Twenty years of CO2-emission monitoring at Cava dei Selci, Colli Albani volcano (Central Italy)

Carbon dioxide flux from the soil has been monitored for 20 years at Cava dei Selci, the main degassing site of Colli Albani quiescent volcano. Cava dei Selci gas discharge occurs at the north-western periphery of the volcano, within an old stone quarry crossed by a NW-SE volcano-tectonic lineament. The area around the manifestation has been densely urbanized and lethal accidents by gas inhalation have occurred to a man and to dozens of animals including cows and sheep. Some houses had to be permanently evacuated because of hazardous indoor gas concentrations. Emitted gas is dominated by CO₂ (>90 vol.%) with <1 vol.% of H₂S. Isotopic composition (δ¹³C and ³He/⁴He) suggests a deep magmatic origin. No significant compositional variations have been recorded during the observation period.

Surveyed area includes a fixed grid of 130 points, regularly distributed over an area of about 5500 m², where soil CO₂ flux surveys have been carried out 55 times from May 2000 to August 2020 by accumulation chamber. Collected data have been reprocessed by sequential Gaussian simulation. The total diffuse CO₂ output is highly fluctuating, with a maximum rate of 24.8 t*d⁻¹ in January 2006 and a minimum value of 5.6 t*d⁻¹ in December 2003; the estimated mean±1σ is 12.1±4.5 t*d⁻¹. All the flux maps show typically a highly emissive area in the internal sector of the investigated grid, with NW-SE elongation. Another anomalous zone, with the same elongation, is found in the SW of the survey area. Diffuse degassing rate (total flux normalized by survey area) is similar to that of active volcanic zones.

In the same zone an automatic permanent station continuously measured the soil CO₂ flux and environmental parameters (which may influence the soil gas flux) from 2004 to 2008 and from 2019 to present. Results of timeseries processing by Multiple Linear regression and Principal Component analysis, commonly used to filtrate and clear data from atmospheric inferences (for example at Stromboli and Campi Flegrei), were unsatisfying for Cava dei Selci. Therefore, we reprocessed the timeseries by the stochastic Gradient Boosting Trees regression technique. This allowed to explain up to 55 % of the CO₂ variance by environmental variations; 45 % of the variance therefore reflects deep-seated processes. This technique looks promising for the regression of soil CO₂ flux timeseries. The results of 20 years monitoring confirm that Cava dei Selci is a convenient site for both monitoring a potential unrest of the volcano and assessing the gas hazard in the nearby inhabited zone.