ADVERTISEMENT
IMD's seasonal successes and forecast failuresThe India Meteorological Department’s predictions have improved manifold in recent years. However, climate change and monsoon variability pose new challenges.
Madhavan Nair Rajeevan
Last Updated IST
<div class="paragraphs"><p>A stallholder sits in the shade, waiting for customers on a hot summer day on the outskirts of Ahmedabad.</p></div>

A stallholder sits in the shade, waiting for customers on a hot summer day on the outskirts of Ahmedabad.

Credit: Reuters File Photo 

The vagaries of the Indian southwest monsoon rainfall (ISMR) have a profound socio-economic impact on agriculture, water resources, energy supply and, above all, the economy. The monsoon is crucial for agriculture and any major deviation in its performance is a matter of public concern. The ISMR exhibits a wide range of fluctuations on daily, sub-seasonal, interannual and decadal time scales. Therefore, it is important to have reliable systems to predict the variability at all these time scales.

ADVERTISEMENT

After the great famine of 1877, the then British government began systematic efforts to make long-term predictions of seasonal rainfall. In the 1920s, Sir Gilbert Walker, the Director General of the India Meteorological Department (IMD), achieved a breakthrough in monsoon forecasting by developing objective statistical models. Since then, IMD's operational system of long-range forecasting has changed in its approach and scope from time to time. Although there have been many forecast failures, the imperatives of economic administration and public demand have compelled IMD scientists to continue with the seasonal forecasts. 

Forecast failures in 2002 and 2004 prompted the IMD to introduce a new Statistical Ensemble Prediction System (SEPS) in 2007 to support the IMD's two-stage forecasting system. The SEPS showed an impressive performance compared to the previous forecasting system that was in operation during the period 1988-2006. 

People wait for a traffic signal to open at a junction covered with tarpaulin to protect commuters from heat on a hot summer day in Bhubaneswar.

Credit: Reuters Photo 

The forecast error during its operational forecast period (2008-2023) was 7.6%, which is lower than the standard deviation of the ISMR (10%). The IMD forecast error was 5% or less in 9 out of 16 years during this time. Interestingly, SEPS predicted the correct sign for the major deficient monsoon years like 2009, 2014 and 2015 with small errors in 2014 and 2015.

Statistical models have inherent limitations to address non-linear feedbacks and interactions in the coupled climate system that ultimately determine monsoon performance. As part of the Monsoon Mission, a prediction system based on a coupled climate model was made operational in 2017.

Later in April 2021, IMD introduced a multi-model ensemble forecasting system to generate probabilistic forecasts of the spatial distribution of monsoon rainfall over India.

Due to systematic and consistent research efforts, our understanding of monsoon variability has improved manifold in recent years, resulting in better forecasting systems. However, the skill of current seasonal forecast models is still below the maximum achievable skill. These models have systematic errors that represent the interrelationships between the ISMR and the oceans, especially the Indian Ocean. There is much room for further improvement.

For daily operations by farmers, short to medium-range forecasts (1-10 days) and extended forecasts (up to 4 weeks) are more useful. During the monsoon season, we experience active periods with increased rainfall activity and break periods with subdued activity. This active-break cycle has a huge impact on agriculture, health, energy supply and water resource management. Therefore, a reliable system to predict these changes at least two to three weeks in advance was needed.

In 2017, the Ministry of Earth Sciences (MoES) introduced a skilful extended-range forecasting system using different coupled climate models. The verification showed that the forecasts for broader regions of the country are good for up to three weeks. The system could make reliable predictions for the onset and withdrawal of the monsoon (early or late) and the start/end of active and break periods, at least two weeks in advance. Some application tools have also been developed for agriculture, water management and the health sector to utilise the extended range forecasts.

During the monsoon season, weather systems such as low-pressure systems bring heavy rainfall, leading to widespread riverine, flash, and urban floods. Observations indicate that the frequency of heavy rainfall events leading to flooding has increased due to global warming and will continue to increase in the future climate. Accurate early warning of these weather systems is essential for a reliable flood warning system.

IMD uses state-of-the-art, high-resolution (12 km) numerical weather prediction models for short to medium-term forecasts (1-10 days). These daily forecasts for rainfall are reasonably accurate up to 3-4 days during the monsoon season. However, there are systematic errors, for example in the heavy precipitation forecasts. Early warning of extreme rainfall events is therefore a major challenge.

Children jump into a lake to beat the beat during summer season, in Chennai

Credit: PTI File Photo 

To this end, the IMD now uses an ensemble forecasting system in which a range of probable forecasts are produced instead of deterministic forecasts. Another index, namely the extreme forecast index, is also used extensively to determine the severity of the subsequent weather event. These two indices correctly indicated the high probability of extreme rainfall affecting northern India on July 9, 2023 almost 3 days in advance.

Impact of climate change

Climate change increases the uncertainties of weather and climate forecasts and thus adds a new dimension to existing problems. IMD has an extensive observation network with quality-controlled and digitally archived data since 1901. They have recently launched a National Framework on Climate Services to provide climate services focusing on agriculture, water, health and energy sectors. With the availability of IPCC climate change projections, we also now have a reliable climate change information system to help us adapt to and mitigate climate change.

Over the last 10-12 years, the MoES has been making systematic efforts to improve weather and climate forecasting in India, which has really changed public perception. We should further accelerate these efforts to improve the reliability of weather and climate services. We need to develop a strategy of ‘Ready, Set and Go’ where we have reliable forecasts from seasonal to extended to short-term and very short-term time scales.

To develop this strategy, we need a more comprehensive observation network, with inputs from crowdsourcing, social media, the public and a constellation of small satellite sensors. We need better coupled data assimilation tools to assimilate asynchronous and non-conventional data. More importantly, we need a modelling framework with much higher resolutions to handle extreme weather events better.

Extensive research to better understand the physical processes of weather extremes is needed to improve the accuracy of the models. Artificial intelligence can bring a paradigm shift in weather forecasting and gives hope that the accuracy of weather forecasts can be improved.

MoES is already planning to move to a 5 km resolution modelling framework and is investing in the development of research test beds across India.

We have one of the best weather and climate services in the world and we should continue to improve the quality of services. We hope that the government will continue to keep confidence in the capabilities of MoES institutions and provide necessary support. Since we are on the right track, let us ensure that we realise the dream of a weather-ready and climate-smart nation by 2047.

(Madhavan Nair Rajeevan is Vice Chancellor, Atria University, Bengaluru, and former Secretary, MoES)