Spatio-temporal changes in seismic velocity associated with the 2000 activity of Miyakejima volcano as inferred from cross-correlation analyses of ambient noise

Miyakejima volcano, Japan, showed strong and interesting volcanic activities in 2000 with the occurrence of dike intrusions, caldera formation and a large amount of gas emission. To understand temporal changes of the volcanic structure associated with this activity, we apply ambient noise correlation analyses to the seismic records of five seismic stations at Miyakejima volcano from July 1999 to December 2002. We calculate cross-correlation functions (CCFs) for eight available station pairs at frequency bands of 0.4-0.8Hz and 0.8-1.6Hz, and retrieve the Rayleigh waves propagating with a group velocity of about 1.0km/s. Comparing CCFs obtained for each day with the reference CCFs that are calculated from a few months of data, we estimate seismic velocity changes before and after the 2000 activity. Results at the two frequency bands show seismic velocity increase and decrease. Shallow structure at the volcanic edifice on the flanks experienced a seismic velocity increase up to 3.3%. On the other hand, the regions located close to the collapsed caldera show a seismic velocity decrease down to 2.3%. To understand the mechanisms that may introduce these seismic velocity increase and decrease, we examine several possible mechanisms of the velocity changes: stress changes due to volcanic pressure sources and caldera formation, and topographic changes. By using the deflation sources previously determined from GPS data, we calculate the stresses at depths of 0km and 1km at the volcano. The results suggest that the seismic velocity increases observed at seismic paths on the volcanic flank are explained by the compression caused by the deflation sources associated with magma activity in 2000. Stress sensitivities of the seismic velocity changes are estimated to be 1.5×10^-2 to 2.2×10^-2MPa^-1 and 4.6×10^-3 to 2.4×10^-2MPa^-1 at frequency bands of 0.4-0.8Hz and 0.8-1.6Hz, respectively, which are in the range of the previously reported results. Our three dimensional numerical calculations on seismic wave propagation show that the seismic velocity decreases observed at seismic paths crossing the collapsed caldera can be explained by topographic changes introduced by a caldera. We further suggest that the seismic velocity decreases detected around the regions close to the caldera may originate from the dilated regions cause by formation of the caldera.