The Pick Aggregator is an open-source ensemble ML workflow that aggregates phase picks from multiple deep learning pickers and ML models to produce high-precision earthquake catalogs. By combining the outputs of ML models with pickers such as EQTransformer and PhaseNet through a voting and aggregation scheme, it achieves pick precision that generalizes to out-of-distribution data, making it particularly suited for hybrid and irregular seismic networks, including combined land and ocean bottom seismometer deployments.

I've tested Pick Aggregator on OBS-land records offshore Puerto Rico and Virgin Islands and results will be published in Seismological Research Letters. Current application includes monitoring natural and induced seismicity in New Mexico.