Hi Kat (and everyone),
as it happens I was watching the live data feed from EMSC when the quake occurred. It was originally published as a 5.1 then got marginally downgraded
to a 5.0. They also located it more precisely and tightened up its time of occurrence. Minor adjustments for time, magnitude and location are quite
normal, as you know. So is a reassessment of the quake's depth. Originally it was published with a depth of 10km -- one of the typical default depths
that are in use these days. Since then it has been adjusted to 52km, which is reasonably deep and a lot less worrying than a shallow quake.
While the USGS has it listed as a 4.9 I've noticed that this organization seems to have a propensity to "downgrade" to a greater degree than
closer, regional agencies do. However where the regional centers have good quality data I think we can go with what they say. After all, the EMSC is a
highly-respected organization with very rigorous standards, and they are still showing it as a 5.0 on their database.
This quake in the Dodecanese Islands was the strongest there since they had a mag. 6.4 on July 15, 2008. According to EMSC data there have only been
six quakes in the Dodecanese Islands region with a mag of 5.0 or greater since January of 2005 -- including yesterday's. So they average about 1 per
year. On that basis, this quake is "significant" and one that seismologists would likely have noted.
For readers who might be interested I have set up a "search" page on the EMSC's website, which lists all quakes of mag. 5.0 or greater from Jan 17,
2005 to yesterday (Jan 17 2010), within a defined region from 32 to 42 deg North, and 19 to 29 deg East. This includes all of Greece, but I should
point out that because Greece is irregularly shaped and has several neighbors, some of the quakes listed are for places like Albania, Macedonia, parts
of western Turkey, and regions of sea that do not belong to Greece.
Here is that
There are 68 events listed. (Notice there are two pages.) Of these, several are clearly not in Greek territory. We can say, though, that there were
about 60 mag 5-plus quakes in or very close to Greek territory during the past 5 years. Twelve per year. I posted my prediction on Friday Jan 15. This
mag 5.0 occurred on Sunday Jan 17. So, if we allow for an average of one mag 5.0 (or greater) quake per month anywhere within Greek territory, the
chances of such an event occurring just 2 days after my prediction are around 1 in 15. (6.67%) If the quake did not happen until the last day of my
5-day time window then the odds go down to around 1 in 6. (16.67%)
In simple terms, this means that if I or anyone here posts a new prediction for "anywhere in Greece" every five days and we keep doing so until one
occurs, then after around six posts we should get a mag 5-plus result purely by chance. We can get the result after just one post, but the odds of it
happening are less. But we don't use that "method" of repeat posting, though some on other sites do. This was my first post about this event.
A note about "proof": from a scientific standpoint, it doesn't actually "prove" that I predicted anything, and it won't even if a mag 6.0 or
greater quake occurs within the predicted region and time frame. (*See below for analysis of the chances of such an event occurring.) None of our
efforts do that and we do not want people to think that we take our results as some kind of scientific "proof" that our methods work.
Even if we were to successfully predict 99 out of 100 larger quakes with well-specified regions, magnitudes and time frames, that still does not
"prove" that our methods work. It just creates a heck of a good argument for them!
Scientists are not always interested in getting "proof" anyway, because they accept that in some cases, proof is either unobtainable or not
essential. Instead, many things they study are analyzed on the basis of probabilities. The lower the probability of a particular result, the more
interesting it is when such (unlikely) results are consistently obtained. If you read research papers in any of the recognized scientific journals,
they very often show results against the probability of them occurring. In other words, they are saying, "Our results don't prove anything, but the
odds of getting them by chance are [x]." The lower the probability the more likely that their peers will take interest.
I'll put it another way: if we can get reasonably consistent results for well-defined predictions that show a probability far beyond mere chance,
then those in the scientific community are far more likely to take note of what we are doing and perhaps investigate further. The simple fact is that
this is how things work and to be fair, it is how things have to work.
Here's why: there are websites where people make predictions which frankly barely qualify to be described as such. For example, I know of a site
where the owner/founder posts "predictions" for various active regions, with a week or longer time window, and with large ranges of magnitude, like
eg 3.5 to 6.5 (or sometimes to 6.0). That is a range with a relative energy-release factor of 1000. (A 6.5 is 1000 times more powerful than a 3.5.)
So, when a 3.5 or more happens some time during the next week (or whatever) in that region, he claims it as a 100% "hit", even though quite often
the odds of it happening were close to 100% anyway
. If it comes in a day late he rates it as a 90% "hit", two days late is 80% and so
I'm not sure how he derived his percentages... They don't seem to be based on the actual statistical probabilities of the events. But whatever...
The fact that he was 3 magnitudes under his stated upper-limit 6.5 doesn't seem to matter.
It matters to me: I get concerned when I'm just ONE magnitude out, because one order of magnitude is a factor of 10 in energy released. As for three
orders of magnitude difference... Well, I'll put it this way: if there is a magnitude 3.5, in most places it means the crockery will rattle. That's
about it. If there's a magnitude 6.5 -- 1000 times more powerful than your crockery-shaker -- it can collapse buildings and lead to piles of dead
bodies in the streets.
I see a big difference between predicting the first versus the second scenario. A very
big difference. And here's the rub: if you predict
within such a big mag range, sooner or later you'll get a 6.5 quake. Then you can say that you "predicted" it! You just don't mention the hundreds
of times you predicted a 6.5 and it didn't happen
and nothing even close to it happened
Predicting within such a big magnitude range ain't accuracy, folks. Not when you do it as an established routine. It has almost no value at all from
any seriously analytical point of view.
In regard to accuracy, I must state that I was expecting a magnitude 6 or bigger
. On that basis it can be reasonably argued that a mag 5.0
doesn't cut the ice and I accept that. It's a full magnitude less, statistically much more likely than a mag 6 and also much less potentially
damaging. On the other hand it was
a statistically significant event, especially as it occurred within the Dodecanese Islands group, of which
Rhodes (which I mentioned as my primary location) is the largest island. So, while it's not bang on target and isn't a 100% "hit", it is at least
interesting and could be useful for future analysis.
Regarding mag 6-plus events, that same search page I gave above can be used to check how many there have been in that given region. (Just change the
"Magnitude min" number from 5 to 6.) In total, there have been 8 such quakes since 17 Jan, 2005, all of them in Greece. There was just one in 2009,
six in 2008 (busy year!), none in 2007, one in 2006, and none in 2005. This shows why it's hard to use simple averaging, but there is not much else
we can do to estimate the odds of an event.
Anyway, if I take data only from 2006 onwards and exclude 2005, we wind up with a much easier-to-handle average of two per year. (It also demonstrates
how statistics are subject to variance in such analyses.) This gives us an average of one mag 6-plus quake every 180-odd days, for anywhere in Greece.
So, the chances of a mag 6-plus quake in Greece in any specified 5-day period are around 1 in 36, which is 2.78%.
Please note that if there is a mag 6 or bigger in Italy, Macedonia, Albania, western Turkey, a northern African country or any other near neighbor to
Greece by this Wednesday, that will not
be taken by me as a 100%
"hit". A "hit" maybe
, and useful to note, but not right on
In Greece, by this Wednesday at the latest, mag 6.0 minimum. That's
"hit". Anything else isn't -- and as the odds of it
happening as I've specified are less than 3%, there's a better than 97% chance I'll be wrong.
I know that and I knew it when I posted. I'd rather
be wrong, because I have predicted an event that -- if it happens -- could
the lives of many people. But if we don't post, then we'll never have any way to show if we're onto something.
Finally I need to say that the 5-day time window I stated for that prediction ends Wednesday. Hopefully there won't be another, larger quake by then,
but all the same, the window is
[edit on 18/1/10 by JustMike]