I. Precursor Identification Seismicity anomaly

1
I. Precursor Identification Seismicity anomaly
Figure 1. Frequency-Magnitude plot for 48835
earthquakes recorded during 1994/1-2002/3 in the
area of (23.6N,25N)x(121.5E,122.5E).

Figure 2. Frequency-Magnitude plot for 21705
earthquakes recorded during 1994/1-1999/9 in the
area of (23.15 N,24.55 N)x(120.12 E,121.52
E).
2
Earthquakes in the northeast and central parts of
Taiwan area
Figure 4. The area (23.15oN, 24.55oN) (120.12oE,
121.52oE) under study includes 7803 M?2
earthquakes with ?the M7.3 Chi-Chi earthquake
occurred on 9/21/99 (921 EQK).
Figure 3. The area (23.6oN, 25oN) (121.5oE,
122.5oE) under study includes 35168 M?2
earthquakes with ?the M6.8 earthquake occurred
on 3/31/02 (331 EQK)
3
Z value and b value

• Z value is the standardized value of the
  difference between two average monthly seismic
  rates over two different time periods, one is
  from 1994/1 to the month under study and the
  other is from the month to a particular
  earthquake, in the region of interest. (Wiemer
  and Wyss, BSSA, 1994)
• b value in the Gutenberg-Richter model is
  computed based on a specified number of
  earthquakes occurred in the region under study.
  (Smith, Nature, 1981)

4
The 331 EQK was preceded by a large b value and a
small z value.

• Figure 5. The b value is computed based on 1000
  events, sliding by 200 events, in the northeast
  part of Taiwan during 1994/1 to 2002/3.

Figure 6. The z value reflects the seismic rate
change between two time periods, one is from
1994/1 to the month under study and the other is
from the month to the 331 EQK, in the northeast
part of Taiwan.
5
The 921 EQK was preceded by a large b value and a
small z value.
day

• Figure 7. The b value based on 100 events,
• sliding by 20 events in central Taiwan
• during 1994/1-1999/9.

• Figure 8. The z value reflects the seismic rate
  change between two periods, one is from 1994/1 to
  the month under study and the other is from the
  month to the 921 EQK in central Taiwan.

6
I. Precursor Identification Ionospheric anomaly

• Using robust statistics of median and
  Inter-quartile range together with an appropriate
  stacking process, earthquake precursor based on
  the maximum plasma density in the ionosphere,
  foF2, is identified as shown in Figure 9.
• The cross-correlation between the occurrences of
  M?5.0 earthquakes and foF2 anomalies (Figure 10)
  indicates that the foF2 anomaly occurs
  significantly 5 days within the earthquakes.
• The successful rate and alarm rate are provided
  in Tables 1 and 2.

7
Ionospheric anomaly

• Figure 9. The occurrences of foF2 anomaly (red
  dot) and M?5.0 earthquakes (blue bar) during
  1994-1999 within 500 km from the Chung-Li
  ionosphere station.

8
Figure 10. The cross-correlation coefficient of
the occurrence of foF2 precursors and that of M?5
earthquakes during 1994-1999.
9
Table 1. The alarm rate of the earthquakes
preceded by foF2 precursors within I days.
Table 2. The successful rate of the earthquake
prediction based on the foF2 precursors followed
by earthquakes within I days during 1994-1999
10
II. Precursor Evaluation Hypothesis testing

• Ho PIP is as good as PSE
• v.s.
• Ha PIP is batter than PSE
• where PIP represents the prediction based on
  the ionospheric precursor and PSE denotes the
  prediction based on the seismic experience.

11
II. Precursor Evaluation Hypothesis testing

• Fit the distribution of the Inter-earthquake time
  during 1991-1993 to be Gamma.

Figure 11. The observed and fitted frequency of
the inter-earthquake time during 1991-1993.
12
II. Precursor Evaluation Hypothesis testing

• The distribution of the Inter-foF2 anomaly time
  during 1994-1999 is fitted by a Gamma
  distribution.

fitted frequency
observed frequency
Weibull distribution with
shape0.871
scale5.184
mean7.09
p-value of the chi-square
goodness-of-fit
frequency
test0.2318
days
Figure 12. The observed and fitted frequency of
the inter-foF2 precursor time during 1994-1999.
13

• Simulation Study A
• Generalize 10,000 sequences of EQK occurrence.
• Calculate successful rate for each sequence.
• Find the proportion of these successful rates
  which are greater than the one obtained from the
  ionopheric precursor.
• Simulation Study B
• Generalize 10,000 sequences of the occurrence of
  foF2 anomaly.
• Calculate successful rate for each sequence.
• Find the proportion of these successful rates
  which are greater than the one obtained from the
  ionospheric precursor.

14
Table 5. The proportion of 10,000 sequences of
predictions with successful rate higher than the
one obtained from the prediction based on foF2
precursor for M ? 5 earthquakes within I days.
15
Simulation results The results of the
statistical tests based on two simulations show
that the foF2 precursor currently observed is
significantly better than the prediction based on
earthquake experience. Moreover, the foF2
precursor currently observed is significantly
(not by chance) related to the earthquakes
occurred during 1994-1999. In fact, the
successful rate about 50 further indicates a
need of an improvement over the current
identification process of the foF2 precursor.