A Magnitude Estimation Approach and Application to the Yunnan Earthquake Early Warning Network

Abstract

Estimating the magnitude based on P waves is important to issue early messages to the general public for earthquake early warning (EEW) purposes. First, we reviewed part of the magnitude estimation results of the current EEW system deployed in Yunnan and found that the magnitude of the mainshock of the Yangbi earthquake sequence (Ms 6.4, as formally determined by the Yunnan Seismic Network) is severe underestimated by 0.9, whereas a number of small earthquakes are overestimated. Then, we determined the reason for this phenomenon and improved the magnitude estimation precision in Yunnan by introducing the fixed effect regression technique, adopting a flexible time window and deriving the magnitude estimation equation strictly. Finally, we examined the deficiencies of this magnitude estimation approach and demonstrated that it is possible to provide instant and reliable parameters of an earthquake through numerous inexpensive accelerometers on the ground.

Intensity Prediction Model Based on Machine Learning for Regional Earthquake Early Warning

“…In EEWs, seismic parameters (magnitude, location, etc.) are obtained by analyzing the information contained within the initial P wave time window [50]- [51]. As the length of the P waves time window increases, more and more information are contained and the seismic parameters are estimated more accurately.…”

“…We simulate this process of gradually decreasing prediction error and test it on Japanese Application Dataset. It is worth noting that we more realistically reproduce the effect of magnitude estimation, which the underestimation of large earthquakes and overestimation of small earthquakes [50]- [51]. We add a biased small error for earthquakes with magnitude greater than 5.7 and a biased large error for earthquakes with magnitude less than 5.7.…”