BioEd Online

When there's a cyclone around, you want to know where it is headed. But that may be harder to predict than we thought.

Melinda Peng and Carolyn Reynolds of the Naval Research Laboratory in Monterey, California, have found that two simultaneous tropical cyclones can influence one another in ways that are subtle and difficult to forecast, even if they are nearly 2,000 kilometres apart.

Weather forecasters currently use computer methods to predict the course of a cyclone - a fierce storm system that gathers around a region of low atmospheric pressure. But when there is another cyclone in the vicinity, Peng and Reynolds warn, small differences in the initial meteorological conditions used in the model can significantly affect the prediction of the first cyclone's path and behaviour two days later.

The distance over which cyclones can affect each other is surprising, the researchers say. Previous observations of twin cyclones had led experts to suspect that they could not affect one another if they were more than 1,500 km apart.

Peng and Reynolds used a cyclone-prediction model developed by the US Navy to forecast the trajectories of two real, coexisting cyclones, Katsana and Parma, that formed in the western Pacific Ocean in late October 2003.

The predicted courses corresponded fairly closely to those of the real storms over a two-day period. But a careful look at the results showed that the forecast for Parma depended on that for Katsana, despite the fact that the two storms were more than 1,500 km apart for most of their lives. (Their closest approach was 1,333 km.) Parma did not, in turn, seem to affect Katsana, which was the stronger of the two cyclones.

What's more, for cyclone Katsana, the researchers showed that the predictions would have been altered significantly if the model's initial conditions 500 kilometres from the cyclone's centre were changed slightly, showing how surprisingly sensitive these storms can be. "The forecasts can be very sensitive to the initial state," Peng and Reynolds say. "Although the forecasts in these cases were good, our tools highlighted the potential for initial errors to grow."

Two cyclones together is not uncommon - it happens about three times every two years in the western North Pacific Ocean, and about once every three years in the Atlantic.

Despite the potential problems for cyclone forecasting, Peng and Reynolds say that their results have a positive spin too. The new technique they used for analysing cyclone interactions, called singular vector diagnostics, seems to give them a better handle on the problem than standard methods. "It may tell us how much confidence we can have in our forecasts," they say.