then be used to predict unknown values in a prediction dataset according to a 2x2km regular grid that has the same associated explanatory variables. The following explanatory variables are included in the model: U content (Bq/kg) of bedrock (Nogarotto et al., 2017), U content of the soil available from GEMAS (Reimann et al., 2014, http://gemas.geolba.ac.at/) and FOREGS database (Salminen et al., 2005; www.gtk.fi/publ/foregatlas), Fine Fraction (FF%) and Available Water Content (AWC %)(Ballabio et al., 2016), Fault Density (FD, Number of Faults/km2) from Italian national and regional database, Heat Flow (HF mW/m2) (Cataldi et al., 1995) and Karst Areas (KA) from the map of the world karst areas (Chen et al., 2017). All these predictors were transformed in 2x2km grid maps and then standardised by using fuzzy classification to transform input data to a 0/1 scale. The final map will be used by national and regional authorities to identify the Radon Prone Areas (RPA) as required by the European Directive 2013/59/EURATOM (art. 103).I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
