Meteorological predictions are not perfect. We never claim that. It is a science, we do our best: Sivananda Pai

Meteorological predictions are not perfect. We never claim that. It is a science, we do our best: Sivananda Pai

Head of climate prediction at IMD, D Sivananda Pai denies allegations that they did not predict the recent heavy rainfall over Mumbai, points out that it is not IMD but govt that issues public alerts, asserts that India’s seasonal prediction is better than many countries, and calls cloud seeding “useless”

 D Sivananda Pai, IMD, IMD climate prediction, Monsoon, Mumbai rains, Mumbai floods, Floods, India floods, Bengaluru floods, Urban floods, Idea Exchange, Indian Express
D Sivananda Pai, Head Climate Prediction, IMD, at the Idea Exchange in The Indian Express Pune office. (Express Photo/Arul Horizon)

Head of climate prediction at India Meteorological Department, D Sivananda Pai denies allegations that they did not predict the recent heavy rainfall over Mumbai, points out that it is not IMD but govt that issues public alerts, asserts that India’s seasonal prediction is better than many countries, and calls cloud seeding “useless”

Why D Sivananda Pai?

The heavy rain in Mumbai at the end of last month that brought the city to a complete standstill, and similar events in Chandigarh and Bengaluru sometime ago, have once again turned the spotlight on the accuracy of predictions made by the India Meteorological Department. As the head of climate prediction at the climate research division of the agency, D Sivananda Pai occupies one of the hottest seats during the monsoon season. He is responsible for devising seasonal monsoon forecasts, that are known to become the subject of intense debate every year. Pai insists that despite limitations, forecasts have improved in the past decade

ALIFIYA KHAN: The Brihanmumbai Municipal Corporation (BMC) said it did not receive any specific warning about heavy rains over Mumbai on August 29.

The India Meteorological Department’s (IMD) Delhi office had issued ‘heavy rainfall’ warning for certain regions, including the Konkan-Goa belt and Madhya Pradesh-Maharashtra, two days prior to August 28-29. Weather prediction models were indicating heavy rain during these two days, almost a week in advance. Extended range predictions too had indicated this phenomenon a fortnight in advance.


IMD cannot forecast an exact rainfall value in millimetres. Instead, there is a prediction of either ‘heavy rainfall’, ‘very heavy rainfall’ or ‘extremely heavy rainfall’. In case of Mumbai, the regional variation was so high that within the city itself, Dharavi recorded 56 cm of rainfall, Santacruz 33 cm and Colaba about 10 cm. Rainfall is never equally distributed throughout the city. So, generally, forecasting the exact rainfall value over a particular location is not possible.

IMD issues warnings in which heavy rainfall is categorised as rain between 6 and 12 cm, very heavy rainfall between 12 and 20 cm, while over 20 cm rainfall is categorised as extremely heavy rainfall. In this case, our forecast predicted ‘isolated extremely heavy rainfall’ over the Mumbai region on August 29. The Regional Meteorological Centre (in Mumbai) shared this information with the local administration. It is important to understand that IMD itself does not issue any public alerts. That is not our mandate. We only issue forecasts. It is for other government authorities to act on those forecasts.

 D Sivananda Pai, IMD, IMD climate prediction, Monsoon, Mumbai rains, Mumbai floods, Floods, India floods, Bengaluru floods, Urban floods, Idea Exchange, Indian Express

ANURADHA MASCARENHAS: But then, contrary to predictions, Mumbai hardly received any rainfall on August 30 and 31.

The low-pressure system, which caused heavy rain in the region, was about 150 km in diameter and was placed over north Mumbai. When heavy rainfall warning is issued for a large region, the peripheral areas are also included. Like, when we issue a cyclone warning, we do not just mention the districts which will be actually affected, but also neighbouring districts, because we cannot afford to take chances.

Nature is not deterministic. It is a chaotic system. A high degree of randomness prevails and even the best of models cannot eliminate this uncertainty. Mumbai and its adjoining areas were put on alert on the days following August 29 because rainfall bands could have extended to these areas as well.

MANOJ MORE: But why are IMD forecasts so unreliable?

I admit that meteorological predictions are not perfect. We never claim that we are 100 per cent right. It is a science and we do the best we can.

MANOJ MORE: How are the forecasts in other countries so accurate?

There is a difference in weather events in other countries and India. England lies in an extra-tropical region. Their weather is dictated by frontal systems which are much more stable and predictable. India, on the other hand, is in the tropical belt. There is more randomness in weather events.

Look at seasonal prediction record of England, they stopped it a few years ago. I am not trying to run them down but seasonal prediction is much more difficult. Our seasonal predictions are more accurate than that of Britain and other places in the extra-tropics.

Also, in the US or Europe, their entire area is tracked by radars. It gives them better short-range forecasts. We don’t have radars that continuously monitor every region because it wasn’t considered very cost-effective. We need 30-40 more Doppler Radars. We are in the process of getting them. But our cyclone forecast is one of the best in the world. Even the World Meteorological Organisation (WMO) acknowledges that.

ANURADHA MASCARENHAS: The intensity of extreme weather events has increased. How has rainfall been affected?

In the past 60 years or so, we have seen an increase in heavy rains — 15 cm or above — over central India, in the Northeast, over the western coast, and in areas close to the Himalayas. Records are being broken. However, the total rainfall in the entire monsoon season has not changed. So, in between heavy rainfall there are long spells of dry days. You would have noticed it this year as well. Wherever the flooding happened, it was because of concentrated rainfall over two to three days.

AMITABH SINHA: We understand that meteorology is an inexact science. Besides, in the tropics, the uncertainties are higher. But given these uncertainties, do we have the best technology, best models and the best skills for weather prediction?

We now have a very good computer system, which will be further upgraded this December. We have very good weather models. There is much better coordination between different agencies now and we have the best talent in the field. Many of our IMD officials are in charge of technical committees at the WMO. I am myself chairing a technical team at the WMO which provides regional climate outlook. If you are not the best, you would not be picked up for these jobs. India is, and now it is globally acknowledged, at the same level as developed countries in weather forecasting.

That is not to say we do not commit mistakes. There are many other issues that apply to meteorology in general. Oceans comprise 70 per cent of the planet’s surface and our data-collecting capability over oceans is not very good. Climate models require data not just from the land but also from over the oceans. It is only in the last10-20 years that the IMD has been getting relatively better quality ocean data. Even on land, it is difficult to get data from certain areas such as Afghanistan, due to the tension in the region. We extrapolate to fill these data gaps, and that introduces inaccuracies. Then for calculations, we make approximations in numbers. Just as an example, we have to restrict the number of decimal points we use. That is because the computing power is limited. This is introducing another inaccuracy. Even if we assume that our data is completely accurate, the model is perfect, and we have an almost unlimited computing ability, two computers will give slightly different results. We might use the same computer to get the same result every time, but since nature does not behave in the same manner every time, there will still be some discrepancy between the model result and actual event. We need to be aware of these inherent limitations.

ANURADHA MASCARENHAS: How accurate have the IMD’s predictions been?

We started using a new model for seasonal predictions in 2007. If you compare the errors in our forecast in the 10 years since 2007 to the decade before that, you will find that it has reduced from 8 per cent on an average to 6 per cent.

Besides, the IMD had never predicted a deficient monsoon year right at the start of the season. We were able to do that in 2014 and 2015, even in 2009 to some extent. So, things have improved.

AMITABH SINHA: But then why does that perception, that the weather forecast in other countries is more accurate, persist?

I would say it is only partially correct. We give very accurate forecasts for cyclones. We are able to predict its path very precisely; five or seven days in advance. Other countries have more accurate forecasts over short-range, day-to-day variation, because of radar coverage and lesser uncertainties in weather systems. However, our extended range and seasonal forecasts are much better than them.

ALIFIYA KHAN: What can we do to improve our accuracy in capturing day-to-day variations?

It is already improving. We have good accuracy over two days, then it diminishes a bit. In other countries, they make good predictions for four to five days. Once we have more radars, our short-range forecasts will also improve.

ANJALI MARAR: Pune has been chosen for housing the Regional Climate Centre for South Asia. What is the progress on that front?

People in the WMO take India as the natural leader in meteorology in this part of the world. We have better capability than any other country. People say that we are even better than China, Korea and other countries. Our scientists are very well respected. When the WMO developed this concept of regional climate forecast, China took the initiative and started giving forecasts for the whole of Asia. Then people realised that there is large variation in climate over China, Japan, Korea on the one hand and monsoon over South Asia on the other. It was then decided to have a separate forum for South Asia. So India, being the natural leader, was asked not just to prepare forecasts for South Asia but also to train other countries in the region.

ANURADHA MASCARENHAS: How useful are cloud-seeding experiments?

I would personally say, and I know many other scientists agree, that cloud-seeding is not the right method or solution. It is not cost-effective. It is totally useless. It has to be understood that it is being carried out as an experiment, to learn more about clouds, not as a solution. It is not a good method.

AJAY KHAPE: Coming back to Mumbai, was it a case of a cloud burst?

Cloud burst has a specific definition. It is actually nothing but a very heavy rainfall event. When within one hour, more than 10 cm of rainfall occurs in a small area, it is called cloud burst. Generally, it happens in hilly regions. It is a very localised event. The heavy rain in Mumbai and the floods were not a case of cloud burst.

ALIFIYA KHAN: Do governments and local authorities take your advisories seriously? Does their inaction on your forecast hurt your reputation?

I think government organisations do take us very seriously and act as well. But there are other issues as well. Like in the case of Mumbai, there was 56 cm of rainfall in Dharavi, which is a low-lying area. Now look at Cherrapunji. It gets the highest rainfall in the country but the area is facing a drinking water crisis because the water just flows out. So when the forecast says that very heavy rainfall is likely, one does not know whether it will happen in Dharavi or some other area. But you can always argue that local authorities must be more pro-active in responding to such forecasts.

You also have to realise that if the authorities act and the rain doesn’t happen, then even that action can backfire. But the IMD is neither mandated nor trained to get into response systems.

Had this 56 cm of rainfall happened in the beginning of the monsoon season, nothing much would have happened. That is because the soil absorbs a lot of water. But, at a later stage in the monsoon season, like in the case of Mumbai, the moisture content of the soil is already saturated.

Engineers have to think of out-of-the-box techniques to take the water out of the city and use it in the future.

MANOJ MORE: Do you have systems for informing civic bodies and other local authorities when you foresee such major events? Do you call up local authorities and tell them to take action?

Our advisories and forecasts go to all relevant people in the government and local authorities. However, it is not the IMD’s job to tell them what to do. But there is regular interaction with government agencies. There is a standard operating procedure for all sorts of situations.

AMITABH SINHA: How difficult is it to forecast an event like the heavy rain in Mumbai recently? Is it possible to predict that there will be rainfall between 30 and 35 cm, instead of, say more than 20 cm?

When rainfall happens in a big area, it is not uniform. It drizzles in one place, and rains heavily at some other place. Since our observatory is not located at every place it is very difficult to exactly say how much rainfall would occur. Sometimes, it is possible that we entirely miss the rain. So that kind of accuracy, predicting that the rainfall will be between 30 and 35 cm at a particular location, is not possible. And this is true everywhere. I don’t think any weather agency makes that kind of prediction anywhere in the world. There are some inherent limitations of meteorology. This is not a perfect science.

AMITABH SINHA: How local can the predictions get?

The grid size we currently use is of 12 km. It is possible to get to a smaller area than that. Sometimes, the model itself shifts the rainfall region. You get an idea that there will be heavy rainfall in this region, but the model has errors. So it is not possible to give the value of rainfall likely to happen in a very small area.

PARTHA BISWAS: This year, the first Agromet Advisory came on June 16. On June 29, there were indications that rain might be less. Immediately, the IMD gave an advisory asking farmers to use sprinkler irrigation. But, by that time, all the sowing was done. In such a short span of time it is not possible for farmers to make the shift. Is there any mechanism to address such problems?

See, the IMD prepares a forecast for 10 days and an extended range forecast for two weeks. The two-week forecast is given to CRIDA (Central Research Institute for Dryland Agriculture) in Hyderabad. They are the experts in agriculture. Based on our forecast, they prepare the advisory. We also share our five-day forecast with agromet field units in every agriculture university. We give them five days’ forecast and they decide what has to be done. This is done twice a week. This has started in the past two-three years. Extended range forecast does not go to farmers, it goes to the planners. It is possible that someone in the field units just goes by the five-day forecast and does not refer to the extended forecast. Also, our agricultural practices have to change. When you get water, you need to use it judiciously.

AMITABH SINHA: How do you see the entry of private companies in weather forecasting?


There is space for everyone to get into forecasting. But sometimes packaging is confused with a better product. We are not into packaging, into building impressive websites. It is possible that they are reaching out to more people than we do. But they do not have any independent data-collection system. They do not have any models.

For latest coverage on Haryana and Maharashtra Elections, log on to We bring you the fastest assembly election 2019 updates from each constituency in both the states.