TANG Jun, MAO Wenfei. Multi-Scale ARMA Residual Correction Model for Detecting Pre-earthquake Ionospheric Anomalies[J]. Geomatics and Information Science of Wuhan University, 2019, 44(6): 791-798. DOI: 10.13203/j.whugis20170321
Citation: TANG Jun, MAO Wenfei. Multi-Scale ARMA Residual Correction Model for Detecting Pre-earthquake Ionospheric Anomalies[J]. Geomatics and Information Science of Wuhan University, 2019, 44(6): 791-798. DOI: 10.13203/j.whugis20170321

Multi-Scale ARMA Residual Correction Model for Detecting Pre-earthquake Ionospheric Anomalies

  • To improve the prediction accuracy of the background values of the ionospheric total electronic content (TEC) disturbance, we have proposed multi-scale autoregressive moving average (ARMA) residual correction model (MARCM) by comparing the precision of this method, ARMA model, inter quartile range(IQR)and sliding window method in predicting the ionospheric total electronic content reference background values. The results show that the average relative accuracy of TEC background value predicted by MARCM is 89.78%, which is higher than the ARMA model, IQR and sliding window method with values of 5.18%, 1.41% and 1.42% respectively and show that the percentage of absolute residual values less than or equal to 3.0 TECU predicted by MARCM is 91.67%, significantly better than the other three methods, It is demorstrated that using MARCM to detect pre-earthquake ionospheric anomalies is feasible. We use this method to detect pre-earthquake ionospheric anomalies of the earthquake happened in Lushan on April 20, 2013, and prove the effectiveness of the proposed method. The results show that obviously ionospheric positive anomalies on 9 and 13 days before the earthquake and apparently ionospheric negative anomalies on 1 to 4 days before the earthquake, are most likely to be caused by the earthquake. And positive anomalies mainly emerge in the 08:00 UT to 10:00 UT, the negative anomalies are mainly concentrated in 00:00 UT to 14:00 UT
  • loading

Catalog

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return