ABSTRACT

This paper reports chaos and long-memory trends hidden in radon (²²²Rn) variations in the groundwater of Panzhihua, Sichuan Province, China, between 2012 and 2017. The analysis is performed using sliding-window (a) detrended fluctuation analysis (DFA), (b) fractal dimension analysis with the methods of Higuchi, Katz and Sevcik and (c) residual radon concentration (RRC). Several fractional Brownian motion (fBm) persistent time series segments of high predictability are found, with DFA slopes above 1.5 and fractal dimensions, below 1.5. Numerous seven-day segments exhibit RRC out of the ±2σ limits and are of noteworthy precursory value. Through a novel two-stage computational approach, the persistent pre-seismic fBm earthquake footprint segments are separated from the low-predictability ones. Several combined segments of dynamical complexity are found with fractal and long-memory behaviour. For these segments, associations are attempted with major (M_w ≥ 6.0) earthquakes occurred in China and border areas of near countries during the period of study. Out of seventeen earthquakes of the period, four earthquakes are identified with all combinations of methods, whereas the remaining earthquakes, with the combination of at least three methods. Trends of long-memory are identified and discussed. The findings are compatible with fractal, and SOC final phases of generation of earthquakes. Finally, potential geological sources are discussed and analysed.