Collaboratory for the Study of Earthquake Predictability: Design of Prediction Experiments

The Collaboratory for the Study of Earthquake Predictability (CSEP) aims to improve our understanding of earthquake physics and predictability through rigorous, prospective testing of earthquake forecast models. The system-science character of earthquake prediction research demands an open and collaborative structure for experimentation in a variety of fault systems and tectonic settings. Several CSEP Testing Centers are now available to provide adequate infrastructure for predictability research. The first began operations at the Southern California Earthquake Center on 1 September 2007, running automated evaluations of daily forecast models for California and monitoring the five-year forecasts of the Regional Earthquake Likelihood Models (RELM) project. Currently, CSEP testing centers are operational in the United States, New Zealand, Japan, and Switzerland; within these centers, prediction experiments are being conducted in regions covering California, the Western Pacific, New Zealand, Japan, Italy, and the globe. More than 80 prediction models are under test and CSEP is expanding into new areas and new experiments. In this presentation, we focus on the design of rigorous and truly prospective prediction experiments, in which all input data are fully specified, the prediction and evaluation rules are pre-defined, and the entire process is transparent. We also discuss the importance of being able to reproduce experiment results and describe how earthquake prediction researchers can benefit from the CSEP approach.