Why re-invent the wheel?

Considering the inherent uncertainties of the weather, meteorological predictions cannot be expected to hold true all the time. But if forecasts err more often than not, the want of prediction skills becomes apparent. That seems to be the case with the India Meteorological Department (IMD), whose prediction that the monsoon would set in over Kerala on May 24 has gone astray, for it eventually arrived on May 28. Whether the IMD's reckoning of the total monsoon rainfall this year to be 95 per cent of the long-period average also meets the same fate will be known only at the end of the monsoon season. The past record does not inspire much confidence in this prophecy.

The IMD has been in the business of meteorological foretelling since 1932, using chiefly empirical models, but its forecasts have been erratic. The exception was a brief six-year spell, between 1988 and 1993, when the then newly-conceived 16-parameter statistical power regression model gave correct predictions. But even this model began to waver subsequently and had to be dumped after it failed to warn of the 2002 drought. Several new statistical models have been devised and tried out since then but without success, indicating that the IMD's forecasting skills have not improved over the past 75 years. This is despite substantial investments in setting up data collection centres, Doppler weather radars, a communication network, meteorological satellites and high-speed computing systems. What also needs to be kept in mind is that this infrastructure has been considerably augmented in the last three years, when the unforeseen 13 per cent deficiency in monsoon rainfall and a prolonged dry spell in July caused a major agricultural crisis.

The IMD has sought to explain the failure of its successive prediction models with the plea that the correlation between different atmospheric factors and monsoon rainfall, assumed normally to remain constant, has actually been changing with time and slowly losing its significance. But what remains unexplained is why it took so long to realise this and why, even after recognising the nature of the problem, it continued to rely mainly on the statistical models that it knew would not hold. Indeed, most countries switched over to dynamic weather-forecasting models, deemed more reliable, in the 1990s. But the IMD has attempted only this year for the first time to put together a dynamic model-based weather prediction on an experimental basis.

The other new initiative taken by the IMD is to take into account the weather predictions made by national and international organisations while formulating its own forecasts. If that holds the key to the reliability of the projections, then why not rely directly on these agencies? In any case, the kind of forecasts the IMD issues is of little avail for agriculture planning and reservoir management, which essentially require location-specific information on timing and extent of rainfall, which the IMD is incapable of generating as yet. Indeed, if the IMD has, for some reasons, to generate a country-specific forecast, it would do well to seek the services of other institutions in evolving a suitable model that would work. These are the issues that the ministry of science and technology needs to ponder over before continuing to support the futile, yet unending, exercise of evolving an indigenous forecasting system through the "hit or miss" approach.