Utilizing Artificial Neural Network to Predict Soil Gas Radon in Lecce Province, Italy: Implications for Environmental Risk Assessment and Forecasting

The presence of radioactivity in the environment results in substantial effects to human health, thus a full understanding of contributing factors in this context is of great interest. The present study considers geological features, permeability, and climatic variations factors, as well as the concentrations of uranium, thorium, and potassium in the soil and the radon in soil gas data collected from 32 sites in the province of Lecce, Italy, in 2012 to assess radon exposure risks by using Artificial Neural Network (ANN) method, which offers valuable insights into potential health implications and aids in devising risk mitigation strategies. In particular, this study uses the distance of lithological units based on uranium content in bed rock to highlight regions with both high and low uranium concentrations. Geological inhomogeneities, including faults, caves, dolines and sinkholes significantly influence radon concentrations, with proximity to faults correlating with higher radon emissions. Distances to these geological features and their densities are considered as an important factor to predict radon concentrations. The effect of soil permeability on radon concentration is evaluated by taking into account the available water capacity, bulk density and fine fraction in the area under study. Finally, the influence of climatic variations, including barometric pressure, air temperature and wind speed, is considered to forecast radon concentration in soil. Using this comprehensive dataset, the ANN method is employed to predict and analyze radon exposure risks. The resulting ANN model demonstrates exceptional accuracy and reliability and a high Goodness-of-Prediction. Subsequently, drawing on forecasting radon concentrations across the province, a detailed contour map is obtained for illustrating the spatial distribution of radon concentration in different areas. Notably, specific regions in the central and south-western areas show radon in soil gas concentrations exceeding 30 KBq/m. This model offers critical insights for evaluating exposure risks and guiding future research and mitigation strategies. It also underscores the importance of considering uranium content, geological factors, and climate variations. Finally, it serves as a robust tool for policymakers and researchers to make informed decisions to address environmental challenges and prioritize public safety. Acknowledgement We would like to extend our heartfelt gratitude to Dr. Giorgia Cinelli for invaluable assistance and guidance throughout the course of this project, which greatly enriched our study. References De Iaco, S., Maggio, S., Palma, M. (2017). Radon Predictions with Geographical Information System Covariates: From Spatial Sampling to Modeling. Geographical Analysis, 49(2), 215-235. Oni, O. M., Aremu, A. A., Oladapo, O. O., Agboluaje, B. A., Fajemiroye, J. A. (2022). Artificial neural network modeling of meteorological and geological influences on indoor radon concentration in selected tertiary institutions in Southwestern Nigeria. Journal of Environmental Radioactivity, 251, 106933. Farai IP, Fajemiroye JA, Oni OM, Aremu AA. Artificial neural network modeling of soil gas radon concentration on different lithologies for Southwest Nigeria. J Radiat Res Appl Sci. 2022 Sep;15(3):34652. Neznal M, Neznal M, Matoln M, Barnet I, Miksova J. The new method for assessing the radon risk of building sites Nov metodika stanoven radonovho indexu pozemku.