Hurricane forecasting made better

By Ajay Singh

A month after Hurricane Katrina devastated New Orleans in 2005, the National Hurricane Center warned that another storm, Rita, was headed for Houston, Texas. As it happened, Rita spared Houston and slammed into the Texas-Louisiana border, about 100 miles to the east.

An error of that size may seem large, especially to residents of Houston — in the chaos caused when 2 million residents hurriedly evacuated the city, more than 100 people were killed in accidents or from exposure. But the inaccuracy is about average for a hurricane forecast two days before landfall — and a substantial improvement over forecasts issued as recently as a decade ago.

One reason for the improvement is that hurricane forecasters now employ the results of many computer models in their calculations, based on the notion that many models are better than just one. But why do different models produce different forecasts? Robert Fovell, a professor of atmospheric and oceanic sciences at UCLA, decided to find out. A cloud dynamist, he teamed up with Hui Su of the California Institute of Technology's Jet Propulsion Laboratory in Pasadena to conduct a series of experiments.

How fast raindrops fall is linked to hurricane direction

The researchers took a computer-simulated model of Hurricane Rita and ran it repeatedly, changing in the model each time only certain microphysical properties of storm clouds. The most important variation they made was the speed at which condensation particles in the clouds fell. To their surprise, the researchers discovered that this one factor — how fast raindrops, both small and large, fall — can have a big effect on where a hurricane is headed. The reason? The slower particles fall, the farther they can spread from the hurricane's core, directly affecting the storm’s horizontal size, the researchers surmise. Among other factors, a storm's size influences its motion.

Fovell's and Su's conclusion is published in the November issue of Geophysical Research Letters, the journal of the American Geophysical Union. Their research promises to further improve hurricane forecasting.

"It has long been known that how microphysical processes are handled has a large impact on hurricane intensity," said Fovell. To predict the paths of hurricanes, known as "track forecasting," meteorologists use computer models and a range of other macroscopic data. But he and Su identified something that was not explicitly cited in previous research to address the problem.

There are many reasons why various computer models give different forecasts, explained Fovell, "and now we have added another — condensation particles."

Seeking the "why" of the cloud particle factor

The idea that even relatively subtle differences in cloud particles can affect forecasts first dawned on Fovell in 2004, when he began comparing his computer-generated forecasts of storms with those of the Miami-based National Hurricane Center on an almost-daily basis.

"I found my forecasts were consistently better — not by much, but just better," recalled Fovell, adding: "Naturally, I wanted to know why."

Fovell is now waiting for reaction from his peers. Meanwhile, the the researchers have already begun to look for answers to the next question. "Our research does not say why condensation particles impact on hurricanes," he said. "That is part of the next study."

While Fovell is happy that one puzzle has been solved, another, more complex one has arisen. "Our research has shown that cloud microphysics is important to the track problem," he said. "The next step is to improve how models handle these processes — and that will be very difficult."