India Meteorological Department
This is a collection of articles archived for the excellence of their content. |
Forecast models
Monsoon may be below par: IMD
TIMES NEWS NETWORK The Times of India
There are several models that the India Meteorological Department (IMD) refers to. For 2014 they issued forecasts by two such models. The monsoon mission model predicted 96% of the long period average while the ESSO IMD seasonal forecast showed 88% of the LPA.
The second model was very close in its assessment in 2013, However, “these are only experimental models and we cannot use their data with any kind of certainty.”
IMD has been periodically updating its methodology for the complex task of predicting the monsoon. The department, however, continues to battle the impression that its accuracy drops markedly when rains fail (see box).
In keeping with the latest updates from international agencies, IMD said that chances of an El Nino occurring in 2014 summer were high. “Latest forecast from a majority of the models indicate a warming trend in sea surface temperatures over the equatorial Pacific reaching to El Nino level during the southwest monsoon season, with a probability of around 60%,” the IMD said.
2013 was an ENSO neutral year — that is, there neither an El Nino nor its mirror opposite, La Nina. The southwest monsoon arrived 15 days before its normal date in June and continued till much after its scheduled date of withdrawal. Against a prediction of 98%, the country recorded an above-normal 106% rainfall.
1990-2002: error margin of forecasts reducing
The Times of India, Apr 13 2016
IMD's error margin on rains reducing
Amit Bhattacharya
The India Meteorological Department has often faced public criticism for getting its monsoon predictions wrong despite the complexities involved in the process.Data, however, suggests IMD may be getting better at the exercise. The average error in IMD's monsoon forecasts in 2003-2015 has come down to 5.9%, from 7.9% in the previous 13-year period (1990-2002), according to the department's analysis.
An error of nearly 6% suggests that the difference between the forecast and actual rainfall is routinely beyond IMD's stated error margin of 5%. Hence, there's certainly room for improvement.
But given the high degree of difficulty in getting forecasts that are spread over four months correct within 6% of the prediction, the performance isn't too bad. Consider this: In the 27 years since IMD began making all India predictions in terms of percentages of normal rainfall, it has been way off the mark seven times -years when the difference between forecast and actual rainfall was 10% or more.
Its worst prediction came in the drought year of 2002, when the forecast-actual rainfall difference was a gaping 20%. The department again erred majorly in 2004, another drought year, when the forecast was off by 14%. These mistakes lead to a relook at the IMD's forecast method. In 2007, a new statistical approach, with eight variables being considered during a two-stage forecast system, was unveiled.
Although IMD was again off the mark in the very first year -having predicted below normal rains (93%) while the actual was above normal at 106% -the department has since got the “direction“ of the monsoon more or less correctly, even though the margins have been high in at least three of the eight years. The year 2015 was another milestone for IMD, when it predicted a drought for the first time and got it right.
Simultaneously , the Indian Institute of Tropical Meteorology has been fine-tuning a dynamical computer climate model, CFS, borrowed from the US in 2012.
Sources said the model's accuracy has been increasing and it could replace the statistical method in a few years.
