Real-Time Prediction of Ground Motion from P-Wave Records

Real-time prediction of ground motion parameters, such as peak values, seismic intensity or spectral response, at a critical site or facility as well as on regional scale, can help to reduce losses by strong earthquakes by the automatic triggering of protective steps seconds prior to the arrival of high amplitude seismic waves. We present a concept for an earthquake early warning system designed for real-time processing of information obtained from the low amplitude P-wave that is recorded at a number of accelerographs installed in the epicentral area. To allow for the adaptation to rupture dynamics, the estimation of ground motion parameters starts as soon as the first station is triggered and is continuously up-dated with ongoing time. Our design is in contrast to other earthquake early warning systems that are based on the explicit knowledge of magnitude and hypocenter location for the application of empirical attenuation laws. The level of the cumulative absolute velocity (CAV) of the P-wave as well as relative values of CAV at different seismic stations give a good indicator for the severity of impending ground shaking. The site-specific prediction of ground motion parameters and interpolation for the generation of regional shake maps is achieved by the application of a three-layer feedforward neural network considering local site effects. The procedure is demonstrated for the Istanbul Earthquake Early Warning System in Turkey.