New Delhi: Starting next year, the Indian weather office’s monsoon forecasts will no longer be based on historical data but on a new complex model that runs on a supercomputer—a move that the India Meteorological Department (IMD) expects to significantly increase the accuracy of predicting the rains that hold the key to the success of India’s summer crop.
The new model, one of the so-called global circulation models (GCMs), are already popular in most other countries, but while a top official of IMD said it would increase the accuracy of the weather office’s forecast by 85%, an expert said such models are usually accurate only for short-term forecasts.
New method: The India Meteorological Department premises in Pune. Ashesh Shah / Mint
IMD’s new model has been facilitated by a €13 million (Rs88 crore) deal with Meteo France International, France’s weather agency. Meteo France will train IMD scientists and offer infrastructure facilities to facilitate this shift, said P. Chakrabarti, who’s in charge of modernization at IMD.
Chakrabarti added that the initiative is part of the Rs900 crore modernization drive under way at IMD, but didn’t disclose other details of the pact with the French agency.
He said that by 2010, the weather department will start using the new model, and bury its traditional method of monsoon forecasting by using a treasury of historical data.
Worldwide, climate scientists and atmospheric physicists use GCMs for medium- and long-term forecasts. Most analyses of climate scenarios, on which the Intergovernmental Panel on Climate Change (IPCC) bases its pronouncements of climate change, are based on such GCMs.
Over the past few years, IMD has relied on a mix of statistical approaches and some GCMs to prepare the monsoon forecast. Once IMD completely adopts the new model, it hopes to bring down the error margin in predicting rainfall to 1%. Currently, this fluctuates between 5% and 10%.
“We will (then) be able to see significant improvements in the way these forecasts are prepared,” Chakrabarti said.
“GCMs require continuous, highly accurate data from as many locations as possible. To this end, we have already upgraded our balloon network (which is used to measure temperature, wind speed and other variables from different heights) and the results of that will reflect in this year’s forecast itself,” Chakrabati added.
Mint had earlier reported how a shortage of these balloons was affecting the quality of IMD’s weather forecasts.
The annual monsoon forecast is eagerly awaited by policymakers, farmers and commodity traders. Agriculture is still rain-fed in India, and 60% of the country’s workforce depends on farming (which roughly contributes 17% to India’s economy). IMD’s predictions of a “normal monsoon” or “drought” are crucial to Central and state level planning.
IMD frequently updates its predictions once the monsoon begins, but these updates often come too late for farmers to change their sowing plans.
“That’s what we will change. Since we can track the weather better, we can give our farmers at least a couple of weeks’ notice before an unforeseen weather anomaly,” Chakrabarti said.
IMD, the only Indian institution allowed to publicize its monsoon forecast for India, has, until now, relied on historical data accumulated over 139 years to prepare each year’s annual forecast.
However, meteorologists in international organizations such as the UK-based Hadley Research Centre have, over two decades, used GCMs—which require supercomputers—to issue weather forecasts.
GCMs use complex mathematical equations that factor in weather variables. These equations are then crunched by supercomputers that come out with weather forecasts.
Not everyone is convinced about the efficacy of GCMs.
Madhavan Rajeevan, a scientist at the Indian Space Research Organisation, or Isro, who was involved in preparing monsoon forecasts when he was at IMD, says GCMs are good only for short-term forecasts and have frequently done a bad job at getting the Indian monsoon right.
“IMD can switch to GCMs and they have the capabilities. But such models still don’t adequately capture the Indian monsoon.” he said. “A good test of such models will be in their ability to predict a drought at least 20 days in advance. I am not sure any GCM in the world can do that for the Indian monsoon.”