Principal component analysis method in the detection of total electron content anomalies in the 24 hrs prior to large earthquakes

Abstract

The goal of this research is to examine one-dimensional total electron content (TEC) data using principal component analysis (PCA) to search for total electron content (TEC) anomalies associated with large earthquakes in 24 h prior to nucleation. The characteristics of principal eigenvalues generated for TEC prior to 24 earthquakes of magnitude scale M ≥ 5.0 and 6 lesser earthquakes of magnitude scale M < 5 that occurred in Taiwan from 01 January 2000 to 31 December 2001 are examined. In an earlier paper, I was able to confirm the statistical findings of Liu et al. (J Geophys Res 111, 2006) that sparse earthquake-associated TEC anomalies existed in 5 days prior to the 12 large earthquakes they examined (Lin, Terr Atm Ocean Sci, 2010). In this paper, I wish to examine the subtlety of principal component analysis in detecting earthquake-associated TEC anomalies by examining if such precursors can be detected in 24 h prior to large earthquakes. Of the earthquakes examined, TEC anomalies given by clear extreme principal eigenvalues were evident within 24 h of nucleation for 21 of the 24 earthquakes of M ≥ 5.0. After making allowance for the general status of background TEC, it is clear that these extreme principal eigenvalues are representative of earthquake-associated anomalies. For the smaller earthquakes (M < 5), it was not possible to differentiate earthquake-associated anomalies from background effects on TEC status. These new findings confirm the validity of PCA in searching for earthquake-associated TEC anomalies and show that it is subtle enough to detect TEC anomalies within 24 h leading to a large earthquake. If this approach continues to prove successful, it could theoretically be used in real-time prediction of large earthquakes through early detection of earthquake-associated TEC anomalies.