Improve weather models

IMD's failures raise questions about monsoon predictions

The India Meteorological Department (IMD) has not had a great track record in forecasting the monsoon in recent years. This year, however, its errors have almost been farcical. In its bulletin on May 30, it predicted the onset of the monsoon in Kerala on May 31; however, it then had to revise it to suggest the monsoon would in fact hit on June 3. The monsoon then spread rapidly across the south-east, including West Bengal — but is yet to arrive in Western Uttar Pradesh, Delhi, and points beyond. This is in spite of the fact that the IMD had originally predicted it would reach Delhi by June 15. This prediction came out on June 13 — meaning that it is hardly a long-term forecast that went wrong. Subsequently, the IMD has repeatedly predicted the arrival of the monsoon in Delhi, only to leave the city disappointed time and again. There is no reason to assume that a late onset means that the overall monsoon will be below par, but it is also true that some sectors, particularly agriculture, need to be warned of a delay or a hiatus in the monsoon rains in a timelier manner.

It is certainly true that predicting the weather is a complex process, and errors are to be expected. But a pattern of errors such as what has been observed over the last six weeks suggests deeper problems. Either predictions are being stated with excessive certainty, or the models need to be re-examined. The Standing Committee on Science of the Parliament recently pointed out that the IMD’s forecasting failures have become particularly problematic because of the need to protect against the increasing incidence of extreme weather events such as flash floods. Areas of Himachal Pradesh were inundated with flash floods after a cloudburst over this weekend. Given the need to manage extreme weather events will become only more urgent as climate change intensifies, the IMD needs to up its game or at least be more open about what its models predict.

Private weather forecasters, including Skymet, Earth Networks, and IBM Weather, have not necessarily done much better than the IMD in recent years, although Skymet’s model has arguably performed more effectively in predicting monsoon onset over past years. At least one state government, Kerala’s, has turned to private providers for “ensemble predictions” of the monsoon this year. The broader question is to what degree end-users, from governments to farmers to industry, can access the output of multiple well-informed models of the weather rather than relying on point forecasts from a single government agency. Legitimate private forecasters should be able to access a broad set of relevant data points, even if collected by state agencies, in order to make their predictions available to a wider audience. This has technically been the case in recent years, and some government agencies like the NITI Aayog have even turned to private weather modelling companies to partner in research into precision agriculture, for example. But the IMD expects that it will continue to be the nodal agency for predictions, while the private companies will simply be service providers or vendors. This is not the best position, given the IMD’s repeated failures. A broader set of models and modellers is essential if farcical situations like this year’s monsoon predictions are not to be repeated.