A new approach to earthquake early warning

Abstract

Earthquake early warning systems utilize the capability of modern real-time systems to process and transmit information faster than seismic waves propagate (3-6 km/sec); they provide first information on forthcoming ground shaking prior to the arrival of seismic waves at potential user sites. The possible warning time ranges up to 70 seconds, depending on the distances between the seismic source, sensor and user sites. The main effort in the design and implementation of early warning systems in the last few years aims at increased warning times and thus at an expansion of areas that can be effectively warned. This requires new strategies and methods for the rapid characterization of earthquakes: on-site warning systems are based on observations at single stations for the quick estimate of source parameters. The on-site warning approach is significantly faster than network based regional strategies for early warning even though often at the expense of accuracy. We have developed an alternative method for earthquake early warning -called PreSEIS (pre-seismic shaking) -that is as fast as the on-site warning approach but combines information from several sensors within small seismic networks with apertures of about 100 kilometers to estimate source parameters from the first few seconds of seismic recordings. Therefore, it is not necessary for seismic waves to have arrived at all sensors before estimates can be issued because even a non-triggered station provides valuable information about the possible location of the seismic source and confines the space of possible solutions to the problem of a magnitude-distance trade-off. With on-going time, longer time series at more stations become available, allowing improved information to be inferred about the earthquake. PreSEIS is based on artificial neural networks. The algorithm will be demonstrated using the example of the mega-city Istanbul, Turkey. © 2007 Springer-Verlag Berlin Heidelberg.