Is There a Season for Earthquakes?
An Analysis of a 92 Year Distribution
Magnitude 7+ Worldwide Earthquakes
The hypothesis of this paper is to suggest that the way data are categorized and presented can make a significant difference in the conclusions which can be drawn from those data.
Suddenly, without warning and without mercy, chaos, destruction and death. A massive telluric upheaval that we know as an earthquake, has struck. The lucky ones die immediately. Trapped survivors linger in pain, darkness and despair for days until they slowly expire from their injuries and thirst. Rescuers frantically search the rubble as hope fades and the stench of death fills the air.
Of the many so called “acts of god” endured by man, the earthquake is by far the cruelest one of them all. If Gods’ Will “be done in earth as it is in heaven” can we not search the heavens and know what is to come? Can earthquakes be predicted? Is there a season for earthquakes?
A statement by the United States Geological Survey (USGS):
“There is no such thing as “earthquake weather”. Statistically, there is approximately an equal distribution of earthquakes in cold weather, hot weather, rainy weather, etc. Very large low-pressure changes associated with major storm systems (typhoons, hurricanes, etc.) are known to trigger episodes of fault slip (slow earthquakes) in the Earth’s crust and may also play a role in triggering some damaging earthquakes. However, the numbers are small and are not statistically significant.” (Contributed by Dr. Malcolm Johnston) (1)
It was the tragic January 2010 Haitian earthquake that challenged me to investigate this phenomenon. For years I had been observing tropical cyclones and other severe weather and it seemed to me that from time to time earthquakes and hurricanes appeared in the news headlines simultaneously. In the course of my observations, I believe I have learned a few things since then, one of which I would like to share with you now.
The hypothesis of this paper is that the traditional way scientific data is organized and recorded, by calendar month, may obfuscate natural patterns existing in those data.
What I have done is to download from the USGS Earthquake Catalogue the 1,205 magnitude 7.0+ earthquakes that have occurred over the 92-year time span of 1925 thru 2016. I then organized this data according to three treatments: First, I summed the data in the traditional manner of categorizing by Calendar Month, as is the usual method of scientific institutions. Next, I arranged the same data according to the Signs of the Zodiac, from approximately the 21(st) of one month to approximately the 21(st)of the next month. Lastly, as a control group, I split the Zodiac by categorizing the data from approximately the 5(th) of one month to approximately the 5(th) of the next. I then applied the basic statistical procedures of graphically illustrating the obtained distributions and then analyzing the same by means of the Chi-square Goodness of Fit (2) and Binomial Distribution (3) tests. Specific details of the variables and tests are presented in an appendix.
A word about Significance Levels (p)
The Significance Level (p) that a researcher chooses as the Critical Value upon which to measure the success/failure (i.e. the presence of ‘evidence’) of a statistical test is the probability of the obtained sample results, assuming the ‘null hypothesis” is true. It is the level of chance he/she is willing to take that he or she is correctly rejecting the ‘null hypothesis’, and that his or her research hypothesis may be correct. Also, the Significance Level is what the audience and other researchers consider when reviewing the test and may thereby decide for themselves if the information is credible, valid or useful.
Significance Levels are expressed as a probability (p), in decimal form ranging from 0.0 (no chance) to 1.0 (certainly chance), which is the possibility that an event occurred by normal sampling variation or mere ‘random chance’.
Significance Levels are not set in stone but are situational, the most common being:
- (p) = .10 which is 10% or 1 chance in 10 (moderate significance)
- (p) = .05 which is 5% or 1 chance in 20 (most commonly accepted level)
- (p) = .01 which is 1% or 1 chance in 100 (very high significance level)
Significance Levels for quantitative tests in the physical sciences are often quite strict: (p) .05 or (p) .01. Levels for qualitative tests in the social sciences (like astrology) are more relaxed and a (p) of .10 is considered moderately significant. (4)
Since our experiments will be comparisons of frequencies distributed by qualitative/categorical variables, I elect to choose a Critical Value of less than or equal to (p) = .10, a moderate Significance Level, as evidence of astrological influence upon this sample of earthquake activity.
Treatment One: By Calendar Month
Chart One graphically displays the distribution by Calendar Month of 1,205 M7.0+ earthquakes (EQ’s) occurring during the years 1925 thru 2016. The Chi-square Goodness of Fit test returns a probability of (p) = 0.30, indicating that a similar pattern could probably occur 30 times out of 100 random samples, or, another way, a 30% probability that this pattern is the result of random chance, much beyond the commonly accepted Significance Levels. Notice that the black distribution line roughly follows the gray expected averages line, with just two sharp departures for September and November.
The adjacent Table One is a summary of (one-tailed) probabilities (5) obtained by means of a Binomial Distribution test. As indicated by the graph, the fewer than expected EQ’s in September has a Significance Level of (p) =.04, and the higher than expected EQ’s for November has a Significance Level of (p) =.01, both of which are each considered highly significant.
However, one must keep in mind that with an overall Chi-square of a 30% likelihood that this earthquake distribution pattern occurred by random sampling variation, the results from a second series of 12 Binomial tests performed on that same 1,205 EQ data set does increase the chance of obtaining a (p) less than .05 result for at least one of the 12 categories (Signs) tested. This is the so called ‘Bonferroni’ effect.
To use a metaphor, that’s like taking 12 shots at a large target that is up close, increasing your chances of a ‘[email protected]’.
When viewed in this traditional way, the ‘null hypothesis’ position maintained by the USGS (“There is no such thing as earthquake weather”) seems to be supported.
Treatment Two: By Signs of the Zodiac
A very different and distinct pattern emerges when the same 1,205 EQ’s are categorized by the Signs of the Zodiac. This distribution displays both symmetry and cohesiveness.
The Chi-square test returns a probability of (p) = .10, a moderate but still significant result, suggesting that a similar pattern would emerge only 10 times out of 100 random samples, or, only a 10% likelihood of this result occurring by random chance, significantly different than the Monthly categories. Also notice the black distribution line only follows the grey expected averages line for 5 of the 12 Signs with sharp departures of 5 Signs along the rest of the wheel, distinctly different than the Monthly distribution.
Furthermore, when each of the 12 categories (Signs) are tested against the Binomial distribution, the Zodiac Signs produce five significant departures from the expected average, versus only two from the Monthly distribution. The Binomial summary of Table Two contains five results ranging from moderate to highly significant and given the low overall Chi-square of (p).10 for this earthquake pattern of distribution, these results obtain a certain credibility that the Monthly distribution does not.
Revisiting the previous metaphor, the ‘Zodiac target’ is smaller, farther away, and we still get five [email protected] out of 12 shots.
In my opinion, when this sample is categorized by Zodiac Sign, the result does credibly challenge the USGS position that “the numbers are small and are not statistically significant”.
Treatment Three: By Split-Zodiac
Lastly, we have Chart Three that graphically demonstrates what happens when we neutralize the assumed influence of the Zodiac in a control group.
Keep in mind that each of the Monthly categories we have observed contain 2/3’s of one Sign and 1/3 of the next, potentially skewing a result in favor of the Sign with the larger proportion of days.
If we dilute the influence of the Zodiac more than the 2/3 – 1/3 proportion of the Monthly categories, we get a very different result. When a control group of 12 categories is created by ‘splitting’ the influence of the Zodiac in half and spreading the distribution of earthquakes evenly into those neutralized categories, the Chi-square test returns a very high probability value of (p)=.56, meaning 56 out of 100 random samples would produce a pattern like this one, or, that there is a 56% likelihood that this distribution is the result of random chance.
Finally, with such a high probability of random chance for this distribution (.56), the two significant Binomial results that were obtained (Table Three) are virtually meaningless, completely the result of random sampling variation.
Again, with the metaphor: a huge target well within arms’ length reach.
In summary, we have three distributions to consider, in the order of their Significance Levels:
- (p) = .10 Signs of the Zodiac (with five significant categories)
- (p) = .30 Calendar Months (with two significant categories)
- (p) = .56 Split Zodiac Control (with two significant categories)
Clearly, the Signs of the Zodiac distribution is the least likely to have occurred by chance and with five of the 12 categories (Signs) showing significant (one-tailed) Binomial probabilities (i.e. five [email protected]), the evidence in this study supports the hypothesis that categorizing by Signs of the Zodiac presents the clearest picture of this natural phenomena upon which to conduct our investigations.
However, I believe the most significant find in this study is the ‘effect’ on each distribution as the actual earthquakes are categorized and re-categorized, first as an all Zodiac distribution, then as a Calendar Month (2/3-1/3 Zodiac) distribution, and lastly by splitting the Zodiac in half in order to minimize its influence as much as is possible while maintaining the integrity of the earthquake data occurring in their natural sequence of time.
The probability then jumps from a low 10% for the earthquake pattern by Zodiac Signs to a high of 56% for the non-Zodiac Signs control group (about the same probability as a coin flip), suggesting evidence for the influence of the Signs of the Zodiac on earthquake activity.
So, does this mean that we can forecast earthquakes with Sun Sign Astrology? No, of course not. What it does mean is that we can know when and where to look in our research of cause and effect. For example, in the 92 years of our sample the faster planets have been around the wheel many times. Pluto and Neptune have not, however, and Uranus only once. Pluto has spent the most time in those Signs showing the greatest significance, so perhaps these slow movers should be the basis of our research.
Oh, and there is one other thing: There does seem to be a season for earthquakes. In this study, the number of significant earthquakes during Cancer and Virgo were 83 but the number jumps 50% to 123 quakes during Leo so the odds of experiencing a significant earthquake during Leo go up considerably.
So, if you have business in a seismologically active part of the world, you might want to skip the trip during Gemini and Leo and go in the months of Cancer or Virgo instead!
That’s all I have for now. Thank you for your kind attention.
Epilogue: An Earthquake Prediction
I do believe that earthquakes, or rather, the timing of them, can be predicted with some specificity, as I have had some success with this in the past. As of the date of submission of this paper to NCGR Publications, June 27, 2018, there have been no M7+ earthquakes since the last eclipse cycle of Jan/Feb 2018. Just on the horizon, I see a number of extreme astronomical events due to occur this summer: A series of eclipses that coincide with the perigee/apogee cycles, and Mars, out of bounds in an extreme southern declination, is aligned with the Lunar Nodes and will be semi-sextile/square to the forming Saturn/Uranus trine in Taurus and Capricorn. So, on the basis of that information, I will take an educated guess that at least one, and probably another, M7+ earthquake will occur on any of the following dates, +/- a day or two:
- July 13, 2018
- July 19, 2018
- July 27, 2018
- August 1, 2018
- August 10, 2018
- August 20, 2018
Binomial Distribution Calculator (6) ????
The probability of exactly 1 (K) out of 6 (n) is p = .179
So, it looks like I have about an 18% chance of a successful prediction. Time will tell.
August 26, 2018 Addendum to “Is There a Season for Earthquakes?”
The Results of the “Epilogue Earthquake Prediction”
In a June 2018 article (submitted to NCGR on June 27, 2018) on the distribution of earthquakes, I included an “educated guess” that at least one M7+ earthquake would strike during an upcoming July/August Eclipse Cycle and posted a series of dates that I ascertained earthquake activity was most likely to occur.
From the “Epilogue: An Earthquake Prediction”
“…. I see a number of extreme astronomical events due to occur this summer: A series of eclipses that coincide with the perigee/apogee cycles, and Mars, out of bounds in an extreme southern declination, is aligned with the Lunar Nodes and will be semi-sextile/square to the forming Saturn/Uranus trine in Taurus and Capricorn. So, on the basis of that information, I will take an educated guess that at least one, and probably another, M7+ earthquake will occur on any of the following dates,
7/13; 7/19; 7/27; 8/1; 8/10; 8/20; +/- a day or two”.
(Summer Eclipse Cycle 2018)
This forecast was very successful: First one large earthquake (a M8.2 EQ) and then another (a M7.3 EQ) occurred each within one day of Target Date 8/20.
The average frequency for M8+ EQ’s is .002 per day and that of M7+ is .0359. The probability of these independent events jointly occurring within a certain number of trials is obtained by multiplying their respective frequencies:
p(M8+EQ, M7+EQ) = .002 x .0359 = .0000718
A total number of trials is determined in this way:
6 Target Dates +/-2 days = 6 Targets Zones of 5 days each = 30 ‘Independent Trials’.
The Binomial probability of this joint occurrence is:
(p)= .002, or in other words, about 2 chances in 1,000. (from MS Excel)
Also worthy of note is that these large quakes occurred in the Zodiac Sign of Leo which has been emphasized in the main article.
Finally, there was an exceptional amount of smaller M6+EQ activity that also occurred on the forecast Target Dates. There were multiple EQ’s (Hits) on 5 of the 6 Target Dates for a Hit Rate of 83%, and half of those quakes seemed to have been associated with the Eclipses, the other half with exceptionally strong planetary influences in late August. I hope to present in a future article.
Conclusion: While this is just a single test, the evidence strongly supports the presence of potent astrological influences on earthquake activity.
In order to avoid a boring redundancy, I will include just the specific details of the variables and statistical tests for the Earthquake Distribution by Signs of the Zodiac.
92yrs x 365.25= 33,603 days in Population Space
1205 eq ÷ 33,603 days = (p) 0.03585 M 7+ EQ per day
M8+ (1901-2016) (for addendum)
116yrs x 365.25 = 42,369 days
93 eq ÷ 42,369 days = (p) 0.00219 M8+ EQ per day
Variables for Zodiac / Split Zodiac Control
365.25 ÷12 signs = 30.4 days per sign
92yrs x 30.4=2800.25 days per sign
(p) .03585 x 2800.25 days per sign = 100.4 expected
Variables for Calendar Months
30 Day Months
92yrs x 30 =2760 days per month
(p) .03585 x 2760 days per month = 99.0 expected
31 Day Months
92yrs x 31 = 2852 days per month
(p) .03585 x 2852 days per month = 102.3 expected
92yrs x 28.25 = 2599 days per month
(p) .03585 x 2599 days per month = 93.3 expected
Chi-square Goodness of Fit
|zodiac||observed||expected||difference diff²||diff² ÷ expected|
|deg of freedom||11|
Degrees of Freedom (df) is a mathematical concept and a parameter that is applied to the formulas of several different statistical tests in order to determine (p), the critical value used to decide whether or not to reject a null hypothesis. The most common tests that require (df) are: The Chi-square Goodness of Fit and Chi-square Test for Independence distribution, the Student’s ‘t’ distribution and ANOVA’s ‘F’ distribution.
The (df) concept is as difficult to explain as it is to understand, but the bottom line is that (df) is an adjustment, a ‘fine tuning’, of the sample size of a distribution to compensate for the tendency of a statistical test to under-estimate that distribution’s critical value, (p). Note: (n) stands for sample size, (k) stands for categories.
The number of Degrees of Freedom are determined differently for each of the applicable tests employed. For a quantitative distribution, the usual formula for (df) is the sample size minus one: df = (n-1). For qualitative distributions, it is the number of categories minus one: df = (k-1)
In the case of a distribution by 12 astrological signs, each sign is a separate category, and the (df) for the Chi-square Goodness of fit Test is determined in this way: (df) = (k-1) = (12-1) = 11.
Included below is a Chi-square Critical Values Table from a typical statistic’s textbook. (7)
Highlighted are the parameters of the Chi-square test of the Earthquake/Zodiac distribution we have examined:
(df) = 11 Chi-square value = 17.12 (p) = .10.
Chi-square Critical Values Table w/ Degrees of Freedom
Binomial Distribution (one-tailed) Probabilities (8)
|zodiac||# days||# earthquakes||frequency||(p)|
References / Resources
(2) Microsoft Excel Spreadsheet Chi-sqr Test
e.g. The (one-tailed) probability of exactly, or greater than, 123(k) out of 2800(n) is p = .012783
(4) https://www.socscistatistics.com/tests/Default.aspx As an example, most of the Tests on this site have Critical Values options of: (p).01; (p).05; and (p).10
(7) Voelker, Orton and Adams, Cliffs Quick Review Statistics, Wiley Publications, New York, 2001