IMD and Skymet differ on specifics of monsoon predictions, again

The big difference in this year's forecast by two agencies has been their inability to assess the rains in the last two months of monsoon season

For years, monsoon forecasting was a stolid subject of intense interest for determining India’s economic fortunes but rarely associated with public-private institutional rivalry. Since 2009, however, a competition of sorts has emerged between the government’s over 140-year-old Indian Meteorological Department (IMD) and Skymet, India’s largest private weather forecaster.  
 

At the heart of this genteel competition is the reliability of the long-range forecast by each: both have come into question in the past few years. This year, the differences surfaced over the onset of the monsoon and when it would enter the lull and active phases. By September, however, the divergences had widened over whether the El Nino phenomenon would impact the performance of the south-west monsoon.  

IMD suggested not; Skymet differed.

On October 1, Skymet posted an article on its website claiming credit for forecasting the monsoon correctly, indirectly criticising IMD for not getting it right.

“We have always been saying that a big break in monsoon, which could lower the overall average, happens in evolving El Nino year,” wrote Skymet CEO Jatin Singh.

Though senior IMD officials have never openly criticised Skymet, they have — behind closed doors — questioned its process of forecasting and also accused it of manipulating forecasts to suit its own business interests.

The big difference in this year’s monsoon forecast by the two agencies has been their inability to assess the rains in the last two months of the June-to-September monsoon season.

Skymet in its forecast had expected August rains to be 7 per cent below normal, for September it said rains would be 4 per cent below normal.

The IMD in its updated forecast said that the August rains would be just 1 per cent below normal, and they would be normal in September. The IMD’s forecast was with a model error of plus and minus 9 per cent.

In reality, the August rains were 13 per cent below normal and September showers 12 per cent below normal. The big and extended break in monsoon in the last two months after a fairly good beginning has raised doubts over the forecast of cumulative average rainfall.

So, what happened? 

In an interview to this newspaper IMD Director General K J Ramesh said the break in the intensity of monsoon over some parts of the country after July was a normal intra-seasonal variability and can’t be forecast in seasonal predictions that are made at the start of the year.

Skymet, on the other hand, said the monsoon break was typical of an evolving El-Nino impact, a point it had made right from the start and which IMD wasn’t acknowledging.

Irrespective of who got it right, for industries that rely on a good monsoon for growth, such breaks can be fatal.  “This year, the extended breaks in August and September, particularly over central and western India, would have impacted sales had the monsoon not revived in time,” Rajesh Aggarwal, managing director, Insecticides India, told Business Standard. 

But the disparities highlighted a key weakness in the models currently used to predict the monsoon: that they still haven’t reached the level at which detailed sales strategies can be worked out on their basis. 

As Ashok Gulati, agriculture economist and former chairman of the Commission for Agriculture Costs and Prices, pointed out, the probability of getting monsoon predictions rights four months in advance should not be expected and neither is it feasible because technological advances are limited.

“Any monsoon forecast made four months in advance must have very low probability as atmospheric conditions can go haywire within 15 days. Marginal errors made either by IMD or Skymet can be overlooked but if there is a big difference in their forecasts, then they should be taken to task,” said Gulati.

As things stand, this break is unlikely to dent the overall foodgrain production in the kharif season, but in parts of Vidharbha, in eastern Maharashtra, and Madhya Pradesh, where the initial rainfall was initially good, the rain break could impact the final yields of some crops.

The rivalry between IMD and Skymet came into the limelight in 2009 (although Skymet started operations in 2003), when the latter correctly forecast the drought. IMD in its first forecast issued in April that year had said that monsoon would be normal at 96 per cent of the long- period average, with a model error of plus and minus 5 per cent, which it accepted went horribly wrong in its end-of-season report.

The drought, which was among the worst in more than three-decades, crippled India’s foodgrain production, dented farm growth and pushed up food prices.

Thereafter, Skymet’s annual monsoon predictions started getting noticed, though it is fair to say that IMD remained the quintessential weather advisor.

After a decent run, however, Skymet’s famous accuracy went askew in 2015 (see chart) and Singh felt compelled to write an oped in the Indian Express explaining what went wrong. His explanation: they over-applied and over-engineered the forecast that year. 

“The monsoon is a beast we thought we knew intimately, but it seems to hold a few more secrets,” Singh wrote in the piece.

He went on to commend the IMD for “very correctly” forecasting the deficient rainfall, something which Singh said the agency seldom did in its long history.

The IMD works from primary information culled from satellite data, whereas Skymet uses a combination of secondary information and statistical models to make its monsoon forecasts.

Asked which of the two he deems more reliable, Gulati said it was too early to judge and any analysis of accuracy should be done over a reasonable period of time.

The IMD on its part has been constantly changing its monsoon forecast models and systems and adopting newer processes to fine-tune its long-range seasonal forecast.

Since 2012, the IMD had been mainly using the dynamical global climate forecasting system (CFS) model, which was developed under the Monsoon Mission.

For the 2017 monsoon forecast, for the first time, it used a combination of statistical ensemble forecasting system and the dynamical global CFS model. 

This constant upgrading of monsoon science means that one prediction is easy to make: the battle of the forecasters will endure for many monsoons to come.