Opinion - Leader Page Articles   

Monsoon forecasting

By N. Gopal Raj

The Indian monsoon remains one of the most maddeningly elusive phenomena, despite all the advances in modern technology.

AFTER LAST year's drought, the question uppermost in everyone's mind is whether this year would be a repeat of 1987, when a second year of poor monsoon rains ensued. These concerns appear to have increased in the wake of the India Meteorological Department's recent advance forecast for the coming monsoon. In reality, the IMD forecast puts only a 21 per cent probability on there being a drought this year too. A good deal of confusion has arisen as a result of changes the IMD has introduced in its monsoon forecasting methodology.

The Indian monsoon remains one of the most maddeningly elusive phenomena, despite all the advances in modern technology, from satellites to powerful supercomputers. Even a gross parameter such as All-India monsoon rainfall can be surprisingly difficult to predict. The IMD was not alone in failing to predict last year's drought. Most others, using both statistical methods and computer-based numerical simulations of atmospheric and ocean circulation, fared no better in their prediction before the monsoon began in June.

India was, in fact, the first country to start operational seasonal forecasting. The first forecast, issued in June 1886, was based on the inverse relationship between the snow depth in the Himalayas and the monsoon rainfall. In the early 1900s, Sir Gilbert Walker, the then Director-General of IMD, laid the foundations for statistical prediction of the monsoon, carefully computing the correlations between a large number of meteorological parameters and India's monsoon rains.

But the IMD's monsoon predictions were given only to the Government to aid its decision-making should an adverse monsoon appear likely. In 1988, however, the IMD went public with its monsoon forecast, based on a new power regression model which used 16 wind, temperature and pressure parameters from around the world. The IMD's monsoon forecast has become an annual media event since then.

The power regression model predicted the All-India monsoon rainfall. If it was within 10 per cent of the long-period average (about 88 cm, according to IMD data), the monsoon was considered normal. Below 10 per cent, as happened last year when it was 19 per cent below average, was considered a drought year and more than 10 per cent, a year of excess rainfall.

The rainfall was normal for all the years from 1989 to 2001. So, criticism of the power regression model could be ignored. In the past nine years, the actual rainfall was within the model's predicted error margin only in one year. But the most important criticism was that the model used too many parameters, some of which were cross-correlated, raising issues of its reliability.

The 16-parameter power regression model's failure to predict last year's drought sealed its fate. But the issue was what to replace it with. At a brainstorming session held at the Indian Institute of Science in November, it was held that efforts should be made to develop general circulation models (GCMs), simulating processes in the atmosphere and ocean, which would provide robust simulations of the monsoon and its variability. As GCMs the world over have difficulty simulating the Indian monsoon and its year-to-year variation, this was obviously going to be a longer-term research goal, rather than an immediate solution.

For its forecast this year, the IMD has thoroughly revamped its power regression model. The number of parameters has been drastically slashed. There are now two versions of the model. One uses just eight parameters. Since all these parameters would be available by March, the first forecast could be issued in April. Then, with two more parameters in June, a 10-parameter model is to provide an update by July. The idea for an update apparently arose after Government agencies found IMD's mid-season drought warning last year useful for contingency planning.

Any statistical model depends on the past relationship between predictor parameters and the predicted value (All-India rainfall) continuing. Unfortunately, such correlations can weaken over time. In 2000, the IMD replaced four parameters in its 16-parameter model. The revamped 10- parameter model will retain six currently used parameters and include four new parameters. According to the IMD, all these parameters are physically well related to the monsoon rainfall. Moreover, they have remained statistically stable over time.

In addition, this year, the IMD has introduced an 8-parameter probabilistic model. This model calculates probabilities for five rainfall categories. While the definitions of the drought and excess categories remain unchanged (less or more than 10 per cent of the long-period average), the `normal' category (within 10 per cent of the average) has been divided into three: Below Normal (90 to 97 per cent), Near Normal (98 to 102 per cent) and Above Normal (103 to 110 per cent). It is understood that this division became necessary because the probabilistic model required that every category have equal probability of occurring based on the historical rainfall data.

But this categorisation has also been the source of much confusion. With its 8-parameter power regression model, the IMD has predicted that the All-India rainfall this monsoon would be 96 per cent of the long-period average. That would make it a normal monsoon year. The probabilistic model put a 21 per cent chance of drought, 39 per cent for below normal rainfall, 14 per cent near normal rainfall, 23 per cent above normal rainfall and a three per cent chance of excess rainfall. If the probabilities for the three normal categories are added together, it can be seen that the probability of normal rainfall is actually 76 per cent. So, there is almost an 80 per cent chance of a normal or an excess monsoon, and only a 21 per cent chance of a monsoon failure.

The coming monsoon will also be a test of some new forecasting methods. An empirical method developed at the Indian Institute of Tropical Meteorology (IITM) in Pune, using only global sea surface temperatures, was one of the very few which had predicted a substantial rainfall deficit last year. There is, of course, the question of how robust this technique is. A. K. Sahai of IITM is unwilling to disclose what the model predicts for this year's monsoon, although the rumour is that it suggests above normal rains.

Peter J. Webster of the Georgia Institute of Technology in the United States and his student, Carlos Hoyos, say that they have a statistical model which can provide 20-25 day forecasts of the monsoon rains. Using this prediction scheme, the extent and duration of the 2002 drought would have been evident in June itself, claims Prof. Webster. The technique, which had been developed for the Ganges region, was useful even for State-level predictions, such as Orissa and Rajasthan, he told The Hindu. However, Prof. Webster and Mr. Hoyos are yet to publish details of their technique.

B. N. Goswami of the Indian Institute of Science believes that he has a method for predicting the dry and active periods which occur during a monsoon season. Frequent or prolonged breaks during the monsoon can lead to a drought, as happened last July. Long breaks during critical growth periods can greatly reduce agricultural yields, as appears to have happened to rice production in 1972, 1979 and 1987. Forecasting the dry and wet spells could be useful for agricultural planning and water management, according to Prof. Goswami.

Prof. Goswami has published research indicating that monsoon breaks can be predicted up to 17 days in advance, while active conditions, which are less easily predicted, might be forecast up to 10 days in advance. His method, however, requires satellite data-based rainfall estimates, which are the only way to get rainfall over the oceans around India. But these satellite estimates are available only after a considerable time lag. "We are examining whether other proxy parameters can be substituted'', he told The Hindu.

Printer friendly page  
Send this article to Friends by E-Mail

Opinion

News: Front Page | National | Southern States | Other States | International | Opinion | Business | Sport | Miscellaneous |
Advts: Classifieds | Employment | Obituary |

Stories in this Section

• SARS wars
• Paying for public safety
• Drowning out the truth
• Monsoon forecasting
• Only member-states can make the U.N. work
• Transforming Indo-U.S. ties
• Rising intolerance
• Working of MPLADS
• Saving heritage

Copyright © 2003, The Hindu. Republication or redissemination of the contents of this screen are expressly prohibited without the written consent of The Hindu