Just as much of India braces for its hot summer, the Indian Meteorological Department projected that the total rainfall, later this year, would be 84.5 centimetres, which is 5% below the average over the last 37 years.
In a closely watched annual exercise, the department, however, said the prediction falls within the 10% error margin considered acceptable. This year too, the department used a two-stage forecast—Thursday’s forecast predicting the level of rainfall between June and September, and the next in June, which would update and forecast how the monsoon would pan out. Usually, the department gives a probabilistic estimate of whether the monsoon would be ‘normal’ or not. However, it refrained from that this year.
“It’s always better to put a number on the rainfall,” said P.S. Goel, secretary, ministry of earth sciences. “As I am doing that, it doesn’t matter what normal is.”
In the last four years, anything between 98% and 102% was considered “near normal”, as per the department’s definition, and 93% and 98% constituted “below normal”. By those standards, this year is “below normal”.
The monsoon is critical to the nation’s economy. Just about one-third of India’s crops are grown on irrigated land and a poor monsoon can have a devastating impact on agriculture and also sharply hurt demand for farm equipment and consumer goods.
This annual prediction dictates sowing patterns across the country and helps decide what is the limit to which sowing can be delayed.
For this forecast, the department used a technique called ensemble forecasting, which is slowly gaining favour among international meteorologists and statisticians. Ensemble forecasting essentially involves taking averages, culled from results generated by a host of models.
This time around, the Department has chosen five geographical parameters that are strongly correlated to the rainfall, based on their performance in the last two decades. These are the North Atlantic Sea Surface Temperature, Equatorial South-East Indian Ocean Sea Surface Temperature, East Asia Mean Sea Level Pressure, North-West Europe Land Surface Air Temperatures, Equatorial Pacific Warm Water Volume. These parameters are permuted, yielding a plethora of models. Results from each are averaged out and the final number is what is given as the April forecast.
In the last four years, the department based its forecast purely on a statistical model called the eight parameter model, which has been accurate only half the time. “The eight parameter model was not very satisfactory,” said Madhavan Rajeevan, director, National Climate Centre, Pune.