An adaptive Butterworth spectral-based graph neural network for detecting ionospheric total electron content precursor prior to the Wenchuan earthquake on 12 May 2008

Abstract

In this study, Global Ionosphere Maps (GIMs) were used to detect all total electron content (TEC) precursors prior to the M_w 7.9 Wenchuan earthquake at 06:28:01 (UTC) on 12 May 2008. The GIMs were converted into GIM-gis between 00.00 (UTC) on 01 December 2004 and 31 March 2005 and were used to build the Spectral-Based Graph Neural Network (SB-GNN) and Adaptive Butterworth Spectral-Based Graph Neural Network (ABSB-GNN) models. A blurred TEC precursor was detected at 2:20 (UTC) on 9 May 2008, i.e. 3 days prior to the Wenchuan earthquake, using the ABSB-GNN model in place of the SB-GNN model, which improved predictive ability. After evaluating the variance in the learning process, the two models were verified for stability with better training and predicted accuracies in the ABSB-GNN model, which might be the reason that a blurred TEC precursor was detectable using the ABSB-GNN model.

Keywords:

• Global ionosphere maps
• total electron content precursors
• GIM-gis
• spectral-based graph neural network
• adaptive Butterworth spectral-based graph neural network
• Wenchuan earthquake

Acknowledgments

The author is grateful to the Jet Propulsion Laboratory, NASA, Global Differential GPS system (GDGPS) and to the World Data Center for Geomagnetism, Kyoto (Nose et al. Kamei [2015], Geomagnetic Dst index, doi: 10.17593/14515-74000). The author is also grateful to several geomagnetic observatories (Kakioka [JMA], Honolulu and San Juan [USGS], Hermanus [RSA], INTERMAGNET, and many others) for providing the final Dst index data. The author is also grateful to Dr. Primoz Potocnik for providing the MATLAB code. The author dedicates this paper to the memory of his mother, who passed away on 9 October 2016.

Disclosure statement

No potential conflict of interest was reported by the authors.

Data availability

The TEC map, data and information can be obtained from the NASA Global Differential GPS system (GDGPS)

(http://www.gdgps.net/products/tec-maps.html)

(https://iono.jpl.nasa.gov/latest_rti_global.html).

The Dst indices can be obtained from http://wdc.kugi.kyoto-u.ac.jp/dst_final/index.html (World Data Centre for Geomagnetism, Kyoto, Japan).

MATLAB code source

http://lab.fs.uni-lj.si/lasin/wp/IMIT_files/neural/NN-examples.pdf

The author modified the code for the objectives of this study.

Author contributions

Jyh-Woei Lin designed the research, performed the research, analyzed the data, and authored the paper.