Exploring the Probabilities

“We learn geology the morning after the earthquake.”
-Ralph Waldo Emerson

Within the area expected to be impacted by a Cascadia Subduction Zone (CSZ) earthquake, only 5% of the population is estimated to be 2-Weeks-Ready. Income, language barriers, and apathy play a role in why that number isn’t higher. Another factor, according to Dr. Howard Kunreuther, Co-Director of the Wharton Risk Management and Decision Processes Center, University of Pennsylvania, is temporal myopia, part of our human nature that often leads us to prioritize short-term (nearsighted) over long-term benefits when making decisions. His FEMA Prep Talk is worth watching.

The way we discuss the CSZ risk plays a major role, as well.

It starts with a question we have all asked—what are the chances that this thing will happen?

Image Credit: National Park Service

The most common answer we hear is:

“There is a 37% chance the earthquake will occur in the next 50 years

For comparison, let’s think about retirement for a minute. Retirement is something most people strive for.

*We know it’s going to happen.
*We know when—usually within a 25 to 30-year timeframe.
*The timeframe is shorter than the 50-year outlook in the earthquake probability
*…and many people still struggle to prepare for it in the United States.

Framing the CSZ risk in a 50-year timeframe makes that risk easy to dismiss altogether or at least push off—largely due to that temporal myopia mentioned above. By the end of this page, it’s my hope that you won’t decide whether to take this hazard seriously based on the “37% chance in 50 years” statement. Before I can do that, though, we need to:

Explore the Probabilities

Step 1: Choose the mean and standard deviation you are curious about from one of the rows below.

*The Gaussian (normal distribution shown on the right of the image) and Log-Normal (shown on the left of the image) results are listed in the image below for 50-year outlooks.

T1 to T19 refer to specific earthquake ‘names’.

Step 2: Use the mean and standard deviation you chose in step 1 to edit the input fields in the embedded Excel spreadsheet below, provided by Oregon State University Professor Chris Goldfinger and his colleagues.

Step 3
Now edit the “Start Year” or “End Year” in the embedded Excel sheet above to see how the probabilities (in blue) change.

According to Professor Goldfinger, the Log-Normal model fits the data for our region better, but the Gaussian model (think bell curve) is better at predicting over time.

How Has The Risk Changed Over Time?

If you took the time to play with the numbers a bit, were you surprised by what you saw? Take a look at this chart, as well as the comparison images below. The Gaussian (normal distribution/bell curve) model continues to show a growing risk over time (assuming an earthquake doesn’t happen).

However, the Log-Normal probability model shows virtually the same risk whether the “start year” entered was 50 years ago, current day, or projected 50 years out, all rounding up to 37%. In other words, this model provided roughly the same risk in 1982 as it will in 2072.

The Log-Normal probability figures from above stretch to 2 decimal places. Did you notice that the values for the start years of 2012 and 2022 match and are the largest two values?

Expanding the values to 5 decimals, here’s the reality. Regarding the ‘37% chance of a CSZ earthquake in the next 50 years”, the Pacific Northwest was at the VERY TOP of the log-normal curve in 2017.

That’s right. According to the model, our risk of an earthquake was highest in 2017 and has been DECLINING ever since. I wish that was the case!

Unfortunately, the probability of this occurring doesn’t decrease over time. Pressure continues to build where the Juan de Fuca plate is stuck beneath the North American plate. It will continue to build until, eventually, the built-up stress ruptures the fault.

This is just one reason why I recommend looking at intervals rather than probability figures.

Did you know:
*81% of the time, during the most recent 10,272 years of history, the fault has not had to wait 323 years for the strain to break it?

Do you also know:
*93% of the time, during the most recent 6,031 years of history, the fault has not had to wait 323 years for the strain to break it? That’s right. 27 of the most recent 29 intervals have been SHORTER than 323 years. The two that were longer measured 330 and 344 years.

Check out Surviving Cascadia’s Likelihood of an 8.0 page for more info! It’s good to know where the numbers come from, but it’s also good to know how much stock to put into the numbers.

A Better Model to Measure Risk?

There are places in the world where earthquakes can occur in clusters. You can read more about the possibility of the Cascadia Subduction Zone experiencing earthquake clustering on the Likelihood of a 9.0 page. Neither the Gaussian nor the Log-Normal calculation accounts for the clustering behavior of the fault, so Professor Goldfinger cautions against getting too consumed by the numbers.

Professor Seth Stein and his team are currently working on a new probability model. For more information, visit their research publication, A More Realistic Earthquake Probability Model Using Long-Term Fault Memory, explanatory YouTube video, and PDF with graphics like the one.

We live in a region where these large earthquakes occur. While we don’t know exactly when the earthquake will happen, there is no doubt that the fault is going to rupture at some point. Being 2-Weeks-Ready is the right choice.

Of note, the “37% chance” isn’t the only number out there. Looking at the percentages below—all based on a 50-year outlook—it can feel like the level of risk depends on who you ask. Click the buttons to read the articles.

Check out this Northwestern University paper for more information on why there are so many different probabilities out there. The answers are complex because earthquakes are complex. In this COGS Interview, Professor Goldfinger explains why we hear so many different probabilities. I highly recommend watching the entire interview, as it’s packed with information, but for a brief intro, start the video at 1:02:00.

%d bloggers like this: