Atmospheric radon progeny and particulate matter (PM) pose significant environmental health risks, yet their interactions and co-variability remain poorly quantified. This study investigates the interplay between short-lived outdoor radon progeny and ambient particulate matter (PM10 and PM2.5) under urban conditions in Bratislava, Slovakia. Continuous measurements of radon progeny, PM levels, boundary layer height and meteorological observations obtained over a period of three years were analysed. To characterize the relationships among variables, a hierarchy of statistical and machine learning approaches was applied, including principal component analysis, multiple linear regression (MLR), generalised additive model (GAM), random forest (RF) and XGBoost. All models identified PM as the dominant predictor of equilibrium-equivalent radon concentration (EEC), with a clear positive association between them. A nonlinear relationship was observed between EEC and PM concentration, with EEC increasing sharply with PM levels (i.e., by more than a factor of two across the observed PM range), before reaching a plateau. The RF achieved the highest predictive accuracy (R2 = 0.72), substantially outperforming GAM (R2 = 0.49) and MLR (R2 = 0.43). These results demonstrate that PM concentration has a significant impact on attached radon progeny in the atmosphere, which are further modulated by meteorological factors. Overall, this study underscores the need for flexible nonlinear modelling to capture complex radon-PM dynamics and highlights the importance of aerosol interactions in radiation dose assessment.
Sultani et al. (Sat,) studied this question.