Can you trust the IMD's weather forecast?

Can you trust India Meteorological Department's (IMD's) forecast that the country will get a rainfall of 98 per cent of the long-period average (LPA) during the monsoon season this year?

Considering the poor accuracy record of IMD's long-term forecasts in the past, the answer to this question cannot be wholly in the affirmative. IMD has erred in its projections far more often than it has managed to get them right (see graph).

Besides, the credibility of IMD's monsoon predictions has been eroded because these forecasts have repeatedly failed to warn against droughts.

This happened in the case of all the three droughts in recent years - in 2002, 2004 and 2009. This year again, IMD has used the same model for predicting monsoon rainfall (five-parameter statistical ensemble forecasting system), which had failed miserably to foresee the severe drought last year.

Indeed, last year's fiasco stands out as a classic case of IMD's inability to precisely assess the rainfall expected during the monsoon season.

IMD went wrong not only in the first stage of the forecast that was issued in April 2009, but also in the "forecast update" issued in June, well after the onset of the monsoon - nor could it foresee the situation accurately even in an unprecedented third attempt to fine-tune its forecast in August when the monsoon had already run through more than half its course.

Its first forecast put the rainfall at 96 per cent, the second one at 93 per cent and the third one at 87 per cent, whereas the actual rainfall turned out to be a mere 77 per cent of the normal, resulting in drought in a large part of the country.

Moreover, the record of IMD's long-range monsoon predictions has not been satisfactory enough to inspire any confidence in the reliability of this year's prediction. Its projections have been way off the mark ever since the monsoon prediction model developed by a team of scientists led by the then secretary of the Department of Science and Technology, V R Gowariker, lost its relevance after 1993.

This is evident in the numbers presented in the graph:

Indeed, a reliable prediction of rainfall is vital not only for agriculture but also for various other purposes, including irrigation, reservoir management, flood control, navigation, power generation and the like.

What is actually needed is not an idea of the total quantum of rainfall in the whole country during the four-month monsoon season (June-September) but also its distribution in terms of space and time.

The advance knowledge about excessive rainfall and heavy downpours is as necessary as it is to know about long breaks in the monsoon activity. IMD, unfortunately, has not been able to meet such demands which are important to different stakeholders.

Though it has been issuing forecasts for the country's four broad regions (north-west, north-east, central India and peninsula) as well as for the agriculturally crucial month of July since 2003, these predictions, too, have yet to establish their credibility.

In 2009, for instance, IMD's forecasts for three of the four regions proved erroneous. For the country's key agricultural belt in the north-west, the projected rainfall was 81 per cent (with a stipulated model error as large as ±8 per cent) but the actual rainfall turned out to be only 64 per cent, far beyond the error range.

Similarly, in the north-east, the actual rainfall was 73 per cent, against the projection of 92 per cent; and in central India, the rainfall was 80 per cent, against the prediction of 99 per cent. Only the forecast for the peninsula held true with the actual rainfall being 96 per cent, against the projection of 93 per cent.

Indeed, the truth is that IMD's monsoon prediction capability has not improved in several decades despite gradual introduction of new technologies, such as Doppler Weather Radars, meteorological satellites and high-speed data communication and computing systems. Substantial additional investments went into augmenting the data collection, communication and processing of infrastructure, especially after the 2002 drought.

The 2004 drought prompted the government to further revamp and strengthen the forecasting capabilities of IMD through measures like creation of a fairly dense network of satellite-based automatic rain gauge stations for online monitoring; augmentation of the network of upper air observations; boosting infrastructure of S-band Doppler radars for complete coverage of coastal areas; and installation of more C-band storm detection radars, besides, of course, introduction of better super-computing facilities for faster data processing.

The only brief period when IMD generated reliable monsoon forecasts in successive years was between 1988 and 1993. This was the time when the 16-parameter power regression model (commonly called Gowariker model) performed well.

Subsequently, even this model began yielding wavering forecasts and ultimately had to be given up after its total failure to visualise the drought in 2002.

It was believed that some of the 16 regional and global parameters related to land, ocean and atmosphere had lost their relevance for the Indian monsoon and needed to be changed.

Consequently, IMD changed the models in 2003 and once again in 2007, using a mix of old and some new parameters, but without much success in moving closer to achieving a reasonable degree of perfection in long-range monsoon rainfall prediction.