Magnitude estimation based on relative power in MEMS-based earthquake early warning system

The cost and performance of earthquake early warning stations are always a trade-off. The high cost of traditional early warning stations limits their large-scale application. The emergence of low-cost MEMS accelerometers makes it possible to build dense earthquake early warning networks. However, the high noise level of MEMS accelerometers cannot be ignored, which affects the performance of magnitude estimation. In this paper, we propose a magnitude estimation method based on relative power which can directly use acceleration records for calculation without the need for integration and filtering. The attenuation relationship of magnitude, relative power and hypocentral distance can be derived from the statistical analysis of some seismic events. Comparative experiments showed that the magnitude estimation performance of Power method is comparable to the method and better than the method. In noise analysis, the Power method demonstrates superior performance in low Signal-to-Noise Ratio data. After station correction, the Power method can achieve robust magnitude estimation results with only five stations triggered following an earthquake. We validated the generalizability of this method using seismic data from Mexico.