Earthquake forecasting probability by statistical correlations between low to moderate seismic events and variations in geochemical parameters

Since late 2002, a network of six automatic monitoring stations is operating in Tuscany, Central Italy, to investigate possible geochemical precursors of earthquakes. The network is operated by the Institute of Geosciences and Earth Resources (IGG), of the National Research Council of Italy (CNR), in collaboration and with the financial support of the Government of the Tuscany Region. The areas of highest seismic risk of the region, Garfagnana, Lunigiana, Mugello, Upper Tiber Valley and Mt. Amiata, are currently investigated. The monitoring stations are equipped with multi-parametric sensors to measure temperature, pH, electric conductivity, redox potential, dissolved CO₂ and CH₄ concentrations in spring waters. The elaboration of long-term time series allowed for an accurate definition of the geochemical background, and for the recognition of a number of geochemical anomalies in concomitance with the most energetic seismic events occurred during the monitoring period (Pierotti et al., 2017).

In an attempt to further exploit data from the geochemical network of Tuscany in a seismic risk reduction perspective, here we present a new statistical analysis that focuses on the possible correlation between low to moderate seismic events and variations in the chemical-physical parameters detected by the monitoring network. This approach relies on the estimate of a conditional probability for the forecast of earthquakes from the correlation coefficient between seismic events and signals variations (Fidani, 2021).

Seismic events (EQ) are classified according to a magnitude threshold, M_o. We set EQ = 0, if no seismic events were observed with M < M_o, and EQ = 1, if at least a seismic event was observed with M > M_o. Chemical-physical (CP) events were defined based on their appropriate amplitudes threshold A_o, being CP = 0 if the amplitude A < A_o, and CP = 1 if A > A_o. Digital time series were elaborated from data collected over the last 10 years, where EQs were declustered and CPs detrended for external influences. The couples of events with the same time differences T_EQ – T_CP, between EQs and CPs, were summed in a histogram. Then, a Pearson statistical correlation coefficient corr(EQ,CP) was obtained starting from the covariance definition.

A conditional probability for EQ forecasting is estimated starting from the correlation coefficient in an attempt to use data from CP network of Tuscany in a seismic risk reduction framework. The approach consists in an evaluation of EQ probability in a defined area, given a CP detection by the station in the same area. The conditional probability P(EQCP), when a correlation between EQs and CPs exists and time difference is that evidenced by the correlation, is increased by a term proportional to the correlation coefficient as

 

with respect to the unconditioned probability P(EQ) when a CP event is detected, where P(CP) is the unconditioned probability of CP.

 

 

Fidani, C. (2021). Front. Earth Sci. 9:673105.

Pierotti, L. et al. (2017). Physics and Chemistry of the Earth, Parts A/B/C, 98, 161-172.