Spectral-decomposition techniques for the identification of periodic and anomalous phenomena in radon time-series

Two hourly-sampled time-series of soil-gas radon concentrations of durations of the order of a year have been investigated for periodic and anomalous phenomena. These time-series have been recorded in locations having little or no routine human behaviour and thus are effectively free of significant anthropogenic influences. One measurement site, Sur-Fretes, is located in the French Alps, with saturated soil conditions; the second site, Syabru-Bensi, is located in Nepal, in a river terrace with unsaturated soil conditions. In such conditions, periodic components with periods ranging from 8 h to 7 days are often weak and intermittent and therefore , even in the presence of stationary forcing, difficult to identify. Two spectral decomposition techniques, Empirical Mode Decomposition (EMD) and Singular Spectrum Analysis (SSA), have been applied to these time series and yield similar results. For Sur-Fretes, weak diurnal and semi-diurnal components are observed with EMD, while SSA reveals only a diurnal component. In Syabru-Bensi, both EMD and SSA reveal a strong diurnal component and a weaker semi-diurnal component. Tidal components M1 and M2 are also suggested by EMD in Sur-Fretes, while these frequencies are not observed in Syabru-Bensi. The development of such analytical techniques can help in characterising the multiple physical processes contributing to the surface and subsurface dynamics of soil gases.