CNET también está disponible en español.

Ir a español

Don't show this again


Scientists tackle midrange weather forecasts

Computer models of a rainfall oscillation pattern can give a glimpse into weather patterns one to three weeks in the future, according to a NASA scientist.

This image shows the alternating periods of stronger and weaker rainfall, called a Madden-Julian Oscillation. Blue and red represent high and low rainfall across the equatorial region of the Earth. The 1987-1988 season had particularly pronounced MJO activity, and MJOs are strongest in the Indian Ocean region. NASA

SAN FRANCISCO--Scientists are trying to peer a bit further into the future than the typical five-day weather forecasts available today.

Forecasting weather is a notoriously tough challenge that combines physics modeling, data collection, and computer processing--and unlike many scientific problems, pretty much everyone on the planet cares how well it's done. But forecasts today peter out after a few days, leaving a cloud of uncertainty (forgive me) that only lifts when it comes to predicting seasonal weather phenomena such as El Nino.

Scientists are now getting a handle on intermediate-term forecasts by computer models of a particular type of large-scale weather phenomenon called a Madden-Julian Oscillation discovered in the early 1970s. These MJOs are linked to phenomena including Atlantic hurricane seasons and South Asian monsoons, and modeling them with a computer can "provide new predictions with lead times of one to three weeks," said Duane Waliser, principal scientist for water and carbon cycles at NASA's Jet Propulsion Laboratory, speaking today at the American Geophysical Union conference here.

When MJOs occur, rainfall alternates between periods of greater and lesser intensity, typically with each period lasting 40 to 50 days and the pattern gradually traveling east across the planet, Waliser said. Although they take place at equatorial latitudes, their effects extend farther north and south into the middle latitudes, Waliser said.

Only recently have MJO computer models become very good at predicting the actual phenomena, he said.

"Five to ten years ago, you would have been hard-pressed to find any models that represented it that well at all," Waliser said. Now, though, multiple institutions are working on the problem, and scientists actually have a choice of models.

One problem with using MJOs to predict medium-term weather, though, is that they are only intermittent phenomena that take place roughly two to six times per year. "If it's there, we have something to go on. If it's not, then we don't."