The effects of environmental parameters on diffuse degassing at Stromboli volcano: Insights from joint monitoring of soil CO2 flux and radon activity

Abstract

Soil CO2 flux and 222Rn activity measurements may positively contribute to the geochemicalmonitoring of active volcanoes. The influence of several environmental parameters on the gas signals has been substantially demonstrated. Therefore, the implementation of tools capable of removing (or minimising) the contribution of the atmospheric effects from the acquired time series is a challenge in volcano surveillance. Here, we present 4 years-long continuousmonitoring (fromApril 2007 to September 2011) of radon activity and soil CO2 flux collected on the NE flank of Stromboli volcano. Both gases record higher emissions during fall–winter (up to 2700 Bq * m−3 for radon and 750 g m−2 day−1 for CO2) than during spring–summer seasons. Short-time variations on 222Rn activity aremodulated by changes in soil humidity (rainfall), and changes in soil CO2 flux that may be ascribed to variations in wind speed and direction. The spectral analyses reveal diurnal and semi-diurnal cycles on both gases, outlining that atmospheric variations are capable to modify the gas release rate fromthe soil. The long-termsoil CO2 flux shows a slow decreasing trend, not visible in 222Rn activity, suggesting a possible difference in the source depth of the of the gases, CO2 being deeper and likely related to degassing at depth of the magma batch involved in the February–April 2007 effusive eruption. To minimise the effect of the environmental parameters on the 222Rn concentrations and soil CO2 fluxes, two different statistical treatments were applied: the Multiple Linear Regression (MLR) and the Principal Component Regression (PCR). These approaches allow to quantify theweight of each environmental factor on the two gas species and showa strong influence of some parameters on the gas transfer processes through soils. The residual values of radon and CO2 flux, i.e. the values obtained after correction for the environmental influence, were then compared with the eruptive episodes that occurred at Stromboli during the analysed time span (2007–2011) but no clear correlations emerge between soil gas release and volcanic activity. This is probably due to i) the distal location of the monitoring stations with respect to the active craters and to ii) the fact that during the investigated period no major eruptive phenomena (paroxysmal explosion, flank eruption) occurred. Comparison of MLR and PCR methods in time-series analysis indicates thatMLR can bemore easily applied to real time data processing in monitoring of open conduit active volcanoes (like Stromboli) where the transition to an eruptive phase may occur in relatively short times.