On modeling nonstationary geomagnetic signal

Based on the smoothness priors constraint, an approach for modeling time series with nonstationary mean is employed in analyzing geomagnetic data during earthquakes. The overall model is fitted by using the Bayesian idea, the Kalman filter and Akeike's AIC modeling criterion. The trends and irregular components of geomagnetic signals during earthquakes can be modeled simultaneously and the spectrum of irregular components is obtained to analyze frequency features of the geomagnetic field during earthquakes.