Analyzing air temperatures and relative humidity in some key areas rather than aggregate rainfall may better and predict the arrival and departure of the monsoon earlier

Monsoon clouds over Chennai. (Image by Kannan Muthuraman)

Monsoon clouds over Chennai. (Image by Kannan Muthuraman)

A team of researchers led by the Potsdam Institute for Climate Impact Research in Germany has identified two regions – the Eastern Ghats and North Pakistan – which they say serve as “tipping elements” in more accurately predicting the arrival and departure of the southwest monsoon.

Unlike current forecasting methods that rely on analysis of rainfall, theirs uses air temperatures and relative humidity, which are well represented both in theoretical models and observations. The proposed method, the researchers say, can predict monsoon onset two weeks earlier and the withdrawal as many as 1.5 months earlier than prevailing methods.

It can also predict the monsoon path in anomalous years, often associated with the El Niño disturbance, leading to a shortfall in rainfall that has impacted India in the last two years. Changing rainfall patterns in recent years have been partly blamed on climate change. In 73% of the years examined, it gives an accurate prediction of the onset with a range of seven days from the actual date. For the withdrawal, it gives a correct prediction in 84% of the years, with a range of 10 days.

“We are planning to test the performance of our method for monsoon timing of this year,” Veronica Stolbova, of Potsdam Institute and Zurich University and lead author of the study to be published in the journal Geophysical Research Letters, told

“We are currently waiting for the data from the National Centers of Environmental Prediction of the US National Weather Service and National Center for Atmospheric Research to be available to test our method for the monsoon of 2016, which should be accessible after May 5,” she said. “We are very curious to see the results for this year. We’ll also discuss with the India Meteorological Department (IMD) how our approach can help further improve their monsoon forecasting.”

The summer monsoon contributes around 80% of India’s annual rainfall. In spite of services and industry contributing the most to the gross domestic product, agriculture still employs almost half of the working population of the country, according to the 2011 Census.

The accurate prediction of the monsoon is a matter of life or death both in terms of employment and farm output, as the current drought underlines. Even the stock market reacted positively when the IMD predicted a higher than normal monsoon this year.

Existing models, such as that developed by the IMD, can accurately predict monsoon patterns only up to four days before the onset or four to 10 days prior to it. The IMD uses a numerical prediction model and most models – including that by the private agency Skymet – are based on such statistical methods, depending on a number of parameters.

Last year, two Indian Institute of Technology Bombay researchers and others pointed to the need to reassess the dates for planting crops, which have hitherto been based on an analysis of long-term rainfall data, without considering the impact of moisture-laden clouds over the Arabian Sea or trends in onset.

The current study treats the Indian monsoon as part of the global monsoon system which itself is one of the tipping elements of the earth. A relatively small disturbance can change the earth’s system and the onset of the monsoon is one of its characteristic features. It is an “abrupt” event and its timing varies from year to year.

The study identifies tipping elements as geographical regions, which demonstrate the biggest fluctuations prior to the onset of the rains. It uses these locations as observation points for collecting long term meteorological data. Analysed over time, the researchers have established a causal relationship between these and the onset or withdrawal of the monsoon, which allows them to predict both events.

“We find that there is an important relation between the two tipping elements of the monsoon: the Eastern Ghats and North Pakistan,” the researchers say. “We found that in North Pakistan and the Eastern Ghats, a mountain range close to the Indian Ocean, changes of temperatures and humidity mark a critical transition to monsoon,” explains Stolbova.

The researchers argue that these tipping elements are an appealing concept because it highlights the interaction of two phenomena. The first is a local weather phenomenon (differential heating between sea and land during the monsoon), which results in the occurrence of the Eastern Ghats tipping element due to the collision of the two branches of monsoon (Arabian Sea and Bay of Bengal).

The second is a global phenomenon, which establishes North Pakistan as the tipping element for the northernmost geographical boundary of the monsoon. It also sheds light on the mechanism underlying drastic change like an abrupt monsoon onset.

This atmospheric feature causes an abrupt onset of the monsoon over large parts of the Indian subcontinent north of the Eastern Ghats. Therefore, these two regions become connected exactly at the monsoon onset date in the ghats. The situation is mimicked during the withdrawal date. “We use this observation of the regional connection in response to monsoon onset and exploit it for a forecasting scheme,” the researchers say.

Monsoon clouds over the Western Ghats. (Image by B. Blalji)

Monsoon clouds over the Western Ghats. (Image by B. Blalji)

Raghu Murtgudde of the University of Maryland agreed with the authors that precipitation (rainfall) was a hard quantity to model and temperature and humidity are better captured. “But at the end of the day,” he told, “rain still has to arrive at the forecasted date and withdrawal has to be accurate also. So it is a bit contradictory to say we have better predictions based on quantities that are better modelled for a quantity that is not modelled well,” he said. “With dynamic forecasting, seasonal totals and spatial distributions will continue to improve till we achieve theoretical predictability. This paper does offer some guidance on where to improve observations but it is a long way to go if having that data will in fact improve dynamic models also.”

“If not, then relying on statistical forecasting for onset and withdrawal may offer only small added value to what is being done at Indian Institute of Tropical Meteorology in Pune. So it is better to translate the tipping point concept they define here to what it actually means for the precipitation processes, which brings us right back to having to do the rainfall right in the models anyway,” Murtgudde said.

“My main point would be that this can be considered as an additional product to the dynamical forecasting but nobody believes the onset date can be predicted months in advance. So more accurate predictions closer to the onset with a lead of about a month is more critical and the same holds for the monsoon withdrawal.”

“The best use of these results would be to understand why these tipping points are able to better capture the onset and withdrawal and how that can be translated into dynamical understanding.”

According to Auroop R. Ganguly of Northeastern University in the US, the paper is certainly an interesting advance in what some experts have been calling physics-guided data sciences. “Innovative algorithms from somewhat different methodological areas have been cleverly adapted for the problem at hand with an attempt to both inform the methods by known physics and interpret the results physically as much as possible,” he told

“However, ultimately it’s a prediction approach and the Indian monsoon is a hard problem. Many such prediction successes have failed when used for the future. As the physicist Neils Bohr is said to have stated: ‘Prediction is difficult especially if it is of the future.’”

“While the authors attempt some validation, more work needs to be done on that front before the results are convincing,” Ganguly said. “However, it may be useful to try this and other approaches for real predictions in the future and evaluating the performance over a few years.”

“Monsoon forecasting is of substantial interest for millions of Indian farmers, which is why the IMD does it,” Stolbova said. “Our approach can advance current prediction methods.”

“We did test it with past weather data and found that it yields correct predictions for onset in more than 70% and for withdrawal in more than 80% of the considered years. Hundred percent would be hardly achievable, given the complexity of the climate system, yet in comparison to existing methods our results show good accuracy.”

“The main advantage of the proposed approach is that it allows us to improve the time horizon of the prediction compared to the methods currently used in India. In addition, the new scheme notably improves the forecasting of monsoon timing during years affected by the global weather phenomenon El Niño Southern Oscillation (ENSO), particularly in its La Niña (withdrawal) phase.”

“This phenomenon significantly alters monsoon timing and decreases the prediction accuracy in existing methods, so we’re quite glad to see that our forecasting approach can be of help for farmers especially in these otherwise difficult years.”

“We agree, indeed, prediction in general and prediction of monsoon onset is a very challenging problem. However, one cannot succeed in the prediction until she/he tries.”

This year, the IMD has predicted a better-than-normal monsoon: 106% of the long period average of 89 cm of rainfall in the four-month period between June and September, measured between 1951 and 2001. See: Good rainfall forecast cheers parched India.

Two private forecasters, Skymet and Weather Risk Management Services, have followed suit. Before IMD, Skymet has put its forecast at 105% of the average a fortnight, which is exceedingly close.

Weather Risk predicts a normal monsoon, based on data from the US National Oceanic and Atmospheric Administration. Skymet erroneously predicted that rainfall would be 102% of the average in 2015 and later downgraded the figure to 98%.

Predictions from these private agencies are subscribed to by the aviation, energy, insurance, banking, chemicals and agriculture sectors, trading companies as well as dealers in commodities and policymakers.

Some of Skymet’s clients include the World Bank, BayerCropScience, Indian Oil and Monsanto. Weather Risk provides information to Syngenta, Pepsico and Mother Dairy.

Share This