Deciphering Radon Variability in the Northern Upper Rhine Graben: An Analysis Using Passive and Active Detection with Random Forest Modelling

Abstract

This study, conducted about 30km south of Frankfurt in the Northern Upper Rhine Graben, focuses on deepening the understanding of Radon concentrations in soil air. The selected area, where neotectonic activity was proven in an accompanying project, provides an ideal setting for investigating Radon variability, particularly its potential correlation with fault zones in unconsolidated rocks or sedimentary basins. Understanding the factors influencing Radon levels in the environment is a complex task, as they are affected by a multitude of variables. Our work aims to decipher these influences and, if possible, quantitatively analyse the contributions of each variable. By doing so, we hope to gain a clearer understanding of how different environmental factors interact to determine Radon levels. A central element of our research is the use of Random Forest models, chosen to handle our multidimensional dataset. This dataset includes a variety of parameters such as Radon measurements, nuclide content, soil grain sizes, weather data, and the distance to fault zones. Random Forest models are particularly effective for this type of complex data because they can analyse many different factors at once and uncover hidden patterns. Contrary to initial hypotheses, our findings indicate that in unconsolidated rocks and sedimentary basins, the grain size of soil is the most influential factor in determining soil air Radon levels, closely followed by soil moisture. These results challenge the previously held belief that fault zones are the primary influencing factors on Radon concentrations in these geological settings.