Comparison of CNN and CNN-LSTM Algorithms based on Earthquake Magnitude Estimation

After the earthquake, the rapid determination and estimation of earthquake magnitude is one of the most important tasks of the earthquake early warning system. In the model construction of magnitude determination, it is necessary to comprehensively consider the seismograph response time, the error of arrival time selection, and the inaccurate velocity model, which will lead to the incompleteness of the current earthquake catalog and the insufficient accuracy of magnitude determination. In this paper, based on the STanford EArthquake Dataset (STEAD), convolution neural network (CNN) and long short-term memory (LSTM) are mainly used for fast estimation. This paper compares the timeliness and accuracy of the two algorithms for magnitude assessment. By comparing the characteristics of absolute error standard deviation and mean value, it provides a new idea for the improvement of subsequent magnitude determination methods to better enhance the disaster reduction effect of earthquake early warning.