Informing out the development, MoES Secretary Madhavan Rajeevan said, “We are ready to roll out a dynamical model, which is being tested at the Indian Institute of Tropical Meteorology (IITM) Pune for a decade. The model would be ready for operational purposes next year.”
Explaining about functioning of the model, Rajeevan said, “A dynamical monsoon model works by simulating the weather on powerful machines and extrapolating it over desired timeframes.”
Such models have been used for research purposes in India for decades; the India Meteorological Department (IMD) has never integrated it into its operational forecast. Though this approach is effective in forecasting weather over a few days, using it to gauge the annual monsoon over 3 or 4 months would prove to be major challenge.
“We hope to be able to launch it next year, though, some discussions are still going on,” Rajeevan told reporters on the sidelines of a function organised to celebrate 10 years of the MoES. The ministry on Wednesday completed a decade as it was established in 2006.
The IMD relies on an ensemble model, a statistical technique that uses an average of models. These models are prepared by relying on a vast, century-old trove of meteorological data linked to the performance of the monsoon.
This traditional approach, in recent decades, has failed to predict droughts, leading to calls by several meteorologists to develop a new, modern approach to monsoon forecasting. Though the dynamical model, called the Coupled Forecast System version 2, has only a 60 per cent accuracy in forecasting the monsoon, it’s good enough for now, Rajeevan said, adding, “No doubt it needs to improve and the aim is to make that 77 per cent, but we have to start somewhere.”
For some years, the IMD has been appending the dynamical model’s forecast along with the traditional forecast. But junking the statistical model signals a new approach in forecasting the monsoon. This is a precursor to giving monsoon predictions over India's 36 subdivisions, rather than the 4 broad geographical regions, within which these 36 subdivisions are subsumed.
Moreover, a dynamical approach can be more easily tuned to account for rapidly changing global weather conditions, Rajeevan added while noting that a big confidence boost was that the dynamical model correctly signalled a drought in 2015. In 2017, the IITM is likely to use a 10 petaflops-per-second system — a 10 fold increase in computing power — for the dynamical model monsoon prediction.