Quick Dips

AI-ing the Weather

Forecasting has moved a long way beyond simple "red sky at night" observations, but the latest high-tech approach is based on a similar concept.

"In those early days people were data processing when their brain told them they had seen a certain pattern occurring several times before and they drew a conclusion from that," says Dr James McGregor, of the Institute of Geophysics in Wellington.

"What we are doing with this latest technique is precisely that -- taking observations and noting the consequences of a particular observation -- only now we are doing it for significantly more complicated phenomena and with vastly more data."

McGregor and PhD student Tim Hume have carried out a pilot study which involved examining fog conditions at Christchurch airport and analysing more than ten years of related historical data.

They examined all the fog events during that period, as well as other data that was available, such as temperature, humidity, wind direction and sunshine hours, and then fed the facts and figures into an artificial intelligence system, which tries to make sense of them and come up with a set of equations.

"At Christchurch airport, for example," says McGregor, "we have found that if you have a windspeed over seven knots then fog doesn't form. That means if you are forecasting winds above seven knots you've got a pretty strong idea that you won't get fog. That's a simple thing but there are lots of other factors which are not so straightforward..."

The researchers are using fog as an example of something which is particularly difficult to forecast. In the future they want to look at heavy rainfall, and in the very long term, McGregor is hoping to use artificial intelligence to look at climate predictions such as drought.