Neural dynamic modelling on earthquake magnitude series

Abstract

Earthquake magnitude prediction is of vital importance for human safety. The earthquake is a very complicated and non-linear dynamic process. It cannot be described adequately by any deterministic models. In this paper a neural dynamic modelling for earthquake magnitude prediction is reported. Historical records of earthquake magnitude series are used to construct the optimal non-linear dynamic model, and the consequent outcome of the earthquake behaviour is then predicted by this model. In turn, the latest recorded data set can be fed back to improve the accuracy of the neural dynamic model. The modelling of experiments of three earthquake magnitude series in China and Japan and their extrapolated predictions are included in this paper. The values predicted by extrapolation are in good agreement with the historical data.