Predictive modelling of radon variations in time series using wavelets, multiple linear regression and ARIMA.

Radon (222Rn), a naturally occurring radioactive gas, is the byproduct of the uranium decay series. As a naturally radioactive gas, radon is frequently used as a geophysical tracer to find underground faults and geological formations, in uranium surveys, and to forecast seismic events. Abnormalities in radon time-series (RTS) data have been studied before seismic events, indicating that it may act as an earthquake precursor. This paper examined complex RTS with various climatological factors, viz. temperature, pressure and humidity, to extract relevant meaningful physical information by employing various simulation techniques. By employing wavelet-based regression (WBR) on RTS data, radon exhibits linear behaviour with temperature, but non-linear behaviour is observed with pressure and humidity. The anomalies in RTS were found before the seismic events. Multiple linear regression (MLR) also shows the positive relationship of radon with pressure and humidity. The autoregressive integrated moving average (ARIMA) model is utilized to analyse patterns, trends and stationarity in RTS data and predict it over a specified period. The method focuses on selecting the optimal model for predicting radon concentration over an uncertain period. This is done by identifying the one model with the lowest Akaike information criterion (AIC) and the Bayesian information criterion (BIC) values. The experimental results indicate that the ARIMA model outperforms others in predicting radon concentrations over an extended period. This research work not only contributes to the domain of earthquake precursors but also aligns with United Nations SDG 3 by understanding environmental health factors. Moreover, SDG 9 applies advanced technologies for infrastructure safety, and SDG 13 enhances disaster risk reduction and seismic resilience.