Tag error: <txp:article allowoverride="0" form="lazydays_default" limit="3" listform="lazydays_default" pgonly="0" searchall="0" searchsticky="0" sort="Posted desc" status="4" /> ->  Textpattern Notice: Form not found: comments_display while_parsing_page_form: lazydays_default, None
Blog: KHPtech: Worldwide Earthquake Statistics

Worldwide Earthquake Statistics

Ok, time for a little Science™.

Recently I was intrigued by some reports on the Internet showing drastically increased earthquake activity in recent years. The data claims to be sourced from the USGS website. So naturally, I decided to download the data and analyze it for myself.

After all that work, I decided I should post something here for anyone else interested.

First off, full disclosure of my sources:

My data comes from the USGS, from two datasets. The first dataset is called the “Centennial Earthquake Catalog”, and it can be found here. It does not include any earthquakes less than M 5.5, and extends from 1900 to April 2002. For reasons mentioned below, I only looked at data from 1973 onward.

The second dataset is called the “USGS/NEIC (PDE)”, and a query tool can be found here. This has a much larger set of entries for the time period we are considering. The NEIC dataset begins at 1973, which is why I only use the Centennial dataset from 1973 onward.

A word about the data

The Centennial dataset is by far the smaller of the two datasets for the time period I am considering. During any particular time period, it appears to have only a small fraction of the entries that the NEIC dataset would have during the same time period, even for M 6.0 and greater. It also terminates in April 2002. I believe this dataset represents earthquake statistics gathered by the USGS from published sources. You can see that it often correlates many different reports on the same quake, if you look at the data.

The NEIC dataset is much denser, and as far as I can tell it is made of up data that is constantly being fed from seismographs worldwide. Thus, I would expect this dataset to have a definite increase in density over time, as earthquake reporting improves and more seismographs are installed. Someone please correct me if I’m wrong.

I downloaded the entire Centennial dataset, and queried the NEIC dataset for the years 1973-2010, M 3.0 or greater. All data was converted to a common csv format and imported into a MySQL database. There are 8333 entries from the Centennial dataset, and 448624 entries from the NEIC dataset.

Frequency Graphs

I produced a number of different graphs to help me analyze the data. First, I wanted to see to total number of earthquakes per year.

USGS NEIC Earthquakes M ≥ 3 Per Year

This graph shows the number of earthquakes per year, M 3.0 or greater, categorized by magnitude range, from the NEIC dataset.

Here is the data from the Centennial dataset. Notice that the data only shows quakes M 5.5 or greater:

USGS Centennial Earthquakes M ≥ 5.5 Per Year

I made another graph of the NEIC data showing only quakes M 5.0 or greater for comparison:

USGS NEIC Earthquakes M ≥ 5 Per Year

Force Graphs

After looking at this data for a while, I started to think “lots of small earthquakes may not be as significant as a few large ones”. This is because the Richter scale is a logarithmic scale. An increase of 1 unit on the Richter scale equals 10x the ground motion (displacement). And furthermore, increase of 1 unit on the Richter scale equals about 32x the destructive force (energy). See here.

So I made some new graphs showing total magnitude, total displacement, and total energy of all earthquakes in a given year.

Now, the “total destructive force” for all earthquakes in a single year is a very large, absolutely meaningless number. As is total ground motion. As is “total” magnitude on the Richter scale. It would be hard to relate these numbers to anything, let alone put them on the same graph. So here’s what I did:

I found the maximum for all 3 sets, and then graphed each set as a percentage of the maximum. This is why I am calling them “normalized” values on the graphs below. In this way, you can easily see the overall trends as well as the relationship between these 3 values. Just keep in mind that there is no “absolute” relationship between the lines, so if one goes above or below another that doesn’t mean anything. What is important is how they track each other following the trends.

Ok, enough of that. If you want the formulas for how I calculated the numbers, they are below the first graph.

First, the NEIC data:

USGS NEIC Earthquakes M ≥ 3 Normalized Total Forces

The blue line shows total magnitude (sum of magnitudes) for all quakes during a given year. (Probably a meaningless metric)

The green line shows total displacement (movement of the earth) for all quakes during a given year. This is calculated as ∑(10M) for each year.

The yellow line shows total energy (destructive power) for all quakes during a given year. This is calculated as ∑((101.5)M) for each year. This is probably the most important metric.

Now the Centennial data:

USGS Centennial Earthquakes M ≥ 5.5 Normalized Total Forces

Finally, the NEIC data restricted to M 5.0 and above, for better comparison with the Centennial data:

USGS NEIC Earthquakes M ≥ 5 Normalized Total Forces


As you can see, there is a lot a variation in the graphs, but there is also an overall constancy. I will refrain from any special interpretation here.

You can definitely see a slightly upward linear trend in the NEIC data, especially when we include quakes below M 5.0. This is probably due to the constant improvement in detection and reporting worldwide, which mostly results in a greater number of recorded small-magnitude quakes. This conclusion is also supported by the USGS’ own explanation.

It is much harder to see a trend in the Centennial dataset. It is a much more sparse dataset and covers a narrow range of magnitudes over a shorter time. Overall the linear trend here seems almost flat.

When we look at the force graphs, here again we can see a slight upward linear trend, especially in the NEIC dataset. The interesting thing to note here is that this trend is still visible even when we restrict ourselves to M 5.0 and higher. Why?

Probably somewhat due to more and better reporting, but if you compare the frequency and force graphs, you will notice that there have been a few more large earthquakes in the last 10 years or so, and because of the math, it makes a big difference.

The most significant real-world impact is probably represented by the yellow “energy” line, representing total destructive force. It peaked in 2004, almost entirely as a result of the Asian Tsunami quake of that year.

2008 had an unusually high number of small-magnitude earthquakes. This gives you peak magnitude and displacement totals in 2008, but yields only an average total destructive energy. 2008 is still the most unusual year in the data, IMHO.

The Centennial graphs show peaks in 1995 and 2000-2001. These look really big but keep in mind they are just normalized values. They correspond to the moderate bumps in 1995 and 2000-2001 on the NEIC graphs.

Also notice that the last datapoint on all the graphs is lower than it should be; the totals for 2002 (Centennial) and 2009 (NEIC) are incomplete. The Centennial data ended on April 1, so you could roughly estimate the correct value by multiplying by 4. I believe The NEIC data is current through December 1, so it’s about 92% there.

If anyone wants a copy of my dataset, shoot me an email.

Posted by Kevin H. Patterson - 2009-12-02 16:31.
Posted in .