Here is some unsolicited advice for the science and technology minister. Please consider abandoning immediately the monsoon modelling and forecasting exercise. It has turned out to be an enormously wasteful exercise, year in and year out.
While you are at it, you may as well consider abolishing all weather forecasts, since what the meteorology department does qualifies to be pastcast more often than not.
First, the season-long monsoon forecast. It cannot simply be an exercise in scientific curiosity; it must serve some purpose. The only two possibly candidates are helping the farmers with their crop cultivation decisions and enabling the government to formulate an annual or a medium-term policy. Both these subsume that the prediction provided is somehow linked to actual production. This is not at all true.
The exercise tries to predict the total quantum of precipitation in the entire Indian land mass in the whole of the south-west monsoon, lasting three months or more. Lately, due to some criticism, the department has attempted a little disaggregation, with four regional and possibly month-wise forecasts. These are expected to be provided only after the onset of the monsoon.
It is not the total quantum of rainfall that determines the crop output, but its distribution. For example, a freak occurrence of the entire season's rainfall in a short period of a week and a complete dry spell thereafter would result in a near-total crop loss, even though the monsoon according to the model terminology would have been average!
The weather patterns of the last two years have amply shown that such skewed distribution does actually occur. On the other hand, as Mr K Rajan, former secretary of agriculture, Government of India, never fails to mention, the best crop achieved in the Khandesh region of Maharashtra was in 1978, when the rainfall was only 50 per cent of the average, but extremely well-distributed throughout the season.
Therefore, no administrator or farmer would ever set his decisions by the pronouncement of a near-normal monsoon, because it is precisely meaningless. At the same time, only a major calamity would produce precipitation that is outside the +/- 10 per cent of the long-term nation-wide average range used by the department to define normal monsoon.
In any given year, some regions would receive excess rain, to make up for the shortfall elsewhere. Given the size of the country, the overall annual variation would be very small. Therefore, without any model building, one would be safe in predicting a near-normal monsoon every year to eternity, global warming permitting, and expect to be wrong only once in a century or so!
What the farmer needs is a week by week, localised forecast of when and how much it would rain. This is a completely different exercise from the one the department is greatly enamoured of, requiring a different approach altogether.
The season-long forecast is not, contrary to popular belief, a cause-and-effect prediction. Prof. Pisharoti of the Physical Research Laboratory of Ahmedabad, the grand old man of this branch of science, explains that nature itself makes up its mind only three weeks in advance.
Therefore, the causative exercise is limited to this narrow period. All the others, regardless of how many parameters they use, are associative. Certain phenomena, such as the ocean surface temperature, are observed to have been behaving in an associative manner with the monsoon and are hence used as part of the modelling exercise.
This is akin to using anti-incumbency as an associative factor in election forecast. You would be right in using it, but the voter makes up his mind finally only when he presses the button on the voting machine and he does what he does because of various factors which we tend to associate with anti-incumbency.
The short-term forecasts, on the other hand, are based on actual observation, such as formation of depressions and movement of cloud masses. They could be very accurate and highly region-specific.
Our Met department, however, looks at its own satellite pictures and assorted evidence mostly after the fact. The Gujarat deluge of June 2005, the Mumbai downpour of July 26, 2005, and the very rare but continued movement of six or seven depressions from the Bay of Bengal westward last August and September with huge precipitation were all possible to be predicted, but the Met department issued alerts only after the main event.
Similarly, the sporadic rain over the entire north-western region this spring or the temperature spikes in summer and winter are noted by the forecasters only after they occur. Hence the term 'pastcast!'
The Accuweather Web site, which used Indian satellite photographs and tracked the movement of troughs and cloud masses last year, predicted the rainfall far more accurately.
After the initial few visits, even I could make reasonable forecasts at my desk; the contractor who carried out my water-proofing work used to call me unfailingly every morning to plan his work!
The work of the IMD, therefore, is certainly not what we need. This is so not any longer because it lacks the tools or equipment. What it lacks is focus and an attitude to be of use and service.
It must give up its obsession with season-long forecasts for the country as a whole, because however many models it uses or strives to achieve a dynamic balance, the plain and simple truth of the matter is that these long-term forecasts are of little value and merit in making farm or sector-level decisions. And this misplaced focus keeps it from making real contributions in area- and time-specific forecasts.
Are you listening, Mr Kapil Sibal?
More from rediff