Incorporate Seismic Activity Prior Information to Earthquake Early Warning through Bayesian Framework

Abstract:

Most of the current Earthquake Early Warning technologies focus on time analysis of wave amplitudes. There are two major drawbacks of these waveform-based techniques: tradeoffs between magnitude and distance estimation for the onsite algorithms, and time latency in alerts for the network algorithms. We are proposing an alternative EEW algorithm that combines the efficiency of onsite algorithms and accuracy of network algorithms, which provides the fastest alert at the moment of station trigger. It is achieved by using observed seismicity from the network as prior information to predict short-term seismic hazards, and then use trigger information from the onsite station as likelihood information to estimate earthquake probability and hypocenter location. This algorithm has numbers of advantages. First, due to the independent data source of this algorithm, results can be directly multiplied to the results of other algorithms such as GPS and waveform data under Bayesian framework to achieve posterior probability function. Second, it is especially beneficial for regions with sparsely distributed station density where it takes longer time for the seismic signals to arrive at the near stations. Lastly, it can significantly speed up warning process during aftershock sequence, swarm earthquake sequence, and mainshocks that had foreshocks. The concept can be further extended to network-based algorithms to incorporate arrived waveform data at more stations.