While India this year may have recorded its highest monsoon rain in 25 years, an analysis suggests that new monsoon models, called the Monsoon Mission Coupled Forecast Model (CFS), deployed by the IMD over the last decade don’t do better than the older ones in long-range forecasting.
This year, India ended up with 10% more monsoon rain (or 110% of the long period average LPA of 887 mm) than usual. However, none of the agency’s models tuned to capture long term forecast trends warned of this. The IMD’s workhorse statistical models said in its last update on August 1 that All India Monsoon Rainfall (June-September) would be 96% of the LPA. The CFS model in April said the monsoon would be 94% of the normal and updated to 99% in August.
Accurate in short term
However, the IMD models that forecast two weeks ahead (called extended range prediction) did warn of increased monsoon activity, as did short-term forecast models (that gauge weather three days ahead). However, the Department doesn’t use these to update estimates of the anticipated all India figures. The newer models were developed as part of a ₹1200 crore ‘Monsoon mission’ that has been underway for over a decade and were meant to improve both short term and long term forecasts.
A perusal by The Hindu of the forecast abilities of the CFS model, used since 2012, show that only twice — in 2013 and 2015 — did the CFS model get the monsoon right. It predicted 104% and 86% of the LPA and India ended up with 106% and 86% respectively. Like the CFS, the older statistical model also got 2015 right — when India saw one of its severest droughts. It also got 2017 right. The statistical model said the country would get 98% and India ended up with 95%. The CFS said 100%.
Independent scientists said they were “surprised” that IMD didn’t update its August forecast to reflect the quantity of the torrential rains in August and September.
“By middle of July, it was apparent that the Indian Ocean Dipole (an anomaly in surface temperatures in the Indian Ocean that correlates with good rains) was favourable,” said K.S. Krishnan of the Indian Institute of Tropical Meteorology, Pune, who works on long term climate forecasting.
“This strengthened in subsequent weeks and it was evident it would lead to very good August and September rains. So I'm surprised that they didn't warn of higher rains.”
Highest since 1983
On August 1, the IMD said August and September rains would be “100% of what’s normal” for the two months with an error window of ±8%. The country ended up with 130%, the highest since 1983.
On September 10, scientists at the Indian Institute of Science, Bengaluru, and the INCOIS, Hyderabad, said the monsoon rains in August and September would be “above normal” — based on data updated to August 10. This was due to the diminished effect of an El Nino (a temperature anomaly in the eastern Pacific, that frequently dries up monsoon rains) that has been compensated for by acitivity in the Equatorial Indian Ocean. Their report appeared in the September 10 issue of the journal Current Science.
IMD scientists admitted that the dynamical models were yet incapable of factoring in changes in the Indian Ocean a month or two in advance. “Anomalies in the Indian Ocean develop rapidly, unlike El Nino which has an 18 month cycle and can be anticipated well in advance,” said D.S. Pai, who heads the IMD's climate forecasting division. “The excess rain in September was attributable to the low pressure systems that suddenly developed. This can’t be anticipated beyond a few days,” he added.
Dr. Pai said the dynamical models gave a good “sign” rather than getting the numbers right and it could be customised for predicting summer heatwaves and cold-waves. It would need much more time to be fully integrated into India's monsoon forecasting system.
Prior to the monsoon hitting the Kerala coast on the June 1, the IMD provides an official forecast in April and an update by late May and yet another update in early August.
Traditionally it has relied on its statistical database of over a 100 years to find correlations between certain weather parameters such as temperatures in the Indian ocean, or the warm water volume in the Pacific, for instance, to estimate the chances of a good monsoon — or a drought. Over the years some parameters have been dropped, some added. But, in consonance with emerging trends in meteorology, the IMD too came around to the view that a new approach to forecasting was needed.
Dynamical models employ a different approach to forecasting the monsoon. Roughly, this relies on capturing the interactions between the land, ocean and atmosphere and tracking how the changes in each affect the other. The land, atmosphere and ocean state at a particulate time, generally March, is mathematically simulated on supercomputers and extrapolated into the monsoon months.
The dynamical model is also called the Climate Forecast Model (CFS) and has been developed based on a climate model developed by the National Center for Environmental Prediction (NCEP), U.S. and it has been implemented on the Prithvi High Performance Computers (HPC) at Indian Institute of Tropical Meteorology (IITM), Pune.