HR: 0800h
AN: NG31A-0869 [Abstracts]
TI: An Earthquake Prediction System Using The Time Series Analyses of Earthquake Property And Crust
Motion
AU: * Takeda, F
EM: takeda@tec21.jp
AF: Takeda Eng Consultant Inc., 2-14-23, Hiroshima, 734-0015
Japan
AB:
An earthquake (EQ) phenomenon has evidence of deterministic chaos. A total of about 17200 EQ's with magnitude M $>$= 4 are
collected from a catalogue of Japan Meteorological Agency from 1983 to June of 2004 for a region of (24 deg - 48 deg N, 124
deg - 150 deg E). Sequencing the five EQ properties of every EQ, which are latitude (LA), longitude (LN), depth (DP),
inter-EQ time interval (IT) and M, in EQ occurrence (event) order, we find that each of the five EQ property series is
deterministic chaos.[1,2] For example, the largest Lyapunov exponents are all positive values, statistically distinct from
those of the series surrogated by randomly shuffling the event order. Each property series in chronological event order i, is
given by
Da = [Da,1, Da,2, . , Da,i-1, Da,i, Da,i+1, .]. (1)
Here index a stands for each property (a = LT, LN, DP, IT and M). Each minimum embedding dimension (ED) of Eq. 1 is only five
(which happens to be our five EQ properties). It is estimated by finding the ED at which the percentage of false nearest
neighbors drops to a constant residual level. The residue is indicative of the contamination level of dynamical noise created
by many other seismogenic variables. To reduce the residue of each series to zero, we take the moving average (denoted by
Da,m,w) of each property element of Eq. 1 over w consecutive events with respect to event i running from m - w + 1 to m. To
extract the rate of change in Da,m,w, we use the second order difference at an event-separation s, defined as Aa,m,w,s =
(Da,m-2s,w - 2 Da,m-s,w + Da,m,w). To extract the determinism in seismogenic evolutions (quiescence cycles) to only large
EQ's, we use the moving sum (CI-w) of element Da,i (a=IT) over w consecutive events from m - w + 1 to m. With an appropriate
choice of w (for example, w = 100), CI-w reveals the deterministic measures that are absent in its surrogate. Thus the
analysis of deterministic Da,m,w, Aa,m,w,s and CI-w should enable us to predict the time, focus and M of a large EQ within
narrow limits in weeks or months (or a few years) ahead of time.
We have developed such a short-term deterministic EQ forecasting system similar to those used for Typhoons and Hurricanes,
which has been under a test operation at http://www.tec21.jp/ since June of 2003. [1,2] The system must divide Japan into a
small area of about 5 deg by 5 deg for which Eq. 1 is reconstructed such that the regional CI-w and Aa,m,w,s (whose ED's are
all reduced to three) become the deterministic measures to make the successful prediction tests on every large EQ (or swarm
if any) of hindsight in the area. The current system has five areas to cover Japan. We focus on two examples to describe the
system operation. One is the 2003/09/26 Tokachi EQ of M 8 in hindsight (the system was yet to cover this area at the event).
To make a prediction test on the M 8 EQ, we assume our current date is back to 2003/08/31. The date prediction of the M
8-class EQ was only one day off. The focus prediction was within the very narrow limits. Another is a successful rollout of
the most recent forecast on the 2004/05/30 EQ of M 6.7 off coast of the southern Kanto (Tokyo) area. Its forecasting at our
website started 16 days ahead of time like forecasting an approaching typhoon. The final forecast issued one day before the
actual event has confirmed a perfect prediction on the focus and date. The system also performs a state-space analysis of the
daily crust motion observed by GPS (F1 data of GEONET) to detect any anomaly including a slow-slip motion precursory to the
M 8 EQ.
1. F. Takeda, and M. Takeo, submitted to Chaos (2003).
2. F. Takeda, and M. Takeo, submitted to AIP Conf. Proc., ECC8 (2004).
UR: http://www.tec21.jp
DE: 7223 Seismic hazard assessment and prediction
DE: 7200 SEISMOLOGY
DE: 3040 Plate tectonics (8150, 8155, 8157, 8158)
DE: 3200 MATHEMATICAL GEOPHYSICS (New field)
DE: 3220 Nonlinear dynamics
SC: Nonlinear Geophysics [NG]
MN: 2004 AGU Fall Meeting