See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/280764638

Innovative Mathematical Model for Earthquake Prediction

Conference Paper · June 2014

CITATIONS READS

2 1,441

1 author:

Suganth Kannan
MathforUS LLC
3 PUBLICATIONS   9 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Suganth Kannan on 08 August 2015.

The user has requested enhancement of the downloaded file.

 Innovative Mathematical Model for Earthquake Prediction

Suganth KANNAN # 1
1. Student, American Heritage School, Plantation, Florida, USA
[email protected]

Abstract
An Innovative Mathematical Model (IMM) analysis was carried out on the past ten to twenty years
of earthquake data with Latitude, Longitude and Magnitude as Variables using Mathematical
principle of Poisson's distribution and Spatial Connections for a large data group for each
earthquake zone, there is an identifiable pattern within the random occurrences of the earthquakes
around each fault zone. Using fault lines and past earthquake data from United States Geological
Survey (USGS) National Earthquake Information Center (NEIC) the following six different zones
were identified for analysis: California, Central USA, North East, Hawaii, Turkey, and Japan.
Using the results of this research, Disaster management agencies like United Nations (UN) , U.S.
Agency for International Aid (USAID), Turkey’s Emergency Management General Directorate
(TEMGD), and Fire and Disaster Management Agency, Japan (FDMA) can allocate and position
their resources in the right location to assist people in evacuation, supply and save lives.

Keywords: earthquake, prediction, geohazard, disaster management, mathematical model, and

spatial connection

.Introduction
Earthquakes are major natural disasters that occur around the world and cause damage to
infrastructure and loss of human life. Several of the current methods of predictions are based on
land deformations, tectonic movements, seismic activity, differences in seismic wave velocities of
different world regions, geomagnetic and geo-electric phenomenons. This research attempts to find
an identifiable pattern within each fault zone based on spatial connection theory which is,
earthquakes occurring within a fault zone are related to one another.
According to seismologist Kossobokov, researching at the International Institute of
Earthquake Prediction Theory and Mathematical Geophysics (IIEPTMG), extreme events, such as
high-magnitude earthquakes, are clustered in the form of obscure, strange patterns. Seismologists
from the IIEPTMG have developed an efficient M8 algorithm with a 99.71% confidence level for
predicting earthquakes. The stochastic-based M8 algorithm designed by scientists from the institute
has accurately predicted the major Spitak Earthquake (1988) in Armenia among other significant
earthquakes. In that time period, the M8 algorithm accurately predicted the location, time, and
magnitude range of six of seven earthquakes. In the time period between 1992 to 2009, the
algorithm has predicted 10 out of 15 major earthquakes. According to Kossobokov, to drive the
confidence level of the algorithm below 95%, the algorithm would have to fail to predict
earthquakes 19 consecutive times. According to other researchers cited in the paper, earthquake
prediction is possible if the “seismic roulette” is in your favor and if you utilize statistics [1].
According to the scientist Shebalin researching at the IIEPTMG, stochastic earthquake
prediction models can be combined utilizing differential equations. This would result in a more
accurate prediction since there are more stochastic models that a seismologist can rely on while
making an earthquake prediction. Since there are more prediction models involved, the reliability
and validity of the seismological prediction is increased. While predicting, seismologists do not
have to choose between one stochastic model and the other; rather they can combine both models to
form a single, more reliable prediction. Suggested by seismologists at the institute, possible
procedures to combine models involve differential combinations. The formula to calculate the new
stochastic model's prediction accuracy is the old model's accuracy times the input model's accuracy.
By mixing and matching stochastic prediction models, scientists can identify a highly accurate
prediction model [2].
According to the researcher Peresan, the stochastic CN Algorithm has been utilized to
predict earthquakes in the Italian region. This research utilizes arcs and curves, geological patterns,
and past earthquake history to predict earthquakes. The researcher geographically and
mathematically analyzes each area utilizing mathematical functions. The researcher is working on
predicting earthquakes in North, Central, and South Italy regions. The researcher also takes
statistical steps to increase the confidence of the model. Peresan predicts earthquakes with his
algorithm such that the predictions will be as spacially accurate as possible. The algorithm utilizes
information from small and moderate earthquakes to predict larger earthquakes. According to the
paper, the CN algorithm has an accuracy of 96% since it has 67 accurate forecasts out of the 70
attempts. Peresan hopes to improve the stochastic algorithm in the future to increase its accuracy
[3].

.Innovative Mathematical Model

The following SIX different zones for analysis were identified using the fault lines and past
earthquake data as shown in figure 1 and 2 below. (Namely California, Central USA, North East,
Hawaii, Turkey, and Japan)

Figure 1 Earthquake Analysis Zones

After studying the past earthquake data for each of the six zones identified above, the following
information about data and locations were decided upon to conduct the IMM analysis. Depending
on the occurrences of earthquakes and their magnitudes, the rectangular area or radius for the data
collection for each zone was chosen as shown in the table Earthquake Meta-Data About Data below.
 EARTHQUAKE META-DATA ABOUT DATA
RICHTOR MAGNITUDE Number of
ZONE LATITUDE LONGITUDE RADIUS RANGE FROM YEAR TO YEAR Earthquakes
California 42, 36 -127, -118 RECTANGLE 5.1 to 9 1991 2001 19
California 42, 36 -127, -118 RECTANGLE 5.1 to 9 2002 2012 14
California 36, 30 -124, -115 RECTANGLE 5.1 to 9 1991 2001 16
California 36, 30 -124, -115 RECTANGLE 5.1 to 9 2002 2012 19
Central USA 39 -99 1000 km 4 to 9 2001 Nov 2011 31
Central USA 39 -99 1000 km 4 to 9 1991 2000 23
North East 39 -75 1000 km 4 to 9 2001 Nov 2011 13
North East 39 -75 1000 km 4 to 9 1991 2000 19
Hawaii 19 -155 150 km 4.5 to 9 2001 Nov 2011 15
Hawaii 19 -155 150 km 4.5 to 9 1991 2000 14
Turkey 38 35 500 km 5 to 9 2001 Nov 2011 19
Turkey 38 35 500 km 5 to 9 1996 2000 12
Japan 35 135 600 km 6 to 9 2001 Nov 2011 11
Japan 35 135 600 km 6 to 9 1991 2000 14
 Using the National Earthquake Information Center (NEIC) database, the data for the six
zones listed above were downloaded using two different ten-year time limits to a total of twelve
data sets in KML format. Then the KML data file for the first fault zone was imported on to Google
Earth program and the earthquakes were arranged to the chronological order.
The data was analysed using the Spatial Connection model, the theory that was based on the
logical assumption that every earthquake within a fault zone is related to the previous earthquake.
Using the California fault zone data points, Spatial Connection lines were drawn between the first
and second earthquake, then between the second and third earthquake, and so on and so forth. See
below for the completed Spatial Connection image for 1991-2001 California earthquake data set.
Using the California fault zone data points, Spatial Connection lines were drawn between
the first and second earthquake, then between the second and third earthquake, and so on and so
forth. See below for the completed Spatial Connection image for 2002-2012 California earthquake
data set.
 There is a relationship existing between the earthquake occurrences with respect to distance,
direction, and time. After spatial connection between earthquake’s epicenters were carried out a
relationship equation between angle of turn and time to predict a distance range for the next
earthquake was carried out as follows.

In the above figure consider the two lines between first, second and third location of earthquakes. If
the angle between the lines is 'theta', distance between first and second location is 'x1' and
between second and third location is 'x2', then Poisson's Range Identifier (Pri) is
Poisson's Range Identifier (Pri) = [(x1 * time lag 2)/ [(COS (theta) * x2 * time lag1)]
Pri 1 = [(128 * 24) / [ COS (2) * 305 * 155 ] = 0.07
Similarly other Poisson's Range Identifiers were worked out for rest of the earthquakes in the zone.
To arrive at a statistically adept group for finding the average (Pri) for the zone two of the highest
values are omitted and cumulative value and mean for rest of the values are found.
Pri (cu) = 89.35
Pri (mean) = 4.70
The Poisson's distribution or Poisson's law of small numbers is a discrete probability distribution
that expresses the probability of a given number of events occurring in a fixed interval of time
and/or space if these events occur with a known average rate and independently of the time since
the last event.
Applying Poisson's Distribution to the Pri Data,
Df = POISSION DIST [Pri; Pri(mean); Pri (cu)]
Df 1 = POISSION DIST [0.07 ; 4.70 ; 89.35 ] = 0.01
Similarly other Distance factors were worked out for rest of the earthquakes in the zone. To arrive at
a statistically adept group for finding the average (Df) for the zone two of the highest values are
omitted and cumulative value and mean for rest of the values are found.
Df(m) = (0.01 + 0.15 + 1 + 0.01 + 0.49 + 0.01 + 0.01 + 1 + 0.01 + 0.49 + 0.01 + 0.49 +
0.31 + 0.01 + 0.01 + 0.90 + 1 + 0.31 + 0.01 + 0.05 + 0.01 ) / 19 = 0.23
From the data table, the Poisson's Range Identifier is determined as 2.99.
As we know
 Poisson's Range Identifier (Pri) = [(x1 * time lag 2)/ [ (COS (theta) * x2 * time lag1)]
The angle theta from the prediction was 40 degrees.
2.99 = [ (64 * t2 ) / (COS (40) * x2 * 23)]
2.99 = [ (64 * t2 ) / .77 * x2 * 23)]
2.99 = [ (64 * t2 ) / 17.71* x2)]
2.99 = [ (3.61 * t2 / x2)]
Re writing the equation x2 = 1.21 * t2
So in the first 100 days,
x2 = 1.21 * 100 = 121 km the Earthquake Epicenter could occur from the last earthquake.
If that does not happen in the first 200 days,
x2 = 1.21 * 200 = 242 km the Earthquake Epicenter could occur from the last earthquake.
Steps shown above in the table California Region (2002-2012) earthquake Data for
Magnitude 5.1 to 9 were repeated for the remaining five fault zones utilizing the earthquake data
set. Please see Appendix for remaining figures and tables.

.Results

CALIFORNIA REGION 1 (2002-2012) EARTHQUAKE DATA FOR MAGNITUDE 5.1 TO 9

Poisson
Range Distance
Year Month Day Quake number Distance Time Lag (Days) Angle cos(Angle) Identifier Factor (Df)
2002 6 17 1-2 50 413 10 0.98 0.12 0.12
2003 8 15 2-3 417 398 2 1.00 0.66 0.12
2004 9 18 3-4 451 283 115 0.42 0.15 0.12
2005 6 15 4-5 49 2 25 0.91 70.43 1.00
2005 6 17 5-6 76 198 60 0.50 2.69 0.63
2006 3 25 6-7 96 336 180 1.00 0.23 0.12
2007 2 26 7-8 276 220 15 0.97 0.93 0.12
2007 10 31 8-9 253 182 100 0.17 9.03 1.00
2008 4 30 9-10 188 212 5 1.00 2.96 0.12
2008 11 28 10-11 123 409 30 0.87 4.03 0.93
2010 1 10 11-12 57 662 5 1.00 0.12 0.12
2012 2 13 12-13 97 130 85 0.09 2.82 0.63
2012 6 21 13-14 370 122
2012 10 21
Cumulative 23.74 3.01
Mean 2.16 0.27

3.1 Summary
This earthquake prediction model is validated using the past earthquake data. Example: The
22 year period was divided into two time periods, 1991 to 2001 and 2002 to 2012. To validate the
Spatial Connection theory, the Spatial Connection Model was built for the data from year 1991 to
2001. Within that model, the predictions were carried out for the 2002 and beyond. When the
Spatial connection model for data from year 2002 to 2012 is built, the predictions corresponded
with the actual 2002 earthquake occurrence in the Spatial Connection Model.
 EARTHQUAKE PREDICTION TABLE FOR SIX ZONES

Zones Possion Ration Factor angle cos of angle last distance last time prediction time prediction distance

California 1 2.99 5 1.00 370 122 300 305

California 1 2.99 5 1.00 370 122 500 509

California 2 2.99 175 1.00 72 182 300 40

California 2 2.99 175 1.00 72 182 500 66

Central 3.99 120 0.50 632 56 50 283

Central 3.99 120 0.50 632 56 100 566

Northeast 1.99 30 0.87 547 40 100 793

Northeast 1.99 30 0.87 547 40 200 1587

Hawaii 2.49 65 0.42 52 120 100 41

Hawaii 2.49 65 0.42 52 120 200 82

Turkey 1.99 15 0.97 269 280 100 50

Turkey 1.99 15 0.97 269 280 200 100

Japan 0.99 110 0.34 67 110 100 180

Japan 0.99 110 0.34 67 110 200 360

3.2 T-TEST Analysis

Statistical Data Analysis was carried out to find whether the data is statistically significant
between the different sets of earthquake data analysed. As shown below in the T-Test Values Graph
and T-Test Values Table, all of the T-Test values were well below the 2.20 limit. So by the reverse T-
Test concept or the null hypothesis, researcher proves that there is no difference between these data
groups, supporting the theory that earthquakes within a fault zone are related to one another.
T-Test Calculation Method Table

Result 1 Difference between mean of the groups

Result 2 Variance of the first group divided by number of subjects

Result 3 Variance of the second group divided by number of subjects

Result 4 Square root of Result 2 & Result 3 addition

T-TEST Value Result 1 divided by Result 4

 T-TEST
Group Result 1 Result 2 Result 3 Result 4 VALUE

Calfornia – Central 38.41 1276.71 20.18 36.01 1.07

Calfornia – Eastern 34.24 1276.71 106.14 37.19 0.92

Calfornia – Hawaii 36.31 1276.71 33.18 36.19 1.00

Calfornia – Turkey 29.11 1276.71 284.81 39.52 0.74

Calfornia – Japan 38.87 1276.71 47.92 36.40 1.07

Central - Eastern 4.17 14.61 106.14 10.99 0.38

Central - Hawaii 2.10 14.61 33.18 6.91 0.30

Central - Turkey 9.30 14.61 284.81 17.30 0.54

Central - Japan 0.46 14.61 47.92 7.91 0.06

Eastern - Hawaii 2.07 106.14 33.18 11.80 0.18

Eastern - Turkey 5.13 106.14 284.81 19.77 0.26

Eastern - Japan 4.63 106.14 47.92 12.41 0.37

Hawaii – Turkey 7.20 33.18 284.81 17.83 0.40

Hawaii – Japan 2.56 33.18 47.92 9.01 0.28

Turkey - Japan 9.76 284.81 47.92 18.24 0.54

T-Test Values Graph

1.20

1.00

0.80

0.60

0.40

0.20

0.00
an
an
y
n

ai
y
n

rke
ter

Jap
rke

p
t er

Ja
Tu

Ha
s
Tu
s

an
Ea

an

ey

i–
ai
Ea

an

Ea y
l

ai

-
tra

ap
p
l-
–

ur k
n-
w

rn
ur k

wa
p
w

Ja
l-
ia –
n

tr a
ia

Ja

Ha

-J
s te
Ha

te r
tr a
Ce

–T
or n

-T

Ha
n

l-
or n

ia –

s
n

Ce
l-
ia –

y
ia –

tr a

Ea
Ce
lf

rke
ai
n
tr a
Ca
lf

or n

ter
n
Ca

w
or n
or n

Tu
n

Ce

Ha
s
Ce
lf

Ea
Ca
lf
lf

Ca
Ca
.

.4 Discussion
Innovative Mathematical Model (IMM) analyses of the California fault zone data showed
the seismic forces travelling long distances along the San Andreas Fault. They released high energy
and formed a zigzag pattern while slowly moving north throughout the fault zone. Comparison
between the models of 1991 to 2001 and 2002 through 2012 showed that during active seismic
cycles the pattern of pressure releases were along zigzag formations with sharp turns.
In the Central part of USA, the IMM analyses showed that seismic forces travelled criss-
crossing large portions of landmass, but most earthquake force releases occurred along the edges of
the region. Seismic force travel patterns were much more predictable since the earthquake
occurrences created similar triangles.
In the North-east part of the USA, the IMM analyses showed that the speeds of travel for the
seismic forces were very high covering large distances from south to north. Within the north end of
the spectrum (near Ottawa region), the forces travelled slow and made sharp and obtuse turns
suggesting high-density material in the stratum. Down towards the south end of the spectrum (near
Virginia region), the forces travelled fast and made acute return angles suggesting low-density
material.
In the Hawaii fault zone IMM analyses, the seismic force releases occurred in concentrated
locations, while the travel of forces followed parallel paths. In several instances, the travel of
seismic forces between two earthquakes followed the same path suggesting high density material in
the stratum, with evident fault lines.
To verify that the Innovative Mathematical Model (IMM) exists between earthquakes
irrespective of the fault zone, two more zones outside of United States of America (USA) were
considered for analysis. In the European continent, within the Turkey fault zone, the seismic forces
travelled along parallel and zigzag patterns. The seismic force releases occurring more along the
east end compared to the west end of the zone and more along the south end compared to the north
end of the zone.
Within the very active Asian Continent, the Japan fault zone with numerous earthquake
occurrences, seismic forces travelled larger distances in short periods of time showing clear fault
line movements. Many seismic force travels occurred in a back and forth fashion similar to “pool
table ball movement” confirming that the repeated occurrences of the earthquake have formed
routine travel routes for the seismic forces with releases occurring when deviations occur due to
slight change of angle.
From the analyses conducted using the Innovative Mathematical Model (IMM), it became
evident that earthquake forces travel at different speed and pattern in different fault zones. Using the
pattern from each zone, the future predictions were carried out within the zones. As the seismic
forces travelled within the fault zones, it was clear that diversions and reactive routes occurred in a
predictable pattern.
This research provides an effective contribution to seismology on earthquake knowledge.
The research results show that there is an identifiable pattern in random earthquake occurrences.
This also provides researchers with knowledge of the seismic patterns in multiple fault planes in
different fault zones around the world.

. Conclusion
After catastrophic earthquakes, there have been major economic consequences. In recent
times, there has been an increase in investments in ports, highways, rail roads, bridges, buildings,
telephones, cable wires, ships, etc. Such valuable properties were greatly damaged. This has caused
financial losses to all scientists, investors, engineers, doctors, etc. working in the earthquake range.
Insurance companies have filed for bankruptcy because of the large demand of money from their
 clients' losses. Ports were extremely dangerous because they were built on loose soil and they
brought income to a particular region. Due to the destroyed highways, rail roads, and bridges, the
goods cannot be transported easily from one business to another. Potential future losses after
earthquakes are rising rapidly demanding we find a mechanism to predict earthquakes. [4]
From the above discussions, it is clear that there is an urgent need to predict future
earthquake occurrences to save precious lives and reduce economic consequences by using
Innovative Mathematical Model (IMM) in tandem with other prediction methods for any particular
location around the world.
5.1 Applications
The researched method of Innovative Mathematical Model (IMM) can be applied to predict
future earthquakes within a fault zone. With over 24 billion dollars in losses from earthquake
damage for the 1994 North Ridge California Earthquake, the potential practical implementation of
these principles could save several hundred millions of dollars and precious lives.
Using the results of this research, Disaster management agencies like United Nations (UN) ,
Federal Emergency Management Agency (FEMA), U.S. Agency for International Aid (USAID),
Turkey’s Emergency Management General Directorate (TEMGD), and Fire and Disaster
Management Agency, Japan (FDMA) can allocate and position their resources in the right location
to assist people in evacuation, supply and save lives.

. References
1. Kossobokov, V. G. (2009, July 7). Earthquake Prediction: 20 Years of Global Experiment
PowerPoint Slides]. http://124.17.4.5/English/Symposium/Slides%20and
%20posters/Keynote/ISESEP09-Kossobokov.pdf

2. Shebalin, P., Narteau, C., Zechar, J. D., & Holschneider, M. (n.d.). Combining earthquake
forecast models using differential probability gains. Geophysical Journal. Retrieved from
http://earth.usc.edu/~zechar/shebalinetal2012.pdf
3. Peresan, A., Costa, G., & Panza, G. F. (1998, February). Seismotectonic Model and CN
Earthquake Prediction in Italy. Published by United Nations Educational Scientific and Cultural
Organization and International Atomic Energy Agency,
http://streaming.ictp.trieste.it/preprints/P/98/019.pdf
4. Jones, B., & Cornell University. (n.d.). The Meeting. In B. Jones & Cornell University (Eds.),
Economic Consequences of Earthquakes: Preparing for the Unexpected. Preface (p. 275).
Retrieved from http://mceer.buffalo.edu/publications/workshop/97-SP01/preface.asp

View publication stats