You are using an old browser with security vulnerabilities and can not use the features of this website.
Timescale: Oct.2015 – Sept.2018
Dr. Sebastian Hainzl, Helmholtz Centre Potsdam GFZ German Research Centre For Geosciences
Prof. Dr. Matthias Holschneider, University of Potsdam
PD Dr. Gert Zöller, University of Potsdam
Natural seismicity is a non-stationary process with various kinds of transient behaviour on different spatio-temporal scales, e. g., aftershocks, foreshocks, swarm activity, and quiescence on time-scales from hours to decades. Man-made earthquakes induced, e. g., by fluid injection in geothermal areas or waste-water disposals, share similar statistical features, but on a smaller spatial scale with transient boundary conditions. The knowledge of the transition times and characteristics is in both cases of particular importance for the estimation of the current state of the seismic hazard. The project aims at improved identification of these transients by developing new statistical methods based on Bayesian methods. This allows for a straightforward assessment of uncertainties, which is crucial in the light of the overall low number of data points in typical earthquake catalogues.
Objectives and Methods
The ultimate goal of the project will be the design of appropriate statistical models for transient periods in seismicity, which includes deterministic (e.g., anthropogenic, via triggering of earth-quakes through fluid injection) in addition to random deviations from stationarity. Achieving this goal requires a sound mathematical framework for the estimation of seismological parameters from the available data that are sparse in many cases.
The project aims at designing appropriate statistical methods for the identification of transient periods in seismicity in space and time, which include natural and anthropogenic (e.g. earthquakes triggered by fluid intrusion) deviations from stationarity. Methods of statistical and physical modelling of seismicity will be used along with modern techniques of Bayesian inference, uncertainty assessment and model selection. We will use stochastic point processes to model seismicity; the identification of transients including uncertainty assessment will be carried out with Bayesian techniques including parameter estimation, model comparison, and selection as well as change-point studies. Information criteria and the Bayes factor will be used for model selection. Physical models might also be useful in order to produce synthetic data for comparison and testing purposes.
Bernhard Fiedler is based at the research teams “Applied Mathematics” of the University of Potsdam and “Physics of Earthquakes and Volcanoes” of the GFZ German Research Centre for Geosciences.