An Analysis of Global Earthquake Data

Victor Shi

12/18/21

Introduction

It has been a decade since the Great Tohoku Earthquake and Tsunami of 2011, in which so many people's lives perished and their homes destroyed. People are still recovering from that disaster. Earthquakes have caused extensive amounts of damage, deaths, and injuries all over the world. Analyzing earthquake data will help find patterns in earthquake occurrences, allowing people to identify which areas are at risk. It may also be used to influence building design decisions so that buildings can be built to be earthquake-resistant, minimizing damage and fatalities in the event of a earthquake. Earthquakes can also provide valuable insights into the movement of the earth's tectonic plates.

An Explanation of Magnitude Scales

Most people have probably heard of the Richter scale, a scale used to measure earthquake magnitude. However, the Richter scale does not measure the magnitudes of very large earthquakes precisely and cannot be used to compare earthquakes around the world. Today, the preferred method of measuring earthquakes is the moment magnitude scale, Mw or Mww. This magnitude scale is applicable to a wide range of magnitudes and can be used globally. There are many other magnitude scales, each which measures something different. We're gonna say that the differences in magnitude scales in our data table aren't significant enough to do anything about it. Magnitude scales are logarithmic (base 10), meaning that the level of shaking increases by 10 times for each whole number you go up.

Magnitude scales can be confusing so you can read more about it here: https://www.usgs.gov/programs/earthquake-hazards/magnitude-types

Data Collection

I used data from the United States Geological Survey (USGS) at https://earthquake.usgs.gov/earthquakes/search/. Since search results were limited to 20,000 results, I decided to focus on earthquakes with larger magnitudes. I filtered out results by selecting events from across the world that occurred from 1900 onwards and had a minimum magnitude of 6.0. I selected the option to output results as a .csv file and then imported the data into a pandas DataFrame. This is what the first 5 rows of the dataset looks like.

Data Cleaning

Let's make sure that the data we're looking at is actually earthquakes and not something else such as an underground nuclear bomb test. We can also remove many of the columns in the data table since they won't be relevant to the analysis.

Data Manipulation and Visualization

In order to see which areas are at most risk of major earthquakes, I used a heatmap from Google Maps to display which areas had frequent earthquake events.

Heatmap of Earthquakes with Magnitudes of 6.0 or greater

Heatmap of Earthquakes with Magnitudes of 7.0 or greater

Image of tectonic plate boundaries from the National Park Service for comparison

https://www.nps.gov/subjects/geology/images/Fig-1-3-New-Tectonic-Map-x10.jpg

Observations

The most frequent area of large earthquakes is around the Ring of Fire, which is a belt consisting of several tectonic plate boundaries. The Ring of Fire stretches from the tip of South America to Alaska, and from Alaska to the Philippines and from the Philippines through New Zealand to the Antarctic Plate. From comparing the plate boundaries map and earthquake locations, it appears that the type of plate boundary has a relationship with the frequency of earthquakes. The Ring of Fire consists of mostly convergent boundaries while being the location of most of the world's earthquakes. And even within the Ring of Fire, areas that aren't convegent plate boundaries, such the coast of California, have less large earthquakes.

I also noticed that there are some earthquakes that don't occur right at a plate boundary, but somewhat close to one. Examining the heatmap of 7.0+ earthquakes, there are multiple earthquakes in Western China, north of the plate boundary at the Himalayan mountains. Another example is the earthquakes that occured inland of South America, east of the coast where the plate boundary is and east of the Andes mountains. In both of these cases, the plate boundaries are near major mountain ranges. This leads me to hypothesize that the large amounts of stress on the tectonic plates, indicated by the mountain ranges, is enough to cause earthquakes farther away.

More resources on the different types of plate boundaries:

How frequent are earthquakes?

Knowing how often earthquakes occur is important for warning against future earthquakes. To start, let's just plot the earthquake magnitudes against time to see if there's any patterns in the data. The frequency of earthquakes should be independent of time but let's see.

There are a lot more data points after 1950 as well as a much different distribution of points. This is most likely because that is when they started to record earthquakes more regularly. To make sure the data is consistent, let's remove all of the data points before 1950.

Now that the data is clean we can do some simple calculations to find the average number of 6.0+ earthquakes per day. Just divide the number of earthquakes by the number of days in the time period. Take the reciprocal of tht to get the average number of days between earthquakes.

We can do the same thing again but with different magnitude ranges this time.

It's amazing how common the lower magnitude earthquakes are and how rare the higher magnitude earthquakes are. The average number of days between earthquakes also appear to grow exponenetially as you increase the minimum magnitude.

So far, we have determined from previous graphs which areas are locations for frequent earthquakes and how frequent earthquakes are globally. However, say you wanted to go to Japan for a month for vacation and wanted to know how frequent the earthquakes are. The above plot doesn't help very much if you are concerned about how regular earthquakes are in an area. So let's narrow down the table to only include earthquakes in Japan.

Aftershocks

Typically after a large earthquake, there are aftershocks, which are smaller earthquakes near the epicenter of the main earthquake. There should be a relatively high number of earthquakes that occur afterwards then. Our loose definition of an aftershock is going to be an earthquake that occurred at least within 10 days of the main earthquake and is no further away from the epicenter of the main earthquake by 1 degree. We're going to investigate if there's a relationship between the magnitude of the main earthquake and the number of aftershocks it produces.

More Data

This time we are using earthquake data from a rectangular region around Japan with a lower minimum magnitude. This is so we can see smaller aftershocks that are under 6.0. I am still using the same website to collect the data: https://earthquake.usgs.gov/earthquakes/search/.

The options if you want to reproduce the results are:
[30.902,45.706] Latitude
[128.848, 145.459] Longitude
1990-01-01 00:00:00 - 2021-12-19 23:59:59 Time
4.5+ Magnitude

There is a weak correlation between number of aftershocks and the earthquake's magnitude. There were a large number of aftershocks for the Great Tohoku Earthquake of 2011. But besides that, all of the other earthquakes seemed to have very random values for the number of aftershocks.

Failing at predicting future earthquakes

Conclusion

Earthquakes are very complicated and are essentially impossible to predict right now. It's simply not possible to predict earthquakes from the patterns of previous earthquakes, and there needs to be more data. Although we can't predict them yet, we can still gain a lot of insight into the nature of earthquakes with the help of data visualization techniques. There is an especially large number of earthquakes on convergent plate boundaries. The world is bound to get an 8.0+ magnitude earthquake about once every 11 years or so.

Further Resources

More earthquake data: https://data.noaa.gov/metaview/page?xml=NOAA/NESDIS/NGDC/MGG/Hazards/iso/xml/G012153.xml&view=getDataView

Deaths from earthquakes: https://data.noaa.gov/metaview/page?xml=NOAA/NESDIS/NGDC/MGG/Hazards/iso/xml/G012153.xml&view=getDataView

What are earthquakes: https://www.usgs.gov/programs/earthquake-hazards/science-earthquakes

Magnitude doesn't mean everything: https://www.usgs.gov/programs/earthquake-hazards/modified-mercalli-intensity-scale