The linked stress release model for spatio-temporal seismicity: formulations, procedures and applications

The linked stress release model is based on the build-up of stress through elastic rebound and its dissipation in the form of earthquakes. In addition, stress can be transferred between large-scale geological or seismic features. The model can be statistically fitted to both historical and synthetic seismicity catalogues and, through simulation, can be used to create probabilistic forecasts of earthquake risk. We review the genesis of the model, provide some observations on forecasting using the model, and follow with a comprehensive review of applications to date. A systematic procedure for identification of the best model is illustrated by data from the Persian region. We then consider the evaluation of fitted models, using residual point processes and information gains. Implications of the use of Benioff strain rather than seismic moment are discussed. The sensitivity of the model to regionalization, magnitude errors, catalogue incompleteness, catalogue size and declustering/magnitude cut-off is then considered in detail with reference to data from north China. The latter data are also used to illustrate the model evaluation techniques introduced earlier. Some technical material on numerical fitting, simulation and calculation of the information gain is given in an appendix.