ABSTRACT
Currently available earthquake early warning systems employ region-specific empirical relations for magnitude determination and ground-motion prediction. Consequently, the setting up of such systems requires lengthy calibration and parameter tuning. This situation is most problematic in low seismicity and/or poorly instrumented regions, where the data available for inferring those empirical relations are scarce. To address this issue, a generic approach for real-time magnitude, stress drop, and ground-motion prediction is introduced that is based on the omega-squared model. This approach leads to the following approximate expressions for seismic moment: M0∝RT0.5Drms1.5/Vrms0.5, and stress drop: Δτ∝RT0.5Arms3/Vrms2, in which R is the hypocentral distance; T is the data interval; and Drms, Vrms, and Arms are the displacement, velocity, and acceleration root mean squares, respectively, which may be calculated in the time domain. The potential of these relations for early warning applications is demonstrated using a large composite data set that includes the two 2019 Ridgecrest earthquakes. A quality parameter is introduced that identifies inconsistent earthquake magnitude and stress-drop estimates. Once initial estimates of the seismic moment and stress drop become available, the peak ground velocity and acceleration may be estimated in real time using the generic ground-motion prediction equation of Lior and Ziv (2018). The use of stress drop for ground-motion prediction is shown to be critical for strong ground accelerations. The main advantages of the generic approach with respect to the empirical approach are that it is readily implementable in any seismic region, allows for the easy update of magnitude, stress drop, and shaking intensity with time, and uses source parameter determination and peak ground motion predictions that are subject to the same model assumptions, thus constituting a self-consistent early warning method.
Real-time Respiratory Tumor Motion Prediction based on Temporal Convolutional Neural Network (Preprint)
