Geochemical continuous signals processing by multiple statistical techniques: application to the Gallicano thermomineral spring

Since early 2003 a network of automatic stations (Geochemical Network of Tuscany, GNT) is operating in Tuscany, central Italy, to investigate possible earthquake geochemical precursors in spring waters. The continuous automatic monitoring stations of the GNT have been designed and realized at IGG-CNR-Pisa in keeping with the recommendations of the International Association for Seismology and Physics of Earth's Interior, and are equipped with sensors for the concurrent measurement of different parameters to ensure cross-checking of independent signals. Here we present data collected over the period 2003-present from the Gallicano thermomineral spring, located in Garfagnana, one of the areas of highest seismic risk in Tuscany. The Garfagnana region belongs to the inner zone of the northern Apennine fold-and-thrust belt consisting of structural units derived from both oceanic and continental domains now represented in the Ligurian, Subligurian and Tuscan Units. The area is tectonically active, and a number of extensional structures (low- and high-angle normal faults) confer a ''semigraben'' type geometry to the local intermountain basins. The Gallicano spring emerges at an altitude of 209 m a.s.l. and is characterized by an average flow rate of about 1.5 L/s. The discharged water has a medium salinity (up to 4.2 g/L) and a temperature of about 25°C. The continuous automatic monitoring station installed at Gallicano is equipped with sensors for the simultaneous measurement of temperature, pH, electrical conductivity, redox potential, and dissolved content of CO2 and CH4. The station operates with flowing water (about 5 L/min), and a frequency acquisition of 1 s for all the parameters. Manual calibration procedures are done on a monthly basis, and/or in conjunction with anomalous variations of the signals. By combining this continuous and discrete monitoring approach, some understanding of the deep circulation paths, and of the geochemical processes governing the chemical evolution of the waters discharged at Gallicano and in the surrounding areas, has been gained, and an integrated hydrogeological and geochemical model of the system has been defined. Data series have been analysed with multiple statistical methods to highlight possible temporal trends (linear vs. sinusoidal, short- vs. long-term), and anomalous variations related to seismic activity. The applied methods include: Fourier analysis, multiple regression, factorial analysis, Census I method, Sinclair method, Neural Network Analysis. A number of CO2 anomalous concentrations have been recognized in concomitance with three major seismic events occurred in northern Tuscany and Emilia Romagna regions, not far from the Gallicano area. Accurate statistical processing of geochemical continuous signals emerges as a powerful and essential tool to ensure a reliable identification of anomalies, and to possibly increase the confidence in the use of geochemical precursors of earthquakes.