More accurate forecasting models being developed

Bangalore, May 9: Encouraged by the good performance of existing long range forecasting models, the Indian Meteorological Department IMD...

Bangalore, May 9: Encouraged by the good performance of existing long range forecasting models, the Indian Meteorological Department IMD is now developing improved versions with better lead time and accurate weather prediction especially for the Indian monsoon rainfall.

The new models gain importance specially considering that a major part of the Indian economy driven by agriculture is still heavily dependent on the performance of the monsoons.

Since 1988, the IMD has been predicting the behaviour of Indian monsoons using basically for long-range forecasting models which are the parametric model, power regression PR model, dynamic stochastic transfer DST model and the multiple regression MR Model.

As the performance of the parametric and the PR models were found to be better, these were used for issuing the official qualitative and quantitative long-range forecast of the IMD, while the DST and the MR Models were basically used for complementing the other two models for operational forecasts.

In his presentation on quot;Long-range forecastingquot; at the recent one-day seminar on atmospheric science here, Dr V Thapliyal, deputy director general, meteorology weather forecasting, IMD, Pune, said now there was a pressing need to improve the lead time for LRF to enable providing as accurate as possible forecasts by April for the monsoons which start in June first week.

Thapliyal said the existing models, especially the parametric and the PR models use a group of 16 parameters from the atmosphere, land and ocean to provide long-range forecasts. However,the forecasts could be delivered by the month of May, just a month in advance of the onset of monsson.

The IMD had zeroed in on the PR model using 16 parameters PR-16 and the DST model to develop a new model for improved and more accurate long-range forecasts, he said.

Story continues below this ad

IMD8217;s new model called the power transfer PT model combines the qualities of the DST as well as the PR models. He said the DST model used the basic climatic association as well as both the time and space components, while the PR model used the basic climatic association and the space component.

The PT model used 12 parameters decided after going through the weather data of 28 years in the period 1967-94. The new model PT-12 and others which were being developed like principal component regression model and the hybrid artificial neural network model would aim at minimising the errors in the existing operational models which were showing an increasing tendency.

The IMD, he said, had also developed 8-parameter models of the existing operational models, of which the PR-8 was considered the best as it could provide forecasts by the end of March to enable the preliminary LRF to be issued in early April for the monsoons.

He said the new models would also be improvised in order to give accurate forecasts in smaller areas in a shorter period. So far, the long-range forecasting techniques are confined to provide forecasts for the whole country.