Unpredictable Models
Listen in RealAudio
Email your weather question

Nowadays, no meteorologist would dream of attempting to forecast the weather without consulting at least one, and more likely a range of numerical models.

Hi, I'm Bryan Yeaton and this is the Weather Notebook.

Models have become increasingly essential to the forecast process. They've also improved . The models are more complex and realistic in replicating conditions in the air and on the ground, and they're more more refined in the imaginary grid they superimpose on the atmosphere. But you could also say they're a bit weird.

Take the grid, for example. A forecast model works by breaking down the continuous flow of the atmosphere into a series of cubes and then assigning numerical values to all the cubes' corners. These are known as grid points. The closer together the grid points, the more accurate the forecast ­ at least that's the theory. Back in 1986 when the first high resolution models became operational, the distance between grid points was 85 km; by the start of 2002, it's expected to be 12 km. So, theoretically, current models should be almost 4 times more accurate than their 1986 predecessors. Why aren't they? Part of the problem, says Ralph Petersen of NCEP, is that we just don't have the detailed, localized data geared to these finer resolutions.

Thanks today to writer David Laskin of Seattle. The Weather Notebook is a production of the Mount Washington Observatory and is supported in part by the National Science foundation. Do you have a question or comment? Call us at 1-888-724-R-A-I-N. That's 1-888-724-6001.