Probabilistic earthquake early warning times in Fujian Province

Earthquake early warning （EEW） systems are a new and effective way to mitigate the damage associated with earthquakes. A prototype EEW system is currently being constructed in the Fujian Province, a region along the Southeast coast of China. It is anticipated that the system will be completed in time to be tested at the end of this year （2013）. In order to evaluate how much advanced warning the EEW system will be able to provide different cities in Fujian, we established an EEW information release scheme based on the seismic monitoring stations distributed in the region. Based on this scheme, we selected 71 historical earthquakes. We then obtained the delineation of the region＇s potential seismic source data in order to estimate the highest potential seismic intensities for each city as well as the EEW system warning times. For most of the Fujian Province, EEW alarms would sound several seconds prior to the arrival of the destructive wave. This window of time gives city inhabitants the opportunity to take protective measures before the full intensity of the earthquake strikes.

Interpretability and spatial efficacy of a deep-learning-based on-site early warning framework using explainable artificial intelligence and geographically weighted random forests

Earthquakes pose significant risks globally,necessitating effective seismic risk mitigation strategies like earthquake early warning(EEW)systems.However,developing and optimizing such systems requires thoroughly understanding their internal procedures and coverage limitations.This study examines a deep-learning-based on-site EEW framework known as ROSERS(Real-time On-Site Estimation of Response Spectra)proposed by the authors,which constructs response spectra from early recorded ground motion waveforms at a target site.This study has three primary goals:(1)evaluating the effectiveness and applicability of ROSERS to subduction seismic sources;(2)providing a detailed interpretation of the trained deep neural network(DNN)and surrogate latent variables(LVs)implemented in ROSERS;and(3)analyzing the spatial efficacy of the framework to assess the coverage area of on-site EEW stations.ROSERS is retrained and tested on a dataset of around 11,000 unprocessed Japanese subduction ground motions.Goodness-of-fit testing shows that the ROSERS framework achieves good performance on this database,especially given the peculiarities of the subduction seismic environment.The trained DNN and LVs are then interpreted using game theory-based Shapley additive explanations to establish cause-effect relationships.Finally,the study explores the coverage area of ROSERS by training a novel spatial regression model that estimates the LVs using geographically weighted random forest and determining the radius of similarity.The results indicate that on-site predictions can be considered reliable within a 2–9 km radius,varying based on the magnitude and distance from the earthquake source.This information can assist end-users in strategically placing sensors,minimizing blind spots,and reducing errors from regional extrapolation.

A τ_c magnitude estimation of the 20 April 2013 Lushan earthquake, Sichuan, China

A crucial part of proposed earthquake early warning systems is a rapid estimate for earthquake magnitude.Most of these methods are focused on the first part of the P-wave train,the earlier and less destructive part of the ground motion that follows an earthquake.A method has been proposed by using the period of the P-wave to determine the magnitude of a large earthquake at local distance,and a specific relation for the Sichuan region was calibrated according to acceleration records of Wenchuan earthquake.The Mw 6.6 earthquake hit Lushan County,Sichuan,on April 20,2013 and the largest aftershocks provide a useful dataset to validate the proposed relation and discuss the risks connected to the extrapolation of magnitude relations with a poor dataset of large earthquake waveforms.A discrepancy between the local magnitude(ML)estimated by means ofτc evaluation and the standard ML(6.4 vs.7.0)suggests using caution when ML vs.τc calibrations do not include a relevant dataset of large earthquakes.Effects from large residuals could be mitigated or removed by introducing selection rules onτc function,by regionalizing the ML vs.τc function in the presence of significant tectonic or geological heterogeneity,and by using probabilistic and evolutionary methods.