Resumen de Contributions to probabilistic earthquake forecasting
Alvaro González
This thesis presents a set of contributions to probabilistic earthquake forecasting, both spatial and temporal.
A novel empirical method with no parameter is used to show that the distances between past earthquakes forecast, in a probabilistic way, the distance to the next earthquake. This procedure is first applied retrospectively to earthquake catalogues of the whole Earth, Southern California and the Iberian region. The resulting probabilistic maps for forecasting the locations of forthcoming earthquakes self-sharpen when updated as new earthquakes occur.
Within the Collaboratory for the Study of Earthquake Predictability, the Southern California Earthquake Center has tested the method prospectively, in an independent and automatic way, calculating daily forecast maps. The results of six years of tests for the whole Earth, California and western Pacific show that the method outperforms the one typically used as a reference.
The earthquake catalogue of the Spanish Instituto Geográfico Nacional, used here for forecast testing in the Iberian region, is analysed in detail with regard to its development, location precision, magnitude of completeness and inclusion of artificial blasts. Spatial and temporal heterogeneities, and overall progressive improvements, are highlighted.
An efficient technique is devised for measuring the areas on the map where each new earthquake is expected. The spherical Fibonacci lattice, a highly homogeneous spiral pattern, is proposed for measuring complex areas on a spherical surface by point counting. Measurement errors are greatly reduced compared to those obtained with latitude-longitude lattices.
A new stochastic model is introduced as an idealization of the seismic cycle of a fault, and applied to the series of earthquakes generated by the San Andreas Fault at Parkfield (California). Along with other, statistical renewal models, it is used to calculate time-depending probabilities for the occurrence of the next large earthquake, and to test a simple forecasting strategy.
Finally, a method is devised to try to synchronize numerical models with the seismic faults that they simulate, by forcing them to reproduce the observed sequence of earthquakes. This procedure is able to partially synchronize models with another, stochastic one, and to forecast its largest synthetic earthquakes more efficiently than with simpler approaches.
