Indian scientists are trying out a new climate model to get monsoon forecasts right. If they can get it to work, expect it to rain when they say it will
I t was going along quite swimmingly till 2002 for the weathermen of India. The Indian Meteorological Department (IMD) forecast a normal monsoon for that year. As the season wore on, it became clear that the monsoon would be anything but normal. It was 19 percent below normal and India experienced its first drought after 1987. This resulted in a 12 percent decline in food production. Weathermen couldn't explain why they had been unable to see it coming. "There was little understanding of what had happened. And different institutes had different explanations for the deficiency," says M. Rajeevan, scientist at the National Atmospheric Research Laboratory.
Given that 65 percent of Indian farmers depend on the monsoon for irrigation, a failure of this kind has serious economic and political consequences. The IMD went back and checked its crystal ball — the climate model, a mathematical formula that incorporates a large number of factors that cause the monsoon and predicts the total rainfall in a season. IMD's model took into account 16 such factors. This model was tweaked. Things improved in 2003 when IMD's forecast more or less matched the real thing. But in 2004, the forecast went wrong again. Another drought hit the country. Things have been up and down since then. After every year or two, the forecasts go awry. "When the monsoon fails, GDP falls by almost 2-3 percent. It has a huge cost on the agricultural sector. Even in 2009, agricultural production fell by about 7 percent from 234 million tonne in 2008 to 218 million," says Dev Raj Sikka, an experienced meteorologist.
Say Hello to My New Friend
The government of India has had enough and wants the weatherman to get his mojo back. In India, there are more than 10 organisations that study climate. One of them, the Pune-based Indian Institute of Tropical Meteorology (IITM), has picked up the gauntlet. "Few institutes want to do it because it involves a lot of investment — in money, training of human resource and time," says B.N. Goswami, Director, IITM. Goswami is 60 years old and stands maybe five foot four. He is light on his feet and scurries about the Pune campus to show off his Brahmastra. In a large building located 80 metre from his office sits the big IBM pSeries 575 supercomputer. It can do 18.8 billion floating point operations in a second. Goswami is now looking to hire more Ph.D.s who understand various aspects of climate and also programming talent to use the supercomputer to look into the future.
But all the manpower and the hardware will be of little use if Goswami can't improve the underlying logic of weather prediction. He is betting on a new method to predict the monsoon accurately. Scientists call it a 'physical' or 'dynamic' climate model. Since it is supposed to be more exact and rigorous, the ministry of Earth Sciences has agreed to foot the bill for the beast in IITM's backyard. If Goswami can prove it works, this is the model that IMD will use for its forecasts.
His enthusiasm is infectious but Goswami knows he is trying to do one of the three most difficult tasks known to mankind. The first task is predicting the stock markets. The second is anticipating the performance of the Indian cricket team. The key thing to understand is that the weather is a result of many different physical phenomena. Measuring them involves different levels of complexity — some are so difficult that it is mostly guesswork. For instance, evaporation as a function of wind speed and humidity, or fluctuations in the ocean. "There are many factors like that. So, what I am saying is that these are all physical processes which can affect our monsoons. Now, in order to predict, actually we have to model them. This is what you call a physical climate model. This model, in principle, is a model of the atmosphere which is nothing but the equations that will solve the motion of the air. Atmosphere is air, right! So ultimately, you have to calculate where there is more heat and less heat in the atmosphere," adds Goswami.
The New 'New Thing'
In the new method, Goswami's team will come with mathematical formulas that will capture the interaction of these different physical phenomena. The formulas will divide the world into tens of thousands of grids of 100 sq km each. Each year, the team will put in fresh values of almost 100 different factors for each of these 100-sq-km grids. Then the man 'o war, the IBM supercomputer, will crunch through millions of calculations to solve the mathematical formulas. When the last calculation is done, the answer left on the screen will be the amount of rainfall India will get in that year. OK, it might not be that simple, but you get the idea.
Now this is a radical departure from what meteorologists have been doing in India till now. The IMD, which is India's official weather forecasting agency, has for years relied on what is called a statistical model, which is developed by using actual, historical data of rainfall over the last 30-40 years. For instance, IMD developed a statistical model in 2007 based on the last 30 years of data to make the forecast for 2008.Based on the data, weights or coefficients are assigned to several parameters that go into the model and that is used to predict the next season's forecast. "Over the years the skill of these models in predicting the monsoon has gone down," adds Goswami.
This is where the new model will be different. It will factor in the data for the current year and it will take into account any change in the relationship between the 100 or so factors that it will use to predict the climate. So, say, historically the relationship between temperature between the Bay of Bengal and land is one of the determinants of monsoon. Now say the emergence of a tornado in that region changes that relationship — the new model will take this into account.