Unmasking the monsoon
But who wants to be foretold the weather? It is bad enough when it comes, without our having the misery of knowing about it beforehand.” This charming piece of dry British wit from the pen of Jerome K Jerome mocks the weather-obsessed Brit, and it may even amuse urban Indians, who every scorching summer, submit themselves to the tantalising tease of the monsoon rains.
But for the millions of poor Indian farmers, whose lives depend on not only generous but also timely rains, Jerome’s wit might appear a tad too blithe, even rude. The recent ill-timed hailstorms that decimated crops of thousands of unsuspecting farmers in north and west India underscore the absolute importance of foreknowledge. If truth about the monsoon could be foretold, the Indian farmer would be certainly better off.
So come April every year, India’s official weather tracker, aka the India Meteorological Department (IMD), goes into a tizzy trying to anticipate - with the help of reams of weather data, complex mathematical models and brute computing power, not to mention a bit of gambler’s gumption, fortified by a sense of history-whether the monsoon is going to be normal, generous, or miserly.
A seasonal or long-range forecast (LRF), as it is commonly called, only gives the odds for the quantum of total rainfall from June to September. Albeit of no practical use to an individual farmer’s sowing decisions, it is, nonetheless, a portent of hope, or possible gloom, for the national GDP and the agrarian economy.
IMD provides LRF for the Indian monsoon rainfall in two stages-first, in the third week of April, and then in the second week of June.
Along with these, IMD also issues monthly forecasts for July and August for the country as a whole. Conventionally, LRF is made using statistical models, which harness historical data to ascertain Indian monsoon’s relationship with a host of atmospheric and oceanic drivers, such as El Niño (a strong El Niño is supposed to be a strong harbinger of a weak monsoon), over different parts of the world in the three-four months prior to the forecast.
At present,the IMD is using an ensemble model, which makes predictions on the basis of monsoon’s relationship with a motley band of atmospheric and oceanic phenomena, called predictors, such as difference in the sea-surface temperatures (SST) between North Atlantic and North Pacific; Equatorial SST of South Indian Ocean; and Northwest Europe Land Surface Temperature.
The forecast gives probabilities for five different categories of outcomes-deficient, normal, below normal, above normal, and excess. For instance, for this year, there is a 66 percent probability that the rainfall will be deficient.
Along with the operational forecasts based on the empirical models, IMD also issues an experimental forecast using dynamic or numerical weather prediction (NWP) models. These are based on complex mathematical models of the atmosphere that simulate the physics of fluids in a rotating system (atmosphere being the fluid and earth the rotating system).
Currently, IMD is using state-of-the-art Climate Forecast System model developed by the US’ National Centers for Environmental Prediction (NCEP). But its forecast capacity is still weak and it is used only to supplement the operational statistical forecast. Indian scientists are trying to improve its predictive accuracy by simulating temperature-related oceanic phenomena such as EQUINOO (Equatorial Indian Ocean Oscillation) and IOD (Indian Ocean Dipole).
Apart from IMD, various domestic and foreign research institutions issue experimental LRF for the monsoon. Though they vary, IMD takes them into account before issuing the final operational forecast.
IMD also issues operational short and medium-range forecasts, using traditional synoptic models and poorly-developed NWP models, for use in public weather services, aviation, agriculture, hydrology, disaster management, etc.
But the trouble with using NWP models for tropical forecasting is that weather systems are more messy and unstable compared to, say, the mid-latitudes, where weather is largely driven by fairly stable and appreciable temperature gradient. In the tropics, the weather is mostly driven by the vicissitude of clouds, a feature that is not captured well by NWP models.
Broadly speaking, numerical forecasting is riddled with uncertainties that stem from inaccurate and insufficient observations of the initial conditions, deficiencies in the model itself, and poor understanding of atmospheric physics. Besides, the effect of factors that cannot be simulated by the model, such as forests, soil moisture and snow, also affect the model’s accuracy.
In general, the accuracy of NWP models in the higher latitudes is improving one day per decade. For the tropics, however, that rate is quite sluggish. All current operational NWP models have limitations in predicting aberrations in the monsoon, particularly extreme events like heavy rainfall.
Under the National Monsoon Mission, launched by the government after IMD failed to predict the 2009 drought, investments are being made to create an infrastructure that would eventually improve NWP models for the tropics. Specifically, observational network for surface, upper air and oceans, is being expanded to improve NWP capacity. Computing power, which is crucial for running models that simulate ever-finer features of the weather, such as cloud dynamics is also being improved.
The success or skill of a statistical or empirical model depends in large part on the interpreter’s ability to tease out the monsoon’s complex links with other large-scale weather phenomena by poring over large datasets. An intimate knowledge of monsoon’s biography, well-organised data archives and a sharp eye for hitherto unknown predictors, contribute to the making of a reliable seasonal forecast. “One recurring problem with statistical models,” explains Laxman Singh Rathore, chief of IMD, “is the evolving relationship between monsoon rainfall and its predictors. Over time, the links might weaken and eventually become irrelevant. This means one has to keep upgrading the models.”
So, what gives weather scientists the temerity to divine the monsoon rains 60 days into the future? On the face of it, it looks like a plain case of tilting at the weather vanes (with due apologies to Don Quixote). But, the whole, as the saying goes, is greater than the sum of its parts. According to Sulochana Gadgil, a veteran student of the Indian monsoon and honorary faculty at the Centre for Atmospheric and Oceanic Sciences (CAOS), Bengaluru, the overall deviation is only 10 percent of the long-term average of about 89 cm.
Conventionally, if the total rainfall is likely to fall short by 10 percent of the long-term average, we are looking at a potential drought; and if it exceeds by 10 percent, we, as a nation, can break into a peacock dance, even though it also raises the spectre of floods. Anything between the two gives us a normal monsoon.
Monsoon’s relative steadiness over long periods is probably because the phenomena that drive its variation, say, for instance, Pacific sea-surface temperatures or Eurasian snow-cover, change sluggishly over time. This, in addition to the perceived economic value of a bankable forecast, perhaps explains why IMD has persevered with the seasonal forecast ever since 1886, when H F Blanford, the imperial meteorological reporter and first chief of IMD, made the maiden long-range prediction in the world.
But IMD’s predictions have often been inaccurate. If one were to judge its seasonal forecasting record for 1993-2014, it made correct predictions in only five of the 21 years. If truth be told, this year is the first time since 1886 that the IMD has forecast a drought, and so far it’s turning out to be right.
“Given our poor record in the last decade, and given that last year was also a drought year, we were apprehensive about declaring this monsoon as deficient. But we decided to trust our new models, and our instinct, throwing political caution to the winds,” says Rathore.
The failure to predict a major drought in 2002 sent meteorologists scurrying back to their weather charts and models. Seeking to know why and how the monsoon changed so much, they discovered some disturbing aspects about its new persona: active and break spells across the summer months had become more frequent, intense, and inconsistent; it had become erratic in its arrivals and departures; there were more frequent extreme weather events such as intense downpour in a single day; its itinerary over the season reflected new and stronger geographical preferences; and finally, there was a distinct thaw in its relationship with old associates such as El Niño or the Eurasian snow-cover, as it warmed up to new partners such as IOD.
The monsoon was always a little mercurial, but in the new century it has apparently become even more shadowy. It posed new riddles to the meteorologist’s attempt to capture it in sub-par models. However, to be fair to IMD, it is the only weather agency in the world that sticks its neck out year after year by making an operational long-range forecast, while everyone else plays it safe by issuing experimental forecasts. And it seems that IMD has been stumped by a very tricky, elusive, and unpredictable customer.
There are a few facts that might offer some perspective on IMD’s record at making long-range monsoon predictions. One, post-Independence, meteorology didn’t get the kind of attention or money it deserved, especially in the 1950s and 1960s when scientists in the US and Europe were making seminal advances. Two, we did not give much importance to the rapidly developing field of computers, without which NWP was a non-starter. India hopped on to the NWP bandwagon only in the late 1980s with the acquisition of advanced computers. Meteorologists in the West had long realised the limitations of the empirical method as a forecasting tool and hence gradually shifted their focus to refining NWP.
For an NWP model to simulate a phenomenon like the monsoon and then predict its behaviour, it is essential to have a good grasp of the underlying physics, plus dense and accurate data.
The question is: do we have them? Monsoon jigsaw
After tracking and mapping it for over a century, we know quite a lot about how monsoon varies between years and within a season. But the causes of this variation are far less understood. “It is necessary to understand the physics of the monsoon and the causes for its variability if we want to construct models for making reliable predictions,” explains J Srinivasan, atmospheric scientist at CAOS.
As if this was not intricate enough, global warming has added another piece to the great monsoon puzzle. Even as some people attribute the recent kinks in monsoon’s behaviour to global warming, science can’t say that with confidence yet. Several climate models predict that with rising temperatures, the monsoon is likely to become more capricious in the latter half of the century and extreme weather events will become more frequent. This is likely to make accurate forecast for any monsoon event extremely difficult.
Some scientists say that in addition to better physics, refining data sets and models could prove handy. Currently, scientists access 90 percent of their data from satellites, which is available in the public domain through the Global Teleconnections System of the World Meteorological Organisation (WMO). To improve the quality and density of weather data, the UK and India recently launched the Drivers of Variability programme, which will use the UK’s BAe-146 atmospheric research aircraft and ocean gliders, and Indian research ships to gather fresh data.
But competent software professionals are needed to accurately translate this data for use in the models. “The trouble is, we don’t pay them well. So the brighter ones go elsewhere”, explains S K Dash, professor emeritus at the Centre for Atmospheric Sciences, Indian Institute of Technology, Delhi.
So, assuming we understand the physics of the monsoon well, and have all the data that we need, including the necessary computing power to crunch it, can we say confidently that we can make reliable predictions across different scales? Theoretically, that was the ultimate dream of the pioneers of the reductionist approach to tame the weather. However, that is fated to remain a holy grail—in 1963, Edward Norton Lorenz, an American theoretical meteorologist, had demonstrated the improbability of predicting weather beyond seven days.
DOWN TO EARTH
(The views expressed are strictly personal.)