Random forest parameterization for earthquake catalog generation

An earthquake is the vibration pattern of the Earth’s crust induced by the sliding of geological faults. They are usually recorded for later studies. However, strong earthquakes are rare, small-magnitude events may pass unnoticed and monitoring networks are limited in number and efficiency. Thus, earthquake catalog are incomplete and scarce, and researchers have developed simulators of such catalogs. In this work, we start from synthetic catalogs generated with the TREMOL-3D software. TREMOL-3D is a stochastic-based method to produce earthquake catalogs with different statistical patterns, depending on certain input parameters that mimics physical parameters. When an appropriate set of parameters are used, TREMOL-3D could generate synthetic catalogs with similar statistical properties observed in real catalogs. However, because of the size of the parameter space, a manual searching becomes unbearable. Therefore, aiming at increasing the efficiency of the parameter search, we here implement a Machine Learning approach based on Random Forest classification, for an automatic parameter screening. It has been implemented using the machine learning Python’s library SciKit Learn. This work is partially supported by the Spanish Ministry of Economy and Competitivity under contract TIN2015-65316-P and by the Catalan Government through the programmes 2017-SGR-1414, 2017-SGR-962 and the RIS3CAT DRAC project 001-P-001723. Moreover, this project has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement No 777778 (MATHROCKS). The research leading to these results has received funding from the European Union’s Horizon 2020 research and innovation programme under the ChEESE project, grant agreement No. 823844.