SCEC Award Number 16068 View PDF
Proposal Category Collaborative Proposal (Integration and Theory)
Proposal Title CSEP participation, including the Canterbury experiment, testing of non-Poisson models, and effects of catalogue deficiencies
Name Organization
Matthew Gerstenberger GNS Science (New Zealand) David Rhoades GNS Science (New Zealand) Annemarie Christophersen GNS Science (New Zealand) David Harte GNS Science (New Zealand) Maria Liukis University of Southern California
Other Participants Maria Liukis, SCEC/USC
SCEC Priorities 2b SCEC Groups CSEP, EFP, Seismology
Report Due Date 03/15/2017 Date Report Submitted 10/18/2016
Project Abstract
The award was for travel to the SCEC annual meeting only. Our associated activities included four main objectives (1) To collaborate with the CSEP team in conducting and analyzing an experiment to retro-spectively test forecasting models based on Coulomb stress calculations using data from the Canter-bury earthquake sequence in the New Zealand CSEP testing center, and to subsequently test the models prospectively on future New Zealand earthquakes. (2) To implement a pilot experiment for non-Poisson testing of one-day models in the New Zealand Earthquake Forecast Testing Centre. At the request of the CSEP working group, we agreed to shift this experiment to the California testing center and provided a non-Poisson ETAS model as a test case. (3) To undertake a retrospective experiment in the New Zealand testing center to evaluate the effect of deficient catalogs on forecasting models. (4) To participate in international CSEP activities, including the design of future experiments, promotion of hybrid modeling capability within CSEP, and preparation of new models for testing in the CSEP framework. Matt Ger-stenberger, David Rhoades and Annemarie Christophersen attended the 2016 SCEC Annual Meeting. David Rhoades visited Masha Liukis at USC for two days after the meeting to collaborate on installation of non-Poisson testing software in the CSEP system.
Intellectual Merit The activities in this project are directly relevant to the SCEC objectives for earthquake forecasting and predictability and development of the CSEP project, including the development of improved forecasting models and testing methods, and the use of hybrid methods to improve forecasting. Our activities are not solely focused on New Zealand, but rather on the general problems of earthquake forecasting and the common international CSEP methodology. Therefore, there are substantial benefits for California re-sulting from our work.
Broader Impacts The collaborative CSEP project is making major advances to the development and testing of earthquake forecasting models on all time scales. Our contributions are a significant component of the overall inter-national collaboration, from which all parties derive mutual benefits. The Canterbury earthquakes have provided a useful data set on which to retrospectively test a range of models, which can subsequently be applied more generally in California and other regional testing centers.
Exemplary Figure Figure 1. Example of a non-Poisson test of CSEP-style grid forecasts with accompanying simulated catalogs. TERD (Target earthquake rate distribution) test for 24-hour ETAS model forecasts beginning one hour before the Christchurch M6.3 earthquake of 22 Feb 2016. The test is calculated using a preliminary cata-log and the final catalog. A Kolmogorov-Smirnov (KS) test compares the distribution of cell rates for target earthquakes expected by the model (calculated from the combination of 5000 simulated catalogs) with the distribution calculated from the actual catalog data. There were 33 target earthquakes in the pre-liminary catalog and 52 in the final catalog. Preliminary catalog data: KS = 0.87, p=10-7; final catalog da-ta: KS = 0.74, p=2×10-6. The low p-values indicate that the model is inconsistent with both the preliminary and final catalogs. The similarity of the distributions calculated from the simulated catalogs shows that the forecast for this 24-hour period was not much affected by catalog deficiencies.