Forecast for mets: grim

Meteorology isn't what it used to be

predicting monsoons may become costlier now. An increase in the frequency of extreme weather events like intense rainfall has magnified the errors in observations and calculations required for predicting the weather.

Chances for an error occurring depend on these parameters: the location of the observation unit, time period and the infrastructure involved.

Weather data like moisture level, temperature, sunshine hours and cloud cover are fed into complex models which are designed to mathematically calculate and predict the weather of a future date. Such data are not exactly the same at different locations within the same region. Due to this variation errors creep in. Besides, none of the existing models is perfect and this contributes further to errors.

The margin of error grows as the period for which prediction is made increases.For example, the error in weather predicted for a date three days later will be greater than that predicted for a day ahead. A study says the rate of growth of errors has increased. Between 1978- 2004, the error doubled in 1.5 days as compared to 3 days recorded in the 26-year-period before that.

The study, carried out by the Indian Institute of Tropical Meteorology, Pune, said the increase in the rate of growth of errors may be due to increased instances of high intensity rainfall. And as scientists earlier argued, the reason for that was rapid warming of the sea surface

A team led by B N Goswami, the director of the institute, studied rainfall data of the past 104 years and compared the deviation of the data of each day from the normal rainfall data. They divided the 104 years into four quarters and calculated the rate of growth of error every quarter. "Errors in prediction come from errors in observations and from model inadequacies. Since the rate of growth of errors has increased, we require a quantum leap in improvement of prediction models for useful predictions," said Goswami.

The observation stations in India are placed at a distance of 50 to 100 kilometres. In the US, stations do not stray beyond a distance of 20 kilometres from each other. Improving observation and hence prediction would require increasing infrastructure (installing more observation stations) and that would cost that much more, added Goswami. The study was published in the April 23 issue of Geophysical Research Letters.