Geographically Weighted Regression technique to construct the Geogenic Radon Potential map of the Lazio region: a methodological proposal for the European Geogenic Radon map

The assessment of high indoor radon concentrations (i.e., radon-prone areas, RPAs) in areas with different size by using only indoor radon measurements may be affected by the geographic scale of the surveys, especially if urban, sub-urban and rural zones are mixed. The spatial distribution of indoor radon samples is often linked to the clustered distribution of houses within the inhabited zones. Furthermore, because of its multifactorial dependence (i.e., physical, meteorological and anthropic parameters), indoor radon usually shows strong variability at least on short geographic scale (i.e., non-autocorrelated and non-stationary spatial behavior). Therefore, the direct interpolation of indoor radon values beyond the boundary of an urban area to identify RPAs could be a difficult and non-robust procedure to accomplish at large scale. The mapping of indoor radon could have a meaning only within the inhabited areas and in standard conditions to help local administrations to adopt simple and cheap recommendations to apply in case of building remediation aimed at preventing and/or reducing radon entry in buildings. An alternative approach considers the construction of Geogenic Radon Potential (GRP) maps by using the geological and geochemical information (i.e., soil permeability, faults, U and Ra content, emanation coefficient, etc.), calibrated through soil gas radon, and then introducing indoor measurements. In this work, different multivariate geospatial techniques, such as Geographical Weighted Regression (GWR) and Empirical Bayesian (EB) were used to construct the GRP map of the Lazio region. Unlike traditional ordinary least squares (OLS) models, GWR allows a better estimate of the local parameters of the relationships under study and has become a complement tool to global modelling, accounting for spatially changing relationships (non-stationarity). Moreover, a comparison between the GWR approach results with the outcomes obtained by Empirical and Full Bayesian inference, applied in order to reduce the uncertainty of the corresponding estimates, is proposed. To get reliable confidence interval for the hyperparameters of the model, a boostrap procedure was also adopted. Geological data, soil gas (about 7000 samples) and indoor data (about 500 samples) were provided by the Soil Protection and Remediation Department of Regione Lazio and by the Fluid Chemistry Laboratory of the Earth Science Department, Rome University Sapienza, respectively. This wide database was elaborated in the GIS environment by using ArcGIS 10.2 (Copyright © 1999-2013 Esri Inc.). All produced maps are constructed according to a grid format with 1000x1000m unit cell resulted by using vector to raster transformation, reclassification and interpolation of primary geological, geomorphological and geochemical data. The obtained results are aligned with the development of the GRP European map, and can also be considered as good first step toward the difficult assertion that there may be a relationship between the radon concentrations in soil gas and those measured in homes. Furthermore, it is worth to note that the two maps can be used both by local authorities for different but complementary purposes: the map of GRP can be used for land use planning and site selection, whereas the Indoor Radon Potential map for targeted monitoring of dwellings. These maps can also help in allocating resources to plan more efficiently denser surveys of both soil gas and indoor radon, remediation of affected houses and the implementation of construction methods for new houses most suitable to the