Seismic Catalogs and Tomographic Velocity Models for the Canary Islands: A Deep Learning Approach

The creation of seismic event catalogs has been revolutionized by the availability of continuous waveform data and the integration of deep learning algorithms for tasks such as event detection, phase picking, association, event localization, and classification. In this study, we showcase the application of these advanced methodologies to generate comprehensive "deep" seismic catalogs for volcanic regions, focusing on the Canary Islands. We also demonstrate how these enhanced catalogs contribute to seismic tomography studies.

 

Our first analysis evaluates the performance of deep learning-based phase pickers when applied to volcano-tectonic events. These pickers, characterized by minimal parameter tuning requirements (typically only a probability threshold for valid picks) offer a significant advantage. However, as they are primarily trained on datasets lacking volcanic events, their sensitivity to such earthquakes may be reduced, and false positives could be more frequent. To address these challenges, we propose a robust workflow combining deep learning-based phase picking, event association, and relocation. This approach yields seismic catalogs that are more complete and accurate compared to those generated using conventional methods.

 

Finally, we utilize these improved seismic catalogs to construct 3D P- and S-wave velocity models for regions within the Canary Islands, including La Palma, as well as the central archipelago's regional structure. These models provide new insights into the subsurface dynamics of this volcanic system.