Wavelet-based filtering and prediction of soil CO2 flux: Example from Etna volcano (Italy)

Abstract

In this work, we propose a wavelet-based filtering for soil CO2 flux time series. The filter relies on the detection of the periodic components achieved by means of the long-term time-frequency characterization of the time series. For this purpose, we exploited the vast data set coming from the monitoring network installed at Mt. Etna volcano (Italy). The network provides hourly measure of CO2 flux together with the measure of the climatic variables. These data allow to investigate the relationships between CO2 time series and the potentially influencing meteorological factors. This has been assessed calculating the wavelet coherence between CO2 time series against air temperatures, atmospheric pressure, and relative humidity in all the sites where these information were available. Results highlight the occurrence of marked cycles at about ∼1 year for the most of the sites while shorter cycles occur only at some sites. From these cycles a periodic signal can be calculated, and therefore opportunely removed from the time CO2 series to enhance the volcano-related anomalies. We found also common cycles among CO2 and the climatic variables, which synchronicity is constant over time but it is site-specific. Starting from this consideration, we calculated a reference signal for CO2 combining analytically the temperature, the pressure, and the humidity cycles: this model of the climatic effect has been used to predict the seasonal trend of the CO2 output.