Worldwide earthquake forecasts

Abstract

We review global earthquake forecast as it was developed by our UCLA Group since 1970s. We discuss the new results on earthquake magnitude/moment distribution, especially the determination of maximum/corner moment magnitude. Recent successes of global geodetic surveys allow us to construct a high resolution map of the Earth surface displacement and, after converting this map to earthquake rate, combine these maps with those based on seismicity smoothing. Thus we are constructing and testing a global earthquake activity rate (GEAR) model as a function of the magnitude at 0.1 by 0.1 $$^{\circ }$$ resolution. Nicknamed “GEAR”, the model relies on a global strain rate model and the instrumental earthquake catalogs. Thus, for our forecast we use datasets that provide uniform global coverage: seismic catalogs, global plate boundary models, and global positioning system geodetic velocities. After testing, the model can be used by Global Earthquake Model (GEM) Foundation and others in seismic hazard estimation. GEAR covers magnitudes 5.8 and larger, the periods are from years to decades with no explicit time-dependence. The normalized magnitude distribution at each location is a combination of the tapered Gutenberg–Richter distributions with b-values and the corner magnitudes determined by a few global parameters that depend on the tectonic style and the focal mechanism properties. The seismic and strain-rate components of the model are specified separately and then combined optimally to fit an earthquake occurrence over the last several years. GEAR performs well in quasi-prospective tests using the global centroid–moment–tensor catalog after 2005 and the GEM catalog from 1918 to 1976; GEAR will be rigorously tested prospectively against future earthquakes. Because of its simplicity, GEAR can serve as a well-vetted reference model and as a null-hypothesis against which more complex models can be tested. With its high spatial resolution, GEAR can also be compared to detailed regional models. We also discuss potential methods to improve and extend the forecasts. Such methods include extending the forecast to lower magnitudes, introducing focal mechanisms into prediction methods, and making the short-term forecast an integral part of the practical forecast implementation.