Weather Prediction: How Met office knows what the monsoon will do next summer | India News

Thara Prabhakaran can stomach all the air turbulence you throw at her. As a weather scientist she flies into the heart of rain clouds to collect samples and conduct experiments, often to a height of 9km. Things sometimes get scary there. Once, during an experiment to understand how water droplets and ice particles are formed, and how aerosol and pollution impact these processes, power went out inside the aircraft and the probes froze. They had to drop altitude to let the ice melt.
That didn’t stop Prabhakaran from going up again, though. “These observations help modify the dynamical models for forecasting weather, climate, and the monsoon rain,” she says.
Prabhakaran is a senior scientist at the Indian Institute of Tropical Meteorology in Pune, which along with the Indian National Centre for Ocean Information Services, Hyderabad, and the National Centre for Medium-Range Weather Forecasting, Noida, monitors weather phenomena – heatwave conditions, cyclones, the monsoon’s arrival – and develops mathematical weather prediction models.
These organisations help the India Meteorological Department (IMD), which will soon complete 150 years, to use satellite data, high-performance computing systems and historical data to forecast the monsoon, and make daily, weekly, and seasonal weather predictions.
Evolution Of Forecasting In India
Sir Henry Blanford, an imperial meteorological reporter who gave India its first official seasonal monsoon forecast on June 4, 1886, founded IMD in 1875. Before that, he had used the inverse relationship between Himalayan snowfall and monsoon rainfall to prepare tentative forecasts from 1882 to 1885.
In 1906, Sir Gilbert Walker used a more complex prediction model based on the link between monsoon rainfall and global circulation parameters.
With time, Indian weather models became richer. Vasant Gowariker’s monsoon prediction model based on 16 global and regional parameters served well from 1988 to the end of the century. But when 2002 – forecast to be a normal monsoon– turned out to be a drought year, a better model had to be made.
An IMD team led by M Rajeevan, former secretary of MoES (Ministry of Earth Sciences), analysed the existing models and came up with a two-stage forecasting system in 2003. “Its first prediction in mid-April was based on eight parameters, and the second in May on 10 parameters. These were followed by a rain forecast for July’s agricultural operations,” says Rajeevan. As the technology evolved, Rajeevan and his team designed the statistical ensemble forecasting system in 2007. But 2009 was a drought year which exposed the limitations of the seasonal forecast models, both statistical and dynamical, Rajeevan and CK Unnikrishnan from National Atmospheric Research Laboratory wrote in a 2011 issue of Breeze – newsletter of the Indian Meteorological Society’s Chennai chapter. “The errors persisted because forecasts were based on empirical data and on dynamical models built on atmosphere-ocean coupled models,” says Rajeevan.
When the government sought forecasts of the spatial distribution of seasonal rainfall along with regional average rainfall forecasts, Rajeevan’s team, consisting of senior scientists DS Pai and OP Sreejith, implemented a multi-model ensemble forecasting system in 2021. It was based on eight coupled global climate models from different prediction and research centres.
A multi-model ensemble has the advantage of presenting a range of future weather possibilities. At present, probability forecasts for rainfall and temperatures are made separately for all 12 months. These are in addition to the seasonal forecasts for the southwest monsoon (June-September), northeast monsoon (October-December) and the premonsoon season (March-May), Pai says.
Progress With Monsoon Mission
Getting monsoon predictions right is crucial for India because if it rains even 10% more than normal, flooding fears arise, and if it rains 10% less, drought is a possibility.
That’s why MoES launched the National Monsoon Mission in 2012 to improve India’s weather and climate forecasts. It combined ocean, land, atmosphere, and sea ice models to make long (seasonal) and extended (four weeks at a time) forecasts, and used standalone atmospheric models for shortto medium-range (7-10 days) predictions.
Mission head Suryachandra Rao says they borrowed the coupled forecast system used at America’s Climate Prediction Center. “Supercomputing facilities in India were enhanced by the MoES to support research and operations. From 2017, IMD startedusing this system to generate experimental seasonal forecasts for the monsoon along with an operational statistical ensemble forecasting system. ”
As a result, the models have become more accurate at the micro level. They can now forecast weather over a radius of 12km, down from 38km before the Mission was set up. IMD now has a fullfledged dynamical seasonal prediction system which serves the whole of South Asia.
Why Forecasts Go Wrong
Tech upgrades play a big role in improving the weather models. For example, in the past 10 years the weather bureau’s processing power has gone up from one petaflop (measure of computing speed) to 10 petaflops. It now has 37 radars in place of 14, and the number of automated weather stations and rain gauges has doubled.
They also have two satellites as against one earlier. ‘Cyclone man’ Mrutyunjay Mohapatra, IMD’s director general of meteorology, says satellite observations are received every 15 minutes and analysed every three hours to determine the status of the atmosphere, oceans and land.
Notwithstanding all these advances, predictions still go wrong, and O P Sreejith, head of climate monitoring and prediction services at IMD Pune, says it is hard to make a perfect forecast in the tropics because many parameters change quickly. It is equally difficult to predict weather in the mountainous regions. “With better computational resources, more observation data and research, predictions can be improved. Still, forecasting tropical weather is challenging, as is 100% accuracy of the long-range forecast,” he says.
This is partly because even the best of weather models have their biases. For example, many climate models have a dry bias over central India during the monsoon. Sreejith’s team looks at different models and generates a forecast after correcting for their biases. He says the multi-model ensemble forecasting system which uses models from India, the US, Japan and Europe, has been used to predict monsoons since 2021 with “good results”.
Mohapatra agrees forecasting is difficult – “we make the best educated guesses based on scientific evidence” – but says they have had a good run so far. “The landfall point error for cyclones was about 150km in 2010, it is about 25km now. The five-day forecast today for heavy rainfall is as accurate as the one-day forecast in 2010. ” Their forecast for Cyclone Biparjoy in June was spot on.