Will IMD get it right this year?

Smart caFE

What are the stages of the operational long-range forecasts made by the India Meteorological Department (IMD) for the southwest monsoon?

The IMD brings out the operational long-range forecast for the southwest monsoon in two stages. While the first stage forecast on the total rainfall for the country as a whole is based on data available up to March and is made in April, the second stage forecast is made in June and is based on data available till May. The second forecast is more detailed as it also provides a forecast of monthly rainfall in July and August, as also the rainfall for the four major geographical regions of the country.

Later, two additional forecasts are made. One is in July when the IMD brings out yet another forecast of the rainfall for August to September, which indicates the rainfall in the second half of the season. And the second is in August when the forecast for nationwide rainfall for the month of September is issued.

This year, how does the first forecast made in April differ from the second that was made in June?

In the first long-range forecast for the southwest monsoon of 2011 made in April, the IMD estimated that the rainfall for the country as a whole would be normal or in the range of 96% to 104% of the long-period average in the current southwest monsoon. This means that rainfall would not be either deficient, that is less than 90% of the long-period average, or below normal, that is in the 90-96% range, or above normal, that is in the 104-110% range, or excess, which is above 110% of the long-period average. In more accurate quantitative terms, the IMD forecast that the rainfall during the southwest monsoon of 2011 will be 98% of the long-period average rainfall of 89 cm during the 1951-2000 period, with a model error of plus or minus 5%.

However, in the second forecast made in June, the IMD revised down its estimates and said that the southwest monsoon would most likely be below normal, that is between 90-96% of the long-period average. It then quantified the monsoon season rainfall for the country as a whole to be as many as 5 percentage points lower than the earlier estimate in April and estimated it to be only 93% of the long-period average of 89 cm, with a model error of plus or minus 4%.

Additionally, IMD also forecast that the rainfall in the month of July is likely to be 93% of the long-period average and would marginally improve to 94% of the long-period average in August. Both estimates have a model error of plus or minus 9%.

Estimating the probable rainfall across the four broad regions of the country in June, the IMD also forecast that the prospects of the southwest rainfall are the best in the northwest region of the country where it is likely to be 97% of the long-period average, followed by the northeast and central India, with each region getting 95% of the long-period average. The rainfall in the south peninsula was forecast to be the lowest, at 94% of the long-period average. The error level of the forecasts for each of the four regions was estimated to be plus or minus 8%.

How accurate are the forecasts of the IMD?

The forecast made by the IMD last year was pretty accurate. In April 2010, the IMD estimated that the rainfall in the southwest monsoon of 2010 will be 98% of the long-term average with an error probability of plus or minus 5%. Later, in June, it revised the estimates and forecast that the rainfall would be 102% of the average, with a probable error of plus or minus 5%. The actual rainfall for the year was 102% of the long-period average.

The forecast for the months of July and August was also correct. While it forecast the rainfall in July at 98% of the long-period average with an error of plus or minus 9%, the actual rainfall was 103%. And in the case of the August rainfall, while the IMD predicted the rainfall for the month to be 101% of the long-period average with the same level of error as in June, the actual rainfall was 106%.

However, the forecasts for different regions sometimes went wide off the mark. While the actual rainfall in north west India and central India was within the range forecast, it went wide off the mark in the case of both northeast India and the southern peninsula. Though the IMD forecast the rainfall in northeast India to be 103% with an error of plus or minus 8%, the actual rainfall was only 82% of the long-period average, which was much lower than the forecast. In sharp contrast, in the case of peninsular India, while the IMD forecast 102% rainfall with an error probability of plus or minus 8%, the actual rainfall was 118% of the long-term average, which was much higher than the forecast.

But the saving grace was that the forecasts made for the second half of the monsoon (August to September) in July and for the month of September in August were more accurate, as they fell well within the projected range.

What about the forecasts made in the earlier years?

When one investigates the rainfall estimates projected in the preceding 10 years between 2000 and 2009, it is seen that the rainfall forecast was below the actual rainfall in 6 years and higher than the forecast in 4 years. The highest upside error was made in 2002, a drought year. While the IMD forecast that the southwest monsoon would be 101% of the long-period average in 2002, the actual rainfall was only 81% of the average, implying an error of 20 percentage points. And the biggest downside error was made in 2007. While the IMD forecast that the southwest rainfall would be 93% of the long-period average, the actual rainfall was 105% of the average, an error of 12 percentage points.

p.raghavan@expressindia.com