Japan Earthquake: An Exploratory View
Thanks to the data provided by the USGS, we can take a look at all earthquakes since 1973, which cover almost the last 40 years of earthquake activity worldwide.
Let’s first take a look at the yearly development of the earthquake activity overall:
The apparent increase in the last 10 years is striking – though I don’t have any explanation for this change, which is most probably not even man-made. Interestingly the magnitude (see next figure) does not increase, though the chance of stronger earthquakes will grow with the overall number.
The distribution of magnitudes (which is used for the coloring) is even more striking, looking at the earthquake in Japan, March 11th which is now rated as a 9.0.
The whole dataset contains only one earthquake at a higher magnitude, i.e., the earthquake originating the terrible tsunami on the 2nd Christmas day in 2004 at a magnitude of 9.1.
Keep in mind that the Richter scale is logarithmic, i.e., stepping up one unit means a 10 times stronger earthquake. The strongest ever measured earthquake was in Chile in 1960 at 9.5.
If we look at the coordinates of the measurements in longitudes and latitudes, we see how much the activity is concentrated on the tectonic hotspots.
We roughly see the shapes of some continents, with one exception. Africa seems to be free of any activity; probably due to the fact that it sits happily on its own tectonic plate.
Looking at this data, we can only start to understand the devastation Japan is facing.
(The data can be loaded directly into Mondrian, which was used to create the graphs above.)
Hi!
The explanation for the increasing number of earthquakes is just two clicks away from the link you gave for the data:
http://earthquake.usgs.gov/learn/topics/increase_in_earthquakes.php
I think the practice to generate visualization after visualization without bothering to understand the data is highly questionable. You should at least say what the first visualization is good for:
a) It indicates that you have correctly imported the data.
b) It shows a feature of the data that needs to be understood before one can reasonably analyse the data.
c) It shows off possibilities of Mondrian for decoration (in this case with too strong contrasts (LOOK AT ME PLEASE!) and meaningless color variation (LOOK AT ME!!! LOOK AT ME!!!)).
These are legitimate reasons to produce this visualization. But the visualization itself does not help to understand the data. Therefore it is no wonder that you jump to the wrong conclusions:
“… though the chance of stronger earthquakes will grow with the overall number.”
No, it doesn’t. The increase in the overall number is due to the registration of more very weak earthquakes.
“Africa seems to be free of any activity; probably due to the fact that it sits happily on its own tectonic plate.”
No, the reason for the small number of earthquakes is probably the lack of sensors.
Best,
Martin K.
Hi Martin K.
thanks for your comment, but I am afraid you mainly missed the spot:
> The explanation for the increasing number of earthquakes is just two
> clicks away from the link you gave for the data:
> http://earthquake.usgs.gov/learn/topics/increase_in_earthquakes.php
>
Unfortunately it does not give the explanation. Let me show you why.
> a) It indicates that you have correctly imported the data.
>
Thanks for confirming, but I was quite sure about that anyway …
> b) It shows a feature of the data that needs to be understood before
> one can reasonably analyse the data.
>
Well, neither the above mentioned webpage, nor your reasoning brings us an understanding. On the webpage it says:
So the data we look at does obviously not show all the small ones, as the full dataset has “only” 34,506 entries for almost 40 years – how does that fit to the mentioned 20,000 incidents per year?
> c) It shows off possibilities of Mondrian for decoration (in this case with
> too strong contrasts (LOOK AT ME PLEASE!) and meaningless color
> variation (LOOK AT ME!!! LOOK AT ME!!!)).
>
Sorry for not explaining the use of color more precisely for you. The coloring reflects the magnitude of each record, ranging from yellow (low magnitude) to purple (high magnitude) across all plots. Having such a broad range of colors, the perfect background color that even respects light colors is black (i.e., no color) – you probably might want to adjust brightness and contrast of your screen.
> These are legitimate reasons to produce this visualization. But the
> visualization itself does not help to understand the data. Therefore
> it is no wonder that you jump to the wrong conclusions:
>
> “… though the chance of stronger earthquakes will grow with the overall
> number.”
>
Hmm, there is nothing to argue about this one. For any “normal” distribution, the probability of observing a given extreme value will rise with the sample size.
> No, it doesn’t. The increase in the overall number is due to the
> registration of more very weak earthquakes.
>
Well, let’s look at the data again, to prove that you are wrong with this one:
http://www.theusRus.de/Blog-files/EQ-Boxplot.png
http://www.theusRus.de/Blog-files/EQ-Zoom.png
The increase did start in the late 90s, so I zoom in on the last 15 years. The boxplot shows an almost constant median within this period; it is 4.4 in 1996 and 4.4 in 2011. If the increased number of registered earthquakes in the dataset would be due “to the registration of more very weak earthquakes.” as you claim, the median must have been gone down – but it didn’t.
> “Africa seems to be free of any activity; probably due to the fact that it
> sits happily on its own tectonic plate.”
>
> No, the reason for the small number of earthquakes is probably the
> lack of sensors.
>
Same as above; and given the enormous number of sensitive sensors worldwide (outside of Africa), they would easily pick up at least the stronger earthquakes in Africa, but they don’t seem to find some!
Let me summarize:
a) yes, we can measure smaller earthquakes today, but that does not result in more entries in the used dataset.
b) the use of colors is always a matter of subjective preferences, and not everything is obvious for everybody
c) we still don’t know where the increase comes from, but maybe (this is my preferred explanation) we are on such a geologically small and insignificant window of time that all we see is just noise, compared to the longterm activity.
d) it might be good to first think twice (or actually look at the data) when writing a comment – this is at least what I usually try to do.
Thanks again for spurring the discussion
Martin
Hi!
> So the data we look at does obviously not show all the small ones,
> as the full dataset has “only” 34,506 entries for almost 40 years –
> how does that fit to the mentioned 20,000 incidents per year?
Part of the answer is given one click away from the link you gave:
http://earthquake.usgs.gov/earthquakes/eqarchives/epic/database.php
Micro-earthquakes having magnitudes below 1.0 are not retained in the data base. Earthquakes with magnitudes less than 2.0 are found in the data base, but in general, the magnitude level of earthquakes in the data base range from 2.5-9.5.
> Sorry for not explaining the use of color more precisely for you.
> The coloring reflects the magnitude of each record, ranging from
> yellow (low magnitude) to purple (high magnitude) across all plots.
Thanks, I indeed missed the reason for the coloring.
> > “… though the chance of stronger earthquakes will grow with the overall
> > number.”
> >
> Hmm, there is nothing to argue about this one. For any “normal”
> distribution, the probability of observing a given extreme value
> will rise with the sample size.
Only if the sample is representative. If the increase of the sample is biased (in this case to earthquakes with smaller magnitudes) the number of earthquakes of higher magnitudes won’t increase. On this page: http://earthquake.usgs.gov/learn/topics/increase_in_earthquakes.php
the providers of the data state quite plainly:
Although it may seem that we are having more earthquakes, earthquakes of magnitude 7.0 or greater have remained fairly constant.
> > No, it doesn’t. The increase in the overall number is due to the
> > registration of more very weak earthquakes.
> >
> Well, let’s look at the data again, to prove that you are wrong with
> this one:
> http://www.theusRus.de/Blog-files/EQ-Boxplot.png
> http://www.theusRus.de/Blog-files/EQ-Zoom.png
> The increase did start in the late 90s, so I zoom in on the
> last 15 years. The boxplot shows an almost constant median within
> this period; it is 4.4 in 1996 and 4.4 in 2011. If the increased
> number of registered earthquakes in the dataset would be due
> “to the registration of more very weak earthquakes.” as you claim,
> the median must have been gone down – but it didn’t.
I find it fascinating that you trust your visualizations more than the explanations of the providers of the data. 🙂
Why doesn’t the median change? I assume partly because of inconsistencies in the databases (apparently some databases include fewer of the very small earthquakes in recent years: http://earthquake.usgs.gov/earthquakes/eqarchives/year/eqstats.php), partly because the number of small earthquakes (4.2 is rather small) is so large that the median doesn’t change much if you include many earthquakes of magnitudes between 3 and 5.
> > “Africa seems to be free of any activity; probably due to the fact that it
> > sits happily on its own tectonic plate.”
> >
> > No, the reason for the small number of earthquakes is probably the
> > lack of sensors.
>
> Same as above; and given the enormous number of sensitive sensors
> worldwide (outside of Africa), they would easily pick up at least the
> stronger earthquakes in Africa, but they don’t seem to find some!
Actually, there were quite a few earthquakes in Africa (and they are visible in your visualization). But you don’t get the high number of small earthquakes (green in your visualization) that you see almost exclusively in the US and Europe. Do you think there are geological reasons why the US and Europe have so many earthquakes of very small magnitude while Canada, Mexico, South America, Africa, Asia, and Australia are all lacking them?
> Let me summarize:
> a) yes, we can measure smaller earthquakes today, but that does
> not result in more entries in the used dataset.
Wow, strong statement. In particular because the providers of the data contradict it. (If you don’t trust them, why do you trust their data?)
> b) the use of colors is always a matter of subjective preferences,
And one might claim that the overuse of too strong contrast and to strong colors has a long tradition in computer visualization! 😉
> and not everything is obvious for everybody
Obivously. Recently, I learned that economists tend to put the independent variable on the y axis and the dependent variable on the x axis. And they think it’s obvious. (They really do.)
> c) we still don’t know where the increase comes from, but maybe (this is
> my preferred explanation) we are on such a geologically small and
> insignificant window of time that all we see is just noise, compared
> to the longterm activity.
Wow. I’d better not ask about your opinion about climate change.
> d) it might be good to first think twice (or actually look at the data)
> when writing a comment – this is at least what I usually try to do.
It might also be a good idea to look at what the providers of data have to say about their data before using it. 😉
> Thanks again for spurring the discussion
>
> Martin
You are welcome. 🙂
Martin K.
First of all, I guess you need to decide whom you mistrust most, the visualizations I put up, the data from the USGS, or the explanations of the USGS (at least the visualizations and the data are in line and don’t contradict).
> > > “… though the chance of stronger earthquakes will grow with the overall
> > > number.”
> > >
> > Hmm, there is nothing to argue about this one. For any “normal”
> > distribution, the probability of observing a given extreme value
> > will rise with the sample size.
>
> Only if the sample is representative. If the increase of the sample
> is biased (in this case to earthquakes with smaller magnitudes) the
> number of earthquakes of higher magnitudes won’t increase. On this
> page: http://earthquake.usgs.gov/learn/topics/increase_in_earthquakes.php
> the providers of the data state quite plainly:
>
> “Although it may seem that we are having more earthquakes, earthquakes of
> magnitude 7.0 or greater have remained fairly constant.”
>
That is certainly true for the fairly small period of time the data covers. Nonetheless, the two strongest earthquakes happened in the last decade! As we talk about rare events here, this does not make me rest in comfort.
> > Same as above; and given the enormous number of sensitive sensors
> > worldwide (outside of Africa), they would easily pick up at least the
> > stronger earthquakes in Africa, but they don’t seem to find some!
>
> Actually, there were quite a few earthquakes in Africa (and they
> are visible in your visualization). But you don’t get the high number
> of small earthquakes (green in your visualization) that you see
> almost exclusively in the US and Europe. Do you think there are
> geological reasons why the US and Europe have so many earthquakes
> of very small magnitude while Canada, Mexico, South America,
> Africa, Asia, and Australia are all lacking them?
>
Take a look at: http://en.wikipedia.org/wiki/File:Plates_tect2_en.svg
Geologically there is nothing that would result in classical (severe) earthquake activity, as all surrounding plates move away from Africa, and most borders are in the oceans around Africa.
> > c) we still don’t know where the increase comes from, but maybe (this is
> > my preferred explanation) we are on such a geologically small and
> > insignificant window of time that all we see is just noise, compared
> > to the longterm activity.
>
> Wow. I’d better not ask about your opinion about climate change.
>
Well that comparison is really bogus. Climate changes happen on much smaller time scales than geological changes, i.e. the chance that your grandfather can remember the last ice-age is much higher than him remembering living in Gondwana ;-). The greenhouse effect we see right now can easily be (mostly) attributed to humans burning fossil fuels and that is happening for a century now on a larger scale. Geological activity though is on a much larger scale, e.g., in its best times the Alps rose by 5mm per year.
> > d) it might be good to first think twice (or actually look at the data)
> > when writing a comment – this is at least what I usually try to do.
>
> It might also be a good idea to look at what the providers of data
> have to say about their data before using it.
>
Well, that is the real problem here. The USGS does not really explain why and how they decide on very small values to enter the database or not. What is even worse IMHO is what they state with:
That may be ok for data shared among geological scientists, but is very questionable when releasing data to the general public on a website. Anyway, nice to get the data so easily, even if we don’t fully understand the censoring mechanisms for small magnitude entries.