Red to blue scale image of Pacific ocean temperatures

Statisticians thrive on the secret stories of numbers

By Joshua Hewitt, 2018-2019 Sustainability Leadership Fellow and Ph.D. Candidate in the Department of Statistics

We’re living in the age of big data, so let’s look at some: “0.67, 0.85, 0.98, 1, 0.71, 0.15, -0.41, -0.87, …” Wait, let’s look at that same data again…

Red to blue scale image of Pacific ocean temperatures
Sea surface temperatures from the ERA-Interim reanalysis project.

Much better. As an image, the data show average sea surface temperatures from winter, 1981. Sea surface temperatures are above average in the red colored regions and below average in the blue colored regions. While we see rich patterns in the data, computers only see the numbers that we first tried to look at. Statisticians and Climate Scientists both have many tools they use to help computers search for patterns in data.  The best tools are based on ideas we can all understand, and stories.  Good stories inspire strong statistical analyses, but Statisticians often listen for the secret stories of numbers.

Building a sustainable future depends on understanding what problems communities face and how data can help suggest causes and effects.  Listening to stories—or even only pieces of stories—can help inspire new ways to use data to solve problems.  Sustainability projects, in particular, challenge us to analyze complex data about how weather patterns impact public health and the environment. Naturally, it is important to have as many people as possible improving the ways we work with data.  Statistics is not just about math.  It’s about math that helps us explore our world.

For example, Peruvian fishers have long known that Peru’s coastal waters periodically become abnormally warm. As coastal waters warmed, fishers would catch more fish, and better ones at that!  However, the number of commercially important Peruvian anchoveta would diminish.  In the worst years of warming, the Peruvian anchoveta was almost overfished to the point of collapse, threatening fishery jobs and economic disaster.  Fortunately, the Peruvian government has avoided complete disaster by working hard to regulate fishing, especially in El-Niño years.

Water scientists in the 19th and 20th centuries had helped link patterns in Peruvian fisheries, worldwide drought conditions, and more phenomena to natural cycles in worldwide ocean temperatures. Today, these natural cycles help scientists and meteorologists anticipate changes in seasonal weather patterns, like average temperature and precipitation. These cycles can also make extreme events more likely, like floods and droughts. But the secret stories of numbers lie in the ideas behind the statistical tools that scientists and meteorologists used.

Statistical analyses of historic ocean temperatures helped scientists identify recurring patterns in sea surface temperatures. Additional analyses linked some of these patterns to predictable changes in weather patterns and ocean conditions that impact people, like the Peruvian fishers.  To date, one of the most predictive ocean temperature patterns remains the El-Niño pattern, shown below. However, the statistical tools scientists often use to identify recurring patterns are largely taught to look for the same type of pattern.  As such, there may be “secret” patterns remaining in the numbers that have been overlooked by traditional analyses.  What if we could teach computers to search for more patterns?

Pacific Ocean surface temperature pattern from El Niño
Natural cycles in ocean temperatures impact global weather patterns. For example, some regions are more likely to experience warmer than normal temperatures, increased precipitation, or extreme conditions like droughts when sea surface temperatures strongly resemble the El-Niño pattern in Pacific Ocean temperatures, shown above.

What if we designed a statistical tool that could look more freely within data for patterns? I have developed a statistical tool that does this and found that such a tool can more reliably estimate changes in seasonal precipitation, for example. The new tool has the freedom to search a near-endless number of patterns, like the examples pictured below. Perhaps unsurprisingly, the El-Niño pattern scientists have known about for decades is still highly effective. But, in developing a new statistical tool, we taught computers about important features that ocean temperature patterns should have. As a result, the computers we trained became better at and more confident in finding patterns in ocean temperatures that impact weather patterns.

Four Pacific Ocean temperature patterns
Examples of a wider range of ocean temperature patterns that new statistical tools can study. New tools help determine if patterns impact weather patterns in different regions of the world.

Many people know that statistics uses lots of math, but as many people might not realize just how much of that math is inspired by the stories we tell each other about our experiences. Next time you see a story that involves data, pay attention to how the numbers add to the story.  If you dig deep enough, you might even see something that people have not considered, like looking more broadly for different types of ocean patterns.  In doing so, you might accidentally find or solve an interesting statistics problem while building a more sustainable future.

Share this post

Share on facebook
Share on google
Share on twitter
Share on linkedin
Share on pinterest
Share on print
Share on email