Joint analysis of seismic velocity change, deformation and meteorological data for volcano monitoring

Detection of subtle precursors to forecast the timing and location of eruptive events are one of the main issues in volcanic risk mitigation. Seismic interferometry of ambient noise allows to detect very small change in the medium, but is affected by numerous processes such as strain and meteorological variations that need to be identified and characterized in order to correctly interpret the observation for operational monitoring. To assess the monitoring potential of the seismic velocity changes, we jointly analyzed 10 year 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 velocitystrainenvironmental interactions. Daily seismic velocity changes were computed considering three frequency bands (0.5-1 Hz, 1-2 Hz and 2-4 Hz). Daily meteorological data were provided by the Japan Meteorological Agency (JMA). Deformation data include (1) areal strain computed from the GNSS stations of the JMA and Geonet network (from Geospatial Information Authority of Japan), and (2) numerically computed tidal strain by using GOTIC2 software (Matsumoto et al., 2001). We assumed the velocity change to results from the linear combination of the strain and meteorological changes (including rainfall, snow load, temperature, atmospheric pressure, wind speed and sea level variations), and inverted their respective contributions to the observed velocity variation using a least-squares method. Over the 21 volcanoes, the inversion of the parameters are able to explain 20-30% of the data in average. We determine that areal strain and temperature have, on average, a high impact on velocities (each representing 20% of the modeled velocities) but also shows a high variability from volcano to volcano, while pore pressure and sea level variations, which shows almost same amount of contribution to the velocities (20%), have a much lower variability indicating a background effect found at more or less all volcanoes. Atmospheric pressure and wind speed show similar behavior but at a lower level (10%). Snow have a relatively low impact over the whole year (5%) but it strongly increases during winter (up to 30%). Tidal strain do not demonstrate significant effect on daily velocity variations. In the details of each volcano, results demonstrate the possibility to retrieve the different component contributing to the measured velocity change. This shows encouraging results for several subsequent application such as: (1) detection of small eruptive precursor by removing the modeled contributions, (2) use of the velocity change as a stress monitoring proxy based on the determined strain sensitivity, (3) monitoring pore pressure change especially for lava dome stability. We also highlight the need for collocated seismic and meteorological sensors to achieve higher accuracy in the environmental effect correction.