IMD’s heavy rainfall forecast accuracy improves to 70%

Published on Aug 21, 2020 12:07 AM IST
HT Image
HT Image
ByJayashree Nandi

New Delhi: The Indian weather office has become more accurate on all four measures it uses to audit the accuracy of its predictions, the top government officer overseeing it said, adding that it got seven of every 10 forecasts of heavy rainfall right between 2017 and 2019, up from under five between 2002 and 2016 -- a change weather scientists attribute to superior weather models, better quality of input data, and more computing power.

The so-called probability of detection (POD) used to forecast heavy rainfall has made a significant progress, said M Rajeevan, secretary, Union Ministry of Earth Sciences (MoES) at an Indian Meteorological Society lecture on Wednesday, going up from 45% in 2002-16 to 70% in 2017-19.

The National Oceanic and Atmospheric Administration describes POD as the number of accurate forecasts divided by the total number of events observed.

“We are using the global forecasting system (GFS) at 12km resolution which is a very good model. We have also started with a probabilistic forecast which tells us which forecast would have more probability. This has given us much better skill and confidence. We are also providing good input for the model by using data from satellite, radar and ground data. To run the model we need higher computing power which has also been made available in recent years. The combination of all three have helped IMD improve,” DS Pai, senior scientist, India Meteorological Department (IMD)

Similarly, the false alarm rate (FAR) has reduced from 42% to 32% over the corresponding period for heavy rain episodes during monsoon season. The four measures of forecast accuracy documented by IMD are false alarm rate, miss rate, probability of detection and critical success index (CSI).

Skymet Weather, a private weather forecasting company claimed it has 80% accuracy in forecasting heavy rain during monsoon. “We are around 80% accurate with heavy rain forecast. Our forecast system is also model based which is then improved on by our meteorologists. World over, forecasts are improving because data points are increasing. We have around 7000 automatic weather stations for surface data,” said Mahesh Palawat, vice president, climate change and meteorology at Skymet Weather.

HT couldn’t independently verify the claims of both IMD and Skymet.

Apart from an overall improvement in forecast reliability, IMD scientists can now make monsoon rainfall forecasts for five days instead of three days when IMD launched its monsoon mission (in 2012). The mission, which involved an expenditure of Rs 400 crore involved IMD acquiring the capability to make accurate short-, medium-, extended- and seasonal-monsoon forecasts, critical in a country where much of agriculture is still rain-fed.

“We can forecast with the same accuracy, but for five days now, as far as the short-range forecast for monsoon season is concerned,” Rajeevan said during the virtual lecture, which was attended by scientists from across the country.

The first phase of the monsoon mission was completed in 2017.

IMD and other teams from MoES are working on the second phase of monsoon mission , which is focused on extreme weather events and development of applications for various sectors such as renewable energy (solar and wind energy projects require accurate forecasts). agriculture and hydrology.

For example, Rajeevan said IMD has already developed a percentile-based forecast of extreme weather.

Similarly, IMD is coming out with 550 new automatic weather stations and state-of-the-art sensors for soil moisture and soil temperature detection. “These can help us understand landslides and help immensely with forecasts,” he added.

Importantly, the weather office has now started issuing forecasts that look at impact and outcome.

“I remember the 2013 Uttarakhand floods. We did forecast very heavy rain. But what does it mean for the common man? What action should he take when he has heavy or extremely heavy rain warning? The impact-based warning takes into account all these factors,” said Rajeevan.

The MoES secretary, however, conceded that though the 25 June lightning events in Bihar was predicted by IMD and Indian Institute of Tropical Meteorology (IITM), several people still died because of the natural disaster.

Independent experts maintained that more long-term data is needed to measure improvements in making accurate predictions.

“These three years are not sufficient enough to judge improvement in forecast skill. There is a method called hind-cast verification. Where you run the current model for earlier years and generate the skill score ...,” said a climate scientist who spoke on condition of anonymity.

Meanwhile, climate change is increasingly making a prediction of extreme rain and seasonal forecasts a difficult proposition.

In an article, co-authored by Rajeevan, published in Nature in 2015, titled: “Rethinking Indian monsoon rainfall prediction in the context of recent global warming,” the authors said: “Despite enormous progress having been made in predicting ISMR (Indian summer monsoon rainfall) since 1886, the operational forecasts during recent decades (1989–2012) have little skill. Here we show, with both dynamical and physical–empirical models, that this recent failure is largely due to the models’ inability to capture new predictability sources emerging during recent global warming.”

“Some extreme rain events pull up or down the average leading to deviation from the seasonal forecast. We have seen it happening,” the MoES secretary said.

“Undoubtedly, the overall weather forecasting accuracy in India has improved in recent years and a 70% POD of heavy rainfall events during 2017-19 is impressive considering challenges associated with the prediction of tropical rainfall. However, verifying forecasts for more years in the past (such as 10 years) is essential since it will add to the robustness of results and eliminate any statistical biases due to below average seasonal rainfall in 2017 and 2018. In addition, exploring other skill scores and regional biases in them will help in further improving forecasting capabilities,” said Akshay Deoras, independent meteorologist and PhD researcher at University of Reading, United Kingdom.

Get Latest India Newsalong with Latest Newsand Top Headlinesfrom India and around the world.
SHARE THIS ARTICLE ON
SHARE
Story Saved
×
Saved Articles
Following
My Reads
My Offers
Sign out
New Delhi 0C
Thursday, September 29, 2022
Start 15 Days Free Trial Subscribe Now
Register Free and get Exciting Deals