Three-dimensional P-wave velocity structure derived from local earthquakes at the Katmai group of volcanoes, Alaska

The three-dimensional P-wave velocity structure beneath the Katmai group of volcanoes is determined by inversion of more than 10,000 rays from over 1000 earthquakes recorded on a local 18 station short-period network between September 1996 and May 2001. The inversion is well constrained from sea level to about 6 km below sea level and encompasses all of the Katmai volcanoes; Martin, Mageik, Trident, Griggs, Novarupta, Snowy, and Katmai caldera. The inversion reduced the average RMS travel-time error from 0.22 s for locations from the standard one-dimensional model to 0.13 s for the best three-dimensional model. The final model, from the 6th inversion step, reveals a prominent low velocity zone (3.6–5.0 km/s) centered at Katmai Pass and extending from Mageik to Trident volcanoes. The anomaly has values about 20–25% slower than velocities outboard of the region (5.0–6.5 km/s). Moderately low velocities (4.5–6.0 km/s) are observed along the volcanic axis between Martin and Katmai Caldera. Griggs volcano, located about 10 km behind (northwest of) the volcanic axis, has unremarkable velocities (5.0–5.7 km/s) compared to non-volcanic regions. The highest velocities are observed between Snowy and Griggs volcanoes (5.5–6.5 km/s). Relocated hypocenters for the best 3-D model are shifted significantly relative to the standard model with clusters of seismicity at Martin volcano shifting systematically deeper by about 1 km to depths of 0 to 4 km below sea level. Hypocenters for the Katmai Caldera are more tightly clustered, relocating beneath the 1912 scarp walls. The relocated hypocenters allow us to compare spatial frequency-size distributions (b-values) using one-dimensional and three-dimensional models. We find that the distribution of b is significantly changed for Martin volcano, which was characterized by variable values (0.8