If the good folks at Mausam Bhawan who just gave us yet another rosy-goosy forecast of a normal monsoon in 2011 were to predict the winner of 2011 Cricket World Cup, they would surely have picked Sri Lanka. Most of their five parameters would have favoured the islanders: batting first, not playing at home, century in the first innings, record in the World Cup against India, average score of teams chasing at the Wankhede. Yet we would rightly not have attached much credence to such a prophecy because these historical parameters bear no cause-and-effect relationship to the outcome of the current match and have little predictive value.
That pretty much sums up the case of the monsoon forecasts, shorn of the jargon, as well. Five parameters (sea and air temperatures, pressure and warm water volumes) are used. But scientists such as the pioneer Dr PR Pisharoty have said that the current weather is the result of factors no more than three weeks old. We cannot, therefore, predict the weather, say, beyond the middle of May with any certainty of causation. What we have then is based on a plethora of associated conditions, not unlike reading tea leaves, only slightly more sophisticated.
This annual rite of (delayed) spring of little logical rigour and less utility has been with us for two decades. Thanks to the India Meteorological Department (IMD) spin (its current press release exhorts the media to give it wide coverage), it is even taken with a measure of entirely undeserved seriousness.
I have grown long in tooth pointing out the futility of season-long forecasts of rainfall over a continental mass such as India. Gresham?s law applies: bad logic drives out good sense! Not only does IMD persist in its quest of this fool?s gold, but also manages to get ever higher allocations from a gullible government.
Logical inconsistencies, inaccurate prognoses and a complete absence of utility are the three main reasons for consigning this elaborate attempt to the dustbin. Averaging the precipitation over a large land mass and a three-month period is conceptually flawed. The trade-off and compensation that averaging implies does not apply to the monsoon. Rain-deficit areas cannot receive excess water from surplus areas for the most part, nor can excess rain in August make up for the shortfall, say, in the sowing period of July. Remember July 26, 2005? Mumbai received 945 mm, or half of its annual precipitation, in 24 hours. The monsoon would have been above average even if most of August and September had been dry! High rainfall in coastal Konkan and Orissa makes central Indian precipitation look normal, even when large tracts in the interior receive scanty rains.
IMD?s forecasts usually range around 96-104% of the long-term average. Given the size of the country and a margin of error of +/-5%, the actual figure would be within the predicted range in most years and IMD would pat itself on the back. Yet IMD?s most abject failure is in chronicling crises, as in 2009. Starting with the onset of the monsoon and continuing through the season, every single forecast went awry and by a large margin at that. I wrote a series of columns pointing this out from July to September. I even found and wrote about an unexplained phenomenon called Madden-Julian Oscillation, which had, in effect, taken our monsoon to China. IMD imperviously put out one feel-good forecast after another. Only at the end did it admit that there was a 23% shortfall!
It failed again to predict a dry June 2010, but was prompt to say that copious rains later in the season made its forecast come good. The most grievous shortcoming was its complete inability to anticipate rain continuing well past September, especially in the peninsula. We are still feeling its aftershocks to vegetable and fruit prices.
Finally, the utility. Even city people know now that it is not just the quantum of rain but its geographic and time distribution that critically affects agricultural production and availability of water for other purposes. Farmers need to know whether and how much it will rain, say, in the next week or so, and planners need to know how much water will be stored in dams and aquifers. IMD now provides district-wise five-day forecasts, presumably to meet this requirement. I have tracked these for the regions I know. They are all right when things are normal, but they predict off-track conditions with a lag, such as heavy rain warnings issued after the deluge. So, even in this endeavour, IMD peddles pastcasts as forecasts! No wonder, farmers, ranging from the urbane Khushwant Singh around Delhi to the rustic dryland growers in Solapur, look at the satellite cloud picture?even IMD can?t mess it up!?and consult weather sites on the Internet.
IMD has provided a background note this year. It mentions a number of other institutions engaged in similar exercises. It does not, however, state what earthly sense a single (or slightly disaggregated) season long forecast makes in a country as large as India, nor the logic behind historical statistical analysis when there are no causal relations.
Bob Dylan didn?t need the weatherman to know which way the wind blows, but we need real weathermen, not IMD, to tell us how the monsoon will behave!
The author has taught at IIMA and helped set up IRMA