Rainstorm over field

Predicting rainfall persistence amid climate change

August 12, 2019

As our climate rapidly changes and human populations continue to stress water resources, predicting rainfall and its persistence – amounts of rain, when it will rain, and how long it will rain – across the U.S. is vitally important. It defines the ways in which farmers, conservationists and even developers think about our ecosystem. But most climatology research has focused on the country’s wettest days, discounting the light rain that also fill streams.

Researchers from the University of Colorado Denver and University of Illinois at Urbana-Champaign used 70 years of precipitation data to explore the predictability of showers and the ways in which rainfall as little as three millimeters affects our ecosystem. The scientists discovered that the number of rainy days across the U.S. has increased since 1948, and while the predictability of rainfall has grown in the West, it’s become more erratic in the East.

The study was published in the August 2019 issue of the Journal of Hydrometeorology.

“Information theory provides a new perspective on predictability,” said Allison Goodwell, PhD, lead author and assistant professor of civil engineering at CU Denver. “We’re not trying to predict the weather like a meteorologist. We’re asking different questions, like how much do we know about whether or not it will rain today, just based on knowing past rainfall?”

How the past predicts future rain

The best predictor of rain today is whether or not it rained yesterday, said Goodwell, because days of rainfall often clump together. Finding the next best predictive day was the challenge for Goodwell and her colleague, Praveen Kumar, PhD.

In the past, hydroclimatology researchers have studied “Markov Chains,” a mathematical method used to statistically model random occurrences, to describe rainfall predictability, usually focusing on the two days that directly precede the day in question.

“But the next best predictive day is not necessarily two days ago,” said Goodwell. “Often it’s seven or eight, or more days ago. When you look at it from a longer term, you might see that the dryness of 10 days ago better predicts whether or not we’re due for rain.”

Using information theory, the researchers looked back as far as 20 days for rainfall events of any size. They discovered patterns that went beyond how each event was changing.

“By comparing different thresholds to define rainfall events, we found that light rain events have changed differently from heavy events,” said Goodwell. “Until now, few scientists have looked at light rain, which also plays a role in stream flow and soil moisture.”

Reducing uncertainty in an era of climate change

Solid red/blue areas indicate regions where predictability has increased or decreased; red/blue hatching represents areas where rainfall frequency (number of rainy days per year) has increased or decreased

In an analysis of the spatial and temporal trends across the U.S., the researchers found rain events in the West became more orderly and more predictable over the past 70 years. In the East, the rain amount remained steady, but became increasingly unpredictable. The divergence between these two regions could have meaning for future predictions.

But the trends also give insight into climate change’s impact in the U.S. as droughts and flooding become more extreme, and human impact continues to change land and water resources. Goodwell hopes to eventually expand her research to include variables like temperature, flows, topography and land cover.

“This is one more way to understand how rainfall connects to the rest of the water cycle,” said Goodwell. “Predictions reduce our uncertainty and reducing any amount of uncertainty is important when it comes to our natural resources.”