English


Electromagnetic Phenomena Associated With Earthquakes

observationsystem-w800px-2
◆Click to view larger

In order to produce our earthquake forecasts, we rely on data from:
– F-net, the broadband seismograph network maintained by the National Research Institute for Earth Science and Disaster Resilience.
– VLF/ LF, ULF and GPS transmitter signals.

There are several earthquake precursory signals. EAL/IS chose F-net, VLF/LF, ULF and GPS data to research and analyze for predicting short-term earthquakes. We have been successfully forecasting earthquakes from 2010. See our outstanding record of on target forecasts.

By investing in EAL/IS’s observation system, many countries facing threat from earthquakes can implement disaster risk reduction. Please contact us for details regarding our monitoring system.
To those residing in earthquake prone countries, we hope to reduce your anxiety due to the lethal potential of seismic disasters by offering our sincere desire that in the very near future earthquake forecasting will be accepted and regarded as an essential service of our lives.


VLF / LF

Actual EQ Event vs Forecasted EQ Event

From 2011 To Present Published on
EAL website, Google Play, iPhone

imgforecasted

F-net

The data observed by the “F-net” is already utilized for studying mechanisms of earthquake occurrence and expected to give a large contribution to understanding of earthquake phenomena.

fnet

med_dataWith the Molchan’s Error Diagram (MED) approach, we evaluate our forecast results based on VLF subionospheric anomalies in this report.
The alarm rate is defined by as follow: the number of cell over Japan area is 2,811 (inside red line in attached PDF document) and we analyzed one year data (2013). So total analyzed area *days = 2811*365 = 1,026,015. The alarmed area *day is computed using individual alarms issued in 2013 and it was 81,516. Therefore,
the alarm rate (AR) is 81,516/1,026,015 = 0.0794.
The number of earthquakes with M>=5 in the target area (see PDF) is 107. Here, we define the success of forecasting as follows:
(1)Time: difference between -2 and +5 days
(2)Location: Within 50km (0.5 deg.) in latitude and longitude
(3)Magnitude: Forecasted M±0.5.
Then, the number of successfully forecasted earthquakes in time, location, and magnitude is 38. Thus,
the detection rate is about 38/107 = 0.355. The gain is 355/79.4 = 4.47.
※(M±0.3. 22/38)
We can conclude that the result is rather good.

alarm-rate2

Evaluation of efficiency of short-term forecast using Molchan’s error diagram

a: the number of successful predictions of EQs
b: the number of false alarms
c: the number of successful predictions
of non-occurrence
d: the number of missed EQ

The proportion of predicted EQs,
V=a / (a+d)
The proportion of alarmed cells,
T=(a+b) / (a+b+c+d)