Sunday, August 5, 2012

Monsoon Chasers - India waits for IMD's most expensive 2017 monsoon prediction model

One day in 1988, life changed for the India Meteorological Department (IMD), a hoary institution founded in 1875 by a British government growing increasingly panicky after three years of bad weather destroyed harvests in its most valuable colony.

A tropical cyclone in Calcutta (now Kolkata) in 1864, and two famines in 1866 and 1871 due to the failure of the monsoon were enough to persuade the British that a single organization was needed, one that would peer into the skies, across the seas and into the ground, and divine the likelihood of another meteorological disaster.

The IMD headquarters in New Delhi.(Pradeep Gaur/Mint)
The IMD headquarters in New Delhi.

Like many institutions founded in the early days of British rule in India, IMD was based in Kolkata. It was headed by Henry Francis Blanford, a paleontologist, weather watcher and member of the Royal Geographical Society. For the next 113 years, headed by men from Gilbert Walker, considered by many to be the “brightest meteorologist ever”, to S.K. Banerji, the first Indian boss of the organization, from British rule through independence, IMD would attempt to predict the future, for the ears of the government only. India Meteorological Department is charged with forecasting the annual monsoon. So why do its predictions go spectacularly wrong sometimes? Here’s a look.

That day in 1988, under then director general S.M. Kulshrestha, India’s weather office dropped its habitual veil of secrecy and went public with its prediction that the monsoon rains that year would be “excellent”. It was right. Rainfall was indeed above average that summer. Since then, the Met department, as it is popularly known across the country, has lost its reputation as a reliable soothsayer—if it ever had one.
Officials measure wind data at the IMD observatory in Pune.
Officials at the IMD observatory in Pune.

Even in 1988, IMD wasn’t exactly the embodiment of accuracy: for instance, it told the government that rains would be normal in 1986. They were 13% below normal. It said the loss would be made up in 1987, which ended up seeing India’s worst drought.
 
Still, IMD’s debut on the public stage set off a chain of events that changed the way the department approached the iffy science of weather forecasting and the equally recondite art of engaging with the public.
Since then, the weather office has weathered the phenomenon called El Nino, once a minor variable and now an entire equation unto itself; two reworkings and an abandonment of its trusted weather models; and lay and scientific criticism of its forecasts. It has also promised to abandon its century-old reliance on statistics-based forecasting for the emerging global trend of so-called dynamical forecasting.

In 2012, when every day seems to push India closer to a drought, despite the weather office’s insistence that the weak rainfall has not yet reached drought levels, it looks like IMD would’ve been better off not going public with its forecast in the first place, but in the late 1980s, the step might not have seemed so drastic. “It wasn’t the media spectacle that we see today,” said Ajit Tyagi, a former director general of IMD, referring to the department’s annual forecast and updates. “It never made it to the papers and was just an internal memo to the government.”

Predicting the unpredictable
Earthquake data being recorded at IMD’s seismological observatory.
IMD’s seismological observatory.
 
The overwhelming importance of the south-west June-September monsoon, which provides 70-80% of the country’s annual rainfall, to both Indian agriculture and the economy has made it the focus of the efforts of at least three generations of meteorologists—both Indian and British. While the relative contribution of agriculture to the Indian economy may have come down from well over 60% in colonial India to a little over 14% in 2011, it continues to be the main source of employment for more than 60% of India’s workforce.

While droughts no longer necessarily mean famines, a poorly distributed network of reservoirs, an inefficient system transporting agricultural produce from farm to market, and the lack of the right kind of seeds and fertilizers mean that poor rain often translates into higher food prices, and hits hard rural households, many of which make a living from farming. The monsoon, then, is a key variable in economic calculations.

The monsoon has always posed a strange conundrum: it is a sea breeze so regular that in three out of every five years of the 100-odd that meteorologists have maintained actual weather records, the total rainfall has remained within 10% of 88cm. Yet, even a seasoned forecaster aided by gluttonous supercomputers measuring windspeeds at various heights, surface temperatures, heat patterns in the Pacific as well as the Indian Ocean, can’t predict, even in late May, whether India will receive normal rains in June. For over a century, IMD’s rather thankless task has been to navigate a fine line between the certain and the unknowable.

An operator observes earthquake data at IMD’s seismological observatory in Pune.
An operator observes earthquake data at IMD’s seismological observatory in Pune.
 
Today’s forecasters at IMD are usually postgraduates in science who have to pass an examination administered by the Union Public Service Commission. Once admitted, they can choose to be a forecaster or an instrumentalist, specialists who manage weather stations.Those who choose to be forecasters follow in a long line of weather predictors.
 
Since the 1680s, European scientists such as Edmond Halley—after whom Halley’s Comet is named—have expounded on the patterns of the monsoon. However, in 1909, it took the genius of a Cambridge mathematician Gilbert Walker, later IMD director general, to draw a statistical relationship between the Indian monsoon and the snowfall in the Himalayas in May; the atmospheric pressure in South America between March and May, and Mauritius in May; and the rainfall in Zanzibar in April and May. Though he never lived to see his theory confirmed, Walker was the first to suggest that the Indian monsoon was the net result of a global, interconnected tussle between the atmosphere and the ocean.

Walker presciently noticed that a cyclical swing in atmospheric pressure between Darwin in Australia and Tahiti in the western Pacific Ocean had something to do—he couldn’t precisely say what—with the occurrence of droughts over India; the phenomenon became known as the Southern Oscillation.

The southern sea breeze
In the 1950s, Walker’s ideas were linked to another centuries-old anomaly, wherein every three-seven years the waters in the Pacific Ocean off the coast of Peru turned unusually warm, especially around Christmas, and disrupted fish catches, an irksome phenomenon that South American fishermen referred to as El Nino. “Among the great discoveries of meteorology was noting that El Nino and the Southern Oscillation (ENSO) are related, a coupling of the sea and sky, and to think that this was intimately connected to droughts over India, several thousands of kilometres away, was fantastic,” said Krishna Kumar, a senior scientist at the Indian Institute of Tropical Meteorology, Pune, who’s studied the ENSO phenomenon for decades.

A self-recording rain gauge being tested at an IMD observatory
A self-recording rain gauge being tested at an IMD observatory
 
The Indian who accidentally brought the Southern Oscillation and the El Nino indices to public attention and midwifed IMD’s exposure to its inclement future wasn’t a meteorologist. A cryogenic expert with the department of space, and then secretary in the department of science and technology, Vasant Gowarikar had an omnivorous appetite for science. He created a population prediction model and authored a comprehensive encyclopaedia on fertilizers. Gowarikar led the development of a weather model that crunched the magnificent unpredictability of the monsoon rainfall into 16 weather variables, including the North Atlantic Sea surface temperature and the El Nino ocean index, and in 1988 used this to make that first publicized forecast.

“When I became secretary in the department of science and technology, I found that 90% of the things were already there,” Gowarikar told Down To Earth magazine in a 1993 interview. “It was really a question of somebody who was not conditioned by the same mindset taking the lead.” Repeated attempts by Mint to reach the 79-year-old Gowarikar were unsuccessful. It was in the 1980s that the Southern Oscillation theory, which had fallen out of fashion with meteorologists, came back into vogue, with help from Gowarikar.

A scientific assistant marks water samples from various parts of India at the air pollution section of the Met department.
A scientific assistant marks water samples from various parts of India at the air pollution section of the Met department.
 
Walker may have invoked a handful of parameters, but Gowarikar took the lead at developing a forecast model with about 200 factors—some as strange as the surface temperature at De Bilt in Holland—all of which had been tested, sifted, abandoned, resurrected and statistically linked to the performance of the south-west monsoon. IMD in the course of time had shifted headquarters from Kolkata to Pune, then Shimla, and finally to Delhi in 1944.

Modernization
The weather office’s forecasting centre in the heart of Lutyens Delhi looks and is as futuristic as it should be. The visible part of this is a multitude of flatscreen monitors. The hidden part of it is the alliance IMD entered into with French weather agency Meteo France to harness trillions of bytes of supercomputing power to assimilate and predict data from across the world, to generate hourly and daily forecasts. Every day, the IMD website records where rainfall is likely across India, and, by evening, it’s possible to know exactly how much rainfall each district received during the day. “There’s been a great improvement in the way we’ve made data accessible to the general public,” said L.S. Rathore, current director general. “In fact, even Pakistan relies on our website for their monitoring and forecasting purposes.”

IMD has a network of around 6,000 rain gauges, of which it relies primarily on about 2,000. It has also invested substantially—no numbers are available—in transmitting the measurements of rainfall, windspeed, surface temperature and cloud patterns from India’s fringes to IMD’s district and state headquarters. “IMD was one of the first organizations to employ everything from telegrams to computerized switching networks, satellite transmission, etc.,” said S.C. Bhan, spokesperson for the agency in Delhi.

Previously, IMD forecasters would be buried in a flurry of charts detailing the exact windspeeds—down from the surface to the stratosphere—and rainfall data from the 2,000 or so stations that would gush in every four hours. “It was then a matter of calculated guesswork,” said A.K. Bhatnagar, a former deputy director general with IMD. “To prepare a usable forecast, you had to apply your experience.”That would mean knowing the local weather history and finding out if there were anomalies—such as a freak wind shear, or unexpected excess rain in a neighbouring district. In Met jargon, such forecasts were called “synoptic forecasts”, and, even today, IMD places a premium on synoptic forecasts for determining the likelihood of rain, storms and cyclones. However, the punctiliousness of its data gathering and the sturdiness of its website notwithstanding, several experts say IMD is yet to achieve the original objective with which the British founded it: forecasting a drought.

In 2007, IMD introduced a new statistical forecasting system called ensemble forecasting, using five predictors from April research and three more from May. In an approach that would befuddle even Wall Street financial wizards, combinations of all parameters are mathematically crunched. Using the first five parameters, it throws up 31 different models, and those which best tally rainfall during a common period in the past are selected. Each of them is given a further statistical weight, just as food articles and consumer goods are given different weights in computing inflation, which in turn generates a new hypothetical model. “That’s the one we finally use,” said Madhavan Rajeevan, a pioneer of the ensemble technique and a former head of IMD’s forecasting division. “This, too, couldn’t predict the 2009 drought, but it did indicate that rainfall would be dramatically less. That’s a victory in itself.” Still, IMD has plenty of critics. “It doesn’t make scientific sense to encapsulate all Indian rainfall into one number,” said a senior weather scientist, who didn’t want to be identified. “No other country reduces its rainfall to a number; most of them give regional indicators.”

Perhaps it’s criticism like this that’s resulted in a Rs. 400 crore programme called the monsoon mission. The aim is to develop a model that relies on actual weather conditions on any given day, rather than statistics, to simulate the output. “We need as much data as quickly as possible, more weather radars, automatic weather stations as well as massive amounts of compu ting power,” said Shailesh Nayak, secretary, ministry of earth sciences. “More than half of it is done.”Perhaps that model, which officials say will be running by 2017, can finally resolve the century-long quest to master the monsoon.

Source  - Live mint
Author - jacob.k@livemint.com
Photographs - Abhijit Bhatlekar/Mint

No comments: