GNGTS 2022 - Atti del 40° Convegno Nazionale

GNGTS 2022 Sessione 2.1 219 cycle component (TCt). The seasonal component also includes earth tides, that can affect the release of deep gas from the rocks (Sugisaki, 1981). In order to separate the irregular component from the original data series, all these components have been combined with the additive model (Xt = St + TCt + It). The additivemodel is applied because the seasonal fluctuations of the recorded geochemical signal are steady, and not dependent on the overall level of the series (Makridakis et al., 1998). The decomposition is performed following a step-by-step procedure: (i) determination of the duration of one season. In this study we consider a reference frame of 12 months, by exploring the effects (always negligible in our case study) of different setupwith seasons of 6 and 3 months; (ii) computation of the centredmoving average of the series, with a windowwidth of one season; (iii) computation of the seasonal component (St) by subtracting the moving average from the original dataset; (iv) computation of the so-called ‘‘seasonally adjusted series’’ by removing the seasonal component from original series; (v) individuation of the trend/cycle component (TCt), by smoothing the seasonally adjusted series; (vi) calculation of the irregular component by subtracting the seasonal and trend/cycle components from the original data series. The irregular component of the CO 2 time series from Gallicano, retrieved during the same earthquake time interval is shown in Figure 2 left. Figure 2 right reports the histogram distribution of the irregular component and its fit. The histogram was fitted by a special function that defined a threshold over which CO 2 recordings had a 90% of probability to be not due by chance. It occurred when the modulus of the irregular component was above 0.0432. A set of 49 CO 2 anomalies were selected from the irregular component and indicated by “1” in the series of daily anomalies. Starting fromdigital time series, the Pearson correlation (known as theMatthews correlation for the digital case) between regional seismic events and the Gallicano CO 2 anomalies was calculated (Fig. 3). Starting from Matthews correlations (Matthews, 1975), it was shown that the conditional probabilities can be defined between two sets of digital events (Fidani, 2018). Moreover, correlation histograms calculated in past publications (Fidani, 2015), were put in relation with the Matthews correlation coefficients (Fidani, 2020). Being so, indicating by EQ the earthquake event, by EC the CO 2 anomaly event, and by R = corr(EQ,EC) the Matthews Fig. 1 - Declustered earthquake dataset displayed by z-map after that two declustering methods were used one after the other, completeness magnitude 1.2. The brown line is the Thyrrenian coastline; the blue line is the main local fault, and the red point along it marks the position of the Gallicano station. Dimensions and colours of square symbols indicate earthquakes magnitude and depths, respectively. The yellow star marks the most energetic seismic event occurred in the studied period. Statistical elaborations are provided in the two right boxes.