10:15 〜 10:30
[SVC33-06] Joint analysis of seismic velocity change, deformation and meteorological data for volcano monitoring
キーワード：Ambient noise interferometry, Strain sensitivity, Ground deformation
Forecasting the timing and location of eruptive events are the main issues in volcanic risk mitigation. Seismic velocity changes are sensitive to strain, meteorological change and other modification of the medium. For these reasons seismic velocity changes show interesting potential to both, (1) temporal forecasting, by detecting “continuous” small precursory structural changes that are not measurable by the other observation techniques (e.g. gas migration), and (2) spatial forecasting of the eruptive vent locations, by allowing stress and structural changes location in the crust. In this study, we aim to assess both, the temporal and spatial monitoring potential of the seismic velocity changes. To do so, we analyzed jointly 10 years long time series of velocity, strain and meteorological changes at 21 Japanese volcanoes. They span different environmental conditions and volcanic behavior, insuring a broad sample of velocity–strain–environmental interactions.
Daily seismic velocity changes were computed using ambient noise interferometry. Three frequency bands were considered (0.5-1 Hz, 1-2 Hz and 2-4 Hz). Daily meteorological data were provided by the Japan Meteorological Agency (JMA) for rainfall, snowfall, temperature, atmospheric pressure and sea level. Deformation data include the GNSS stations of the JMA and Geonet network (from Geospatial Information Authority of Japan). Aerial strain was computed using a triangulation-based Bayesian approach (Xiong et al., 2021). We first averaged velocity and strain changes between all the stations available at each volcano, and removed effect of large earthquakes. We assumed the velocity change to results mainly from the linear combination of the strain and meteorological changes and jointly inverted their respective contribution expressed as follow (Wang et al., 2017):
(dv/v)synth = a1.Pfluid + a2.hsnow + a3.T + a4.Patm + a5.hsea + a6.ε + Cst
where, Pfluid is the pore fluid pressure in the rocks estimated from rainfall (Talwani et al., 2007), hsnow the snow thickness, T the temperature, Patm the atmospheric pressure, hsea the sea level change, ε the strain and Cst a constant. ai being the coefficient controlling the relative contribution of each component. To avoid overfitting using unnecessary contribution, we tested each possible combination of contribution and computed their Akaike Information Criterion (AIC).
We select two volcano showing clear signals to test our approach: (1) Izu-Oshima, which has been shown to have velocity changes closely related to strain (Takano et al., 2017) and (2) and Asamayama, which show a very strong seasonal effect, attesting of the environmental influence. For Izu-Oshima, we confirm that strain is the main parameter controlling the velocity change, representing 38-49% of the modeled velocity change, with strain sensitivity of 2x102, 3.4x102 and 3.6x102 strain-1 for the 0.5-1 Hz, 1-2 Hz and 2-4 Hz frequency bands, respectively. At Asamayama, for the 0.5-1 Hz band, despite a relatively poor fit, the velocity change was determined to results mainly from pore pressure (48%) and strain change (37% with a sensitivity of 2x102 strain-1). For the 1-2 Hz band, we found the velocity change to results mainly from the combination of temperature (50%), pore pressure (24%) and strain (22%; 3x102 strain-1). While the 2-4 Hz band seems to rather show relations with temperature (47%) and sea level (24%) variations. The remaining contribution being distributed between pore pressure, strain and snowfall.
Our preliminary results demonstrate the applicability of our approach to retrieve the different component contributing to the measured velocity change while determining coherent strain sensitivity values. This would allow (1) subsequent use of the velocity change as a stress proxy for vent location forecasting and (2) detection of potential small eruptive precursor by removing the modeled contributions.
Daily seismic velocity changes were computed using ambient noise interferometry. Three frequency bands were considered (0.5-1 Hz, 1-2 Hz and 2-4 Hz). Daily meteorological data were provided by the Japan Meteorological Agency (JMA) for rainfall, snowfall, temperature, atmospheric pressure and sea level. Deformation data include the GNSS stations of the JMA and Geonet network (from Geospatial Information Authority of Japan). Aerial strain was computed using a triangulation-based Bayesian approach (Xiong et al., 2021). We first averaged velocity and strain changes between all the stations available at each volcano, and removed effect of large earthquakes. We assumed the velocity change to results mainly from the linear combination of the strain and meteorological changes and jointly inverted their respective contribution expressed as follow (Wang et al., 2017):
(dv/v)synth = a1.Pfluid + a2.hsnow + a3.T + a4.Patm + a5.hsea + a6.ε + Cst
where, Pfluid is the pore fluid pressure in the rocks estimated from rainfall (Talwani et al., 2007), hsnow the snow thickness, T the temperature, Patm the atmospheric pressure, hsea the sea level change, ε the strain and Cst a constant. ai being the coefficient controlling the relative contribution of each component. To avoid overfitting using unnecessary contribution, we tested each possible combination of contribution and computed their Akaike Information Criterion (AIC).
We select two volcano showing clear signals to test our approach: (1) Izu-Oshima, which has been shown to have velocity changes closely related to strain (Takano et al., 2017) and (2) and Asamayama, which show a very strong seasonal effect, attesting of the environmental influence. For Izu-Oshima, we confirm that strain is the main parameter controlling the velocity change, representing 38-49% of the modeled velocity change, with strain sensitivity of 2x102, 3.4x102 and 3.6x102 strain-1 for the 0.5-1 Hz, 1-2 Hz and 2-4 Hz frequency bands, respectively. At Asamayama, for the 0.5-1 Hz band, despite a relatively poor fit, the velocity change was determined to results mainly from pore pressure (48%) and strain change (37% with a sensitivity of 2x102 strain-1). For the 1-2 Hz band, we found the velocity change to results mainly from the combination of temperature (50%), pore pressure (24%) and strain (22%; 3x102 strain-1). While the 2-4 Hz band seems to rather show relations with temperature (47%) and sea level (24%) variations. The remaining contribution being distributed between pore pressure, strain and snowfall.
Our preliminary results demonstrate the applicability of our approach to retrieve the different component contributing to the measured velocity change while determining coherent strain sensitivity values. This would allow (1) subsequent use of the velocity change as a stress proxy for vent location forecasting and (2) detection of potential small eruptive precursor by removing the modeled contributions.