A Statistical Method for Associating Earthquakes with Their Source Faults in Southern California

Abstract

We present a new statistical method for associating earthquakes with their source faults in the Southern California Earthquake Center’s 3D Community Fault Models (CFMs; Plesch et al., 2007) in near‐real time and for historical earthquakes. The method uses the hypocenter location, focal mechanism orientation, and earthquake sequencing to produce the probabilities of association between a given earthquake and each fault in the CFM as well as the probability that the event occurred on a fault not represented in the CFM. We used a set of known likely associations (the Known Likely Sets) as training or testing data and demonstrated that our models perform effectively on these examples and should be expected to perform well on other earthquakes with similar characteristics including the full catalog of southern California earthquakes (Hauksson et al., 2012). To produce near‐real‐time associations for future earthquakes, the models have been implemented as an R script and connected to the Southern California Seismic Network data processing system operated by the California Institute of Technology and the U.S. Geological Survey to automatically produce fault associations for earthquakes of M ≥ 3.0 as they occur. To produce historical associations, we apply the method to the most recent CFM version (v.5.2), yielding modeled historical associations for all events of M ≥ 3.0 in the catalog of southern California earthquakes from 1981 to 2016. More than 80% of these events and 99% of moment within the geography covered by the CFM had a primary association with a CFM fault. The models can help identify clusters of small earthquakes that indicate the onset of activity associated with major faults. The method will also assist in communicating objective information about the faults that source earthquakes to the scientific community and general public. In the event of a damaging southern California earthquake, the near‐real‐time association will provide valuable information regarding the similarity of the current event to forecast scenarios, potentially aiding in earthquake response.