The much-awaited long-range monsoon forecast of the India Meteorological Department, the country’s official weather forecasting agency, is out. On April 16, Met department issued its first stage long-range forecast for this year’s southwest monsoon, predicting “normal rainfall” of 97% of the long period average, or LPA, with a model error of plus or minus 5%. The LPA is the weighted average of rainfall that India received in June-September from 1951 to 2000 and is pegged at 89 cm.

The latest forecast suggests maximum probability of normal rainfall, defined as 96%-104% of the LPA, and minimum probability of deficient rainfall, defined as less than 90% of the LPA, between June and September this year.

The forecast has brought cheer, and for good reason. The southwest monsoon is often described as India’s “real finance minister”, determining the performance not just of the agriculture sector but the entire economy.

“As per our forecast, there is 42% probability, the highest probability, of normal monsoon rainfall this year,” said DS Pai, the Met department’s head of climate prediction in Pune.

Skymet Weather, a private forecasting agency, has also predicted normal southwest monsoon with 100% rainfall of the LPA, and assured that there is no chance of drought in the country.

But amid the excitement about plentiful rainfall, meteorologists offer a word of caution. “An all-India average monsoon rainfall forecast of 97% can be deceptive as it masks rainfall variability in various meteorological subdivisions of the country,” said Akshay Deoras, an independent meteorologist. “All-India forecast of normal monsoon doesn’t mean all 36 subdivisions will receive normal rainfall.”

As per Deoras’ preliminary estimates, as reported in the Hindustan Times, the Konkan coast and most of Maharashtra’s Vidarbha region can expect normal rainfall, but there is a greater probability of below average rainfall in Marathwada and central Maharashtra, especially during July and August.

Rainfall distribution and variability in a “normal monsoon year” is a valid concern. “Last year, the IMD forecast normal rainfall of 96% of the LPA and the actual rainfall was 95%,” Avinash Tathe, a scientist at the Met department’s Regional Meteorological Centre in Nagpur, pointed out. “Take the case of Central India. While Gujarat, central and western Maharashtra, and western Madhya Pradesh received excess rainfall, there was deficient rainfall in Vidarbha and eastern Madhya Pradesh.”

Forecasting southwest monsoon

The India Meteorological Department issues two long-range forecasts for the southwest monsoon. The first stage forecast is issued in April and the second stage in early June. “The first stage monsoon forecast provides a broad national picture and has its own limitations due to the spring predictability barrier,” explained Pai. “But by the end of May or early June things become clearer and we issue the second stage of forecast covering the four homogenous regions of the country and provide monthly monsoon rainfall forecast for the months from June to September.”

The four homogenous regions are Central India, Peninsular India, Eastern India and Northwestern India.

Both first stage and second stage forecasts are prepared using a statistical model called the Statistical Ensemble Forecasting System. For the April forecast, the Met department uses five predictors that are based on data monitored up to March. Sea surface temperatures in the equatorial Pacific and the Indian Ocean Dipole are closely monitored as well. The Indian Ocean Dipole is the difference in sea surface temperature between western and eastern ends of the South Indian Ocean. If positive, the Indian Ocean Dipole favours a normal monsoon.

Other than the statistical model, the Met department has been using the dynamical global Climate Forecasting System for predicting the southwest monsoon. This model, developed under the Union ministry of earth sciences’ Monsoon Mission, has been in use since 2012 but is still at the testing stage. It is borrowed from the United States National Centers for Environmental Prediction.

The dynamical model has predicted normal rainfall of 99% of the LPA, with an error margin of plus or minus 5%, this monsoon.

Not exactly normal rainfall

An anaysis of the rainfall data of the past few years shows that several subdivisions received deficient rainfall even in normal monsoon years. For instance, the country received 95% rainfall of the LPA last year, but Vidarbha had 23% deficient rainfall. The East Madhya Pradesh subdivision reported 24% less rainfall than normal, East Uttar Pradesh 28% and West Uttar Pradesh 30%. The Haryana, Delhi and Chandigarh subdivision and the Punjab subdivision also had deficient rainfall. On the other hand, the subdivisions of West Rajasthan and Saurashtra and Kutch received excess rainfall.

Similarly, in the normal monsoon year of 2016, at least 10 of the 36 subdivisions reported deficient rainfall while four – West Rajasthan, East Rajasthan, Marathwada, and Konkan and Goa – had excess rainfall. In 2011, seven subdivisions had excess rainfall while three subdivisions in North Eastern India reported deficient rainfall.

“Clearly, a normal monsoon year does not mean normal rainfall for all subdivisions, which is why we need more focused long range monsoon rainfall forecasting at state-level or subdivision level,” said Deoras. “Such a forecast can help state governments and local authorities prepare contingency plans and implement mitigation efforts at the right time.”

For instance, the Met department had forecast normal monsoon rainfall of 96% of the LPA last year. As such, farmers in Marathwada and Vidarbha sowed their kharif crops. But a long dry spell of four-five weeks, which was not forecast, led to sowing failure and reduced crop productivity.

In the normal monsoon year of 2016, at least 10 of the 36 subdivisions had deficient rainfall. Maharashtra's Marathwada region faced a drought. Photo credit: Nidhi Jamwal

Difficulty of localised forecasting

While long range monsoon forecasting at state or subdivision level is desirable, it is also extremely challenging. “Forecasting monsoon rainfall at subdivision level is not easy,” said RR Kelkar, former director general of meteorology at the India Meteorological Department. “At national average, the error margin is plus or minus 5%. But if we go down to subdivision level, then the error margin will increase significantly. But having said that, it also needs to be noted that subdivision level forecast should be attempted and improved over a period of time.”

Pai, the Met department’s head of climate prediction, said that his colleagues were working to improve the long range monsoon forecast. “We can provide state or subdivision level forecast but there is no point issuing a forecast just for the sake of forecasting,” he said. “The present knowledge on this subject across the world is limited and there is limited skill to issue such a focused forecast. We must remember that forecasting monsoon is not a perfect science but a probability science as monsoon is affected by several global and local factors.”

Kelkar said the challenge in monsoon forecasting lay in the model used by the Met department. “The statistical model of the IMD has its limitations and cannot provide subdivision level forecast without increasing the error margin,” he said. “The dynamical model doesn’t have these limitations, but it is still at a testing phase and the IMD is trying to fix it.” Eventually, he added, the statistical model should be replaced by the dynamical model, which can provide more accurate monsoon forecast.

Private weather forecasters agree. “The monsoon forecasting models we use in India are borrowed from the West and are not meant for a tropical country like India with extreme weather conditions,” said GP Sharma, president of meteorology and climate change, Skymet Weather. “These western models have been improved upon to provide monsoon rainfall forecast, but have their own inherent limitations.”

He referred to error margins in the current model to make his point. “While issuing an all-India monsoon forecast, IMD’s model has an error margin of plus or minus 5%,” Sharma explained. “But in June, when the IMD issues second level forecast for the four regions, the error margin increases to 8%. This further increases to 9% when issuing monthly monsoon rainfall forecast.”

Incidentally, the plus or minus 5% error margin has not come down in many years. “It has remained the same since 1998,” said Kelkar. “This error margin at the national average level is too wide and the IMD plays safe using this wide error range. A 5% error means that eight out of 10 times IMD’s forecast would be correct. Such a forecast doesn’t even need a model.”

For example, the Met department has forecast normal rainfall of 97% of the LPA with an error margin of plus or minus 5%. This means if India gets rainfall between 92% and 102%, the Met department’s forecast would be correct. “But a rainfall of 92% of the LPA is not at all a happy or ‘normal situation’,” said Kelkar. He added that the error margin could be brought down but only through the dynamical model.

M Rajeevan, secretary, Union ministry of earth sciences, said the Met department will “very soon” shift completely to the dynamical model. “Since many states have demanded it, we are exploring the possibilities of developing suitable prediction systems for state-level forecasts,” he said. But he cautioned: “As we go down to smaller spatial scale, seasonal forecast skill will fall down appreciably. The whole premise of seasonal forecast is for a larger area, averaged over a longer time scale. We may not have useful skill for many states but we expect some reasonable skill for some states, especially over central and northwest India.”

Facing uncertain conditions 

It may be premature to celebrate the impending “normal monsoon” because there is uncertainty around the major factors that impact the southwest monsoon. Apart from the five parameters used by the Met department in its statistical model, two additional crucial factors are the El Niño Southern Oscillation and the Indian Ocean Dipole. The former is one of the most important climate phenomena that can change global atmospheric circulation, influencing temperatures and precipitation across the world. The El Niño Southern Oscillation can occur in three phases – El Niño, a warming of the ocean surface in central and eastern tropical Pacific Ocean; La Niña, a cooling of the ocean surface in the same area; Neutral, when sea surface temperatures settle close to the average. El Niño tends to suppress monsoon rainfall in India whereas La Niña is considered to have the opposite effect.

“Sea surface temperatures have started increasing in the equatorial Pacific Ocean and the heating is expected to increase in the coming months,” said Deoras. “This may have some impact on our southwest monsoon and can suppress rainfall in parts of Maharashtra.”

Kelkar agreed. “The problem with the current monsoon forecast by the IMD is that it hangs heavily on El Niño Southern Oscillation and Indian Ocean Dipole,” he said. “So when these are neutral, we need to find out what other factors will determine the behaviour of the monsoon.”

All eyes are now looking skywards, hoping for bountiful monsoon rainfall. As many as 404 districts in the country are already facing dry conditions, according to the The Times of India, with 140 of them being categorised as severely to extremely dry.

Nidhi Jamwal is an independent journalist in Mumbai.