Towards a comprehensive High Resolution 3D P- and S-Wave Velocity Model for the Alpine Mountain Chain using Local Earthquake Data

Seismic data availability and automated picking algorithms drastically improved in the European Alps since the last orogen wide crustal P-wave velocity model was compiled by Diehl et al. (2009). Especially, the abundant seismic data recorded by the AlpArray Seimic Network (AASN) which was in operation from 2015-2021 provides a unique high resolution seismic data set. The aim of our project therefore is to create a comprehensive 3D P- and S-wave crustal velocity model for the European Alpine region using Local Earthquake Tomography (LET). Such a model is not only needed to sharpen high resolution teleseismic tomography studies imaging subducted slabs but also to relate surface structures to mountain building processes in the mantle.
To achieve this aim precise onset times of seismic crustal phases are needed. Here we show our first results of automatic onset time determination obtained through the deep-neural-network PhaseNet. When compared to catalogues of manual travel time picks, we find its performance as accurate as a human analyst's. This confirms the transferability of machine learning approaches to our area and data set.
The large amount of evenly distributed seismic stations yields up to a total of 720 P and S arrival picks with epicentral distances up to 700km for events with ML > 3.5. Earthquakes with magnitudes of ML=2.5 are generally detectable for epicentral distances up to at least 200km and contribute approximately 200-300 arrivals per event.
As a first step towards a 3D model we present a thorough analysis of the consistency of the automatically determined arrival times, which facilitates a reliable removal of outliers. 
Furthermore, we show visualizations of our preliminary tomography model and its resolution.