Speakers
Description
Radon concentrations in enclosed subterranean environments exhibit strong temporal variability driven by environmental conditions, making reliable short-term forecasting challenging. Elevated radon levels and the decay products of $^{222}$Rn are particularly relevant in underground laboratories, caves, and mines, where they pose health risks and can contribute significantly to the radiation background.
The HUN-REN Wigner Jánossy Underground Research Laboratory is a shallow underground facility with a maximum depth of approximately 30 m below the surface. Its 40-cm-thick outer walls, constructed from reactor concrete, provide substantial shielding from external radiation, making the laboratory suitable for low-background radiation measurements. However, elevated radon concentrations and $^{222}$Rn progeny increase the gamma-ray background and can adversely affect measurement conditions.
To monitor and mitigate elevated radon levels, we installed a local sensor network with remote data access. The monitoring system is integrated with an edge-deployed predictive model that forecasts radon concentrations and issues alerts for anticipated threshold exceedances. To address limited data availability while quantifying predictive uncertainty, we employ a Bayesian modeling framework that provides probabilistic forecasts of future radon concentrations. Forecasted ambient pressure, temperature, and relative humidity from open-source meteorological services are incorporated as exogenous covariates. This approach enables operational radon forecasting based solely on information available in real time and provides a framework for proactive radon mitigation in low-background underground environments.