Latest news with #HenryFrancisBlanford
&w=3840&q=100)
Business Standard
20-05-2025
- Climate
- Business Standard
Delhi weather: Why IMD keeps predicting rain but the heat won't quit
On paper, Delhi was supposed to have a 'pleasant day' on Tuesday (May 20). The India Meteorological Department had issued a yellow alert — thunderstorms, dust storms, some strong surface winds, maybe a touch of lightning. But as any Delhiite could tell you, the city felt more like a 'tandoor.' A blazing, suffocating heat wrapped the capital, and even the winds did little more than shuffle around the hot air. Pleasant? Not by a long stretch. It's not just a one-off mistake either. Time and again, the IMD has struggled to forecast Delhi's weather with any meaningful accuracy. From missing sudden storms that blow roofs off metro stations to underestimating rainfall that paralyses entire cities, India's official weather agency finds itself repeatedly caught off guard. And that begs the question: after more than 150 years of operations, why is the IMD still struggling to get it right? How often do IMD forecasts go wrong? On May 2 this year, Delhi was battered by a torrential downpour — 77 mm of rainfall in a single day, the second-highest May total since 1901. It was the wettest day in over eight months. And yet, not even a whisper of this made it to the IMD's forecasts the night before. On June 28, 2023, an unpredicted storm dumped 91 mm of rain on Delhi in a single hour. The agency had warned of 'light to moderate' rainfall. What came instead looked like the start of a monsoon apocalypse — flooding roads, disrupting traffic, collapsing infrastructure. The story repeats across the country. In December 2023, Tamil Nadu was devastated by heavy rainfall that killed at least 10 people. Once again, the IMD failed to forecast the intensity of the storm. Its Director General, Mrutyunjay Mohapatra, later admitted that while rain had been predicted, the storm's ferocity caught the department off guard. A warning was only issued in the early morning hours—too late for most to prepare. Ironically, the IMD is globally respected for its monsoon modelling. It uses the Dynamical Model, a complex climate system-based tool involving ocean-atmosphere coupling, and has recently integrated data from the high-performance computing system Mihir, one of India's most powerful supercomputers. Why did India start forecasting weather in the first place? The story of India's weather forecasting woes begins not today, but nearly a century and a half ago. Founded in 1875, the IMD's first mission was to crack the code of the monsoon. Back then, it wasn't just about rain. Famine stalked the land, and agriculture was wholly dependent on monsoon rainfall. The British colonial government saw meteorology as a matter of life, death and revenue. The IMD's first meteorologist, Henry Francis Blanford, tried to connect Himalayan snow cover with monsoon rainfall. His successor, John Eliot, added data from Australia and the Indian Ocean. But no matter how many charts they drew or patterns they spotted, the forecasts still failed to prevent famines. In 1899–1900, more than a million Indians died in a famine that Eliot had confidently predicted would not come. In 1904, Sir Gilbert Walker took the reins. A statistician by training, Walker introduced the idea of global pressure patterns—including what we now call the Southern Oscillation, part of the El Niño system—into monsoon prediction. He identified 28 variables that seemed to influence rainfall. But even then, forecasting was more art than science. How accurate are IMD forecasts today? India's seasonal monsoon predictions, vital for both agriculture and the broader economy, have historically had limited precision. Over the past 20 years, the IMD has achieved an average accuracy of just 42 per cent for its initial monsoon forecasts. That means in 14 out of 24 years since 2001, the actual rainfall deviated by more than five percentage points from the early forecast—making it statistically less reliable than a coin flip, according to a Hindustan Times analysis. That said, IMD's forecasting reliability has improved over the last five years, with fewer large deviations than in the past. The IMD's official monsoon outlook carries a standard error margin of ±5 per cent. For instance, the 2025 forecast anticipates rainfall at 105 per cent of the Long Period Average (LPA), allowing for this range. In recent years, the IMD has enhanced its precision for short-range and extreme weather forecasts by 40–50 per cent, largely due to high-resolution numerical models and AI tools. What models has IMD used and replaced? For nearly a century, India's monsoon forecasting relied on tweaking statistical models. In 1988, a new power regression model introduced by Vasant Gowariker promised more accurate forecasts using 16 parameters. But over time, many of those predictors lost relevance. The model failed to predict droughts in 2002 and 2004, and confidence in it eroded. From 2007, IMD began using ensemble statistical forecasting—blending multiple models to generate a more accurate estimate. Between 2007 and 2018, the average forecast error dropped from 7.94 per cent to 5.95 per cent of the LPA. That's progress — but still not precision. What are the latest forecasting systems IMD is using? The real shift came with the launch of the Monsoon Mission Coupled Forecasting System (MMCFS) in 2012. Unlike statistical models, MMCFS integrates data from oceans, land and atmosphere to simulate monsoon behaviour. In 2021, a multi-model ensemble (MME) system was added, incorporating forecasts from climate centres in the US, Japan and other countries. The rationale: if one model fails, maybe five together won't. Since these upgrades, IMD forecasts have become more accurate—at least on paper. The government claims that seasonal rainfall forecasts now have 21 per cent less error than those issued in the 1990s and early 2000s. But public confidence remains low. Ask someone in Delhi when they last trusted an IMD forecast, and you're likely to get a sarcastic shrug. Why are extreme events still so hard to predict? Part of the issue is scientific. Thunderstorms and cloudbursts are inherently difficult to predict. They form quickly, behave erratically, and require ultra-high-resolution data. For such events, IMD relies on nowcasts—forecasts issued just 2–3 hours in advance. But systemic constraints persist. Countries like the US and UK have long benefited from dense weather station networks, rapid data transmission and vast computing power. India is still catching up. Under Mission Mausam, the IMD has increased its radar capacity from 26 to 39, with plans to scale up to 126. Doppler radars are being installed nationwide, and a new high-density mesonet—automated local weather stations—is being rolled out in major cities. New tools like microwave radiometers and wind profilers are also being introduced. How does IMD compare to international weather agencies? Unlike the US National Weather Service or the European Centre for Medium-Range Weather Forecasts, the IMD often struggles with clarity and urgency in communication. Alerts may go out late at night. Warnings often lack actionable detail. And public trust—once lost—is hard to win back. In July 2023, Bengaluru received 132 mm of rain in under four hours. The IMD had issued only a 'moderate rain' alert. Lakes overflowed, tech parks shut down, and people were seen kayaking through flooded streets. Compare that with the Netherlands—one of the most flood-prone countries—where smart drainage systems are linked to forecasts, adjusting pumping capacity in real time. At its core, weather forecasting is not just a scientific challenge. It's a public service. Getting it wrong isn't just embarrassing—it's dangerous.
Indian Express
28-04-2025
- Climate
- Indian Express
The history and evolution of monsoon forecasting in India
The India Meteorological Department (IMD) has forecast 'above normal' rainfall — 105% of the long-period average (LPA) — during the June-September southwest monsoon season. The IMD said earlier this month that all major drivers of the Indian monsoon, such as El Niño-Southern Oscillation (ENSO) in the equatorial Pacific Ocean and the Indian Ocean Dipole (IOD), were favourable. The four-month southwest monsoon season brings almost 70% of the country's annual rainfall. It is critical for agriculture and crops, for the economy as a whole, and to recharge reservoirs and aquifers. Accurate forecasts of the monsoon are key for the government to prepare for a range of eventualities. The April 15 forecast was the first of the IMD's long-range forecasts for this year's monsoon. A second-stage or updated forecast will be made in the last week of May, ahead of the monsoon striking the coast of Kerala. Long-range forecasts can be made for 30 days to up to two years into the future. The first forecasts A systematic effort to forecast monsoon rainfall began in 1877, two years after the IMD was established with the British meteorologist and palaeontologist Henry Francis Blanford as the first Meteorological Reporter to the Government of India. Crop failure that began in the Deccan plateau in the previous year had set off the Great Famine of 1876-78, and the effects were felt across the country by 1877. The colonial administration saw an acute need to understand the arrival of the monsoon and the distribution of rain over the country. 'The success of the monsoons dictated agricultural production and the health of rivers, coasts, and shipping lanes — i.e., revenue generation for British interests,' Ramesh Subramanian of Quinnipiac University in the US wrote in his paper 'Monsoons, Computers, Satellites: History and Politics of Weather Monitoring in India' (2021). SNOW & RAIN …: The first tentative forecasts of the monsoon were provided by Blanford between 1882 and 1885, who analysed the relationship between Himalayan snow cover and the amount of rainfall over the Indian region. Blanford's forecasts were 'based on the inverse relationship between Himalayan winter and spring snow accumulation and subsequent summer monsoon rainfall over India. It was assumed that, in general, varying extent and thickness of the Himalayan snow has a great and prolonged influence on the climate conditions and weather of the plains of northwest India,' the IMD says in its official account of the evolution of meteorology in India. In 1886, Blanford made the first long-range forecast (LRF) of monsoon rainfall for the whole of India and Burma, based on this inverse relationship hypothesis. Blanford was succeeded by Sir John Eliot, who was appointed the first Director General of Indian Observatories, equivalent to the position of the head of the IMD today, in May 1889 at its Calcutta headquarters. …PLUS SOME OTHER FACTORS: Eliot took forward Blanford's work, combining data on Himalayan snow with factors such as local Indian weather conditions in April-May and conditions over the Indian Ocean and Australia to issue his LRFs. But like Blanford, Eliot still could not effectively predict droughts or the famines that followed, bringing starvation and deaths. The Indian Famine of 1899-1900, which is estimated to have killed between a million and 4.5 million people, struck in a year for which Eliot had predicted better-than-normal rain. The first colonial official who sought to incorporate the influence of global factors on the Indian monsoon was the physicist and statistician Sir Gilbert Walker, who succeeded Eliot in 1904. 28 PREDICTORS, STATISTICAL CORRELATIONS: Walker developed the first objective models based on statistical correlations between monsoon rainfall and antecedent global atmospheric, land, and ocean parameters. To make his forecasts, Walker identified 28 parameters or predictors with a significant and stable historical relationship with the Indian monsoon. Walker described three large-scale see-saw variations in global pressure patterns — Southern Oscillation (SO), North Atlantic Oscillation (NAO), and North Pacific Oscillation (NPO). 'Among these, SO was found to have the most significant influence on the climate variability of India as well as many parts of the globe… The SO…was later linked to the unusual warming of sea surface waters in the eastern tropical Pacific Ocean or El Niño by Jacob Bjerknes in the 1960s,' says the IMD. Walker also reasoned that the Indian subcontinent could not be considered as an undivided whole for the purpose of forecasting the measure of rainfall, and divided the region into three subregions: Peninsula, Northeast, and Northwest India. After Independence The IMD stayed with Walker's model of monsoon forecasting until 1987. The forecasts were not very accurate. 'The average error of the predictions for the peninsula was 12.33 cm and 9.9 cm for NW India during the period 1932-1987,' M Rajeevan, a former Secretary to the Ministry of Earth Sciences, and IMD Scientist D R Pattanaik wrote in their paper, 'Evolution of Monitoring and Forecasting of Southwest Monsoon' (Mausam, IMD's quarterly journal, 2025). The main problem was that several of the parameters identified by Walker had lost significance over time — meaning their relationship with the monsoon was no longer the same. IMD scientists attempted several tweaks to the model, but its accuracy did not improve greatly. GOWARIKER MODEL: In 1988, the IMD began to issue operational forecasts of the monsoon based on a power regression model developed by scientists led by Vasant R Gowariker, which used 16 empirically derived atmospheric variables as predictors in a statistical relationship with the total rainfall. The forecast for geographical regions was discontinued in favour of a forecast for the season over the country as a whole. Operational forecasts for Northwest India, Peninsular India, and Northeast India were reintroduced in 1999, but the geographical boundaries of these regions were different. Similar issues emerged in the new model as well. 'In the year 2000, it was realised that out of the sixteen parameters, four of them have lost their correlation' with the monsoon, 'and hence they were replaced by other predictors', wrote Suryachandra A Rao, Prasanth A Pillai, Maheshwar Pradhan and Ankur Srivastava in their 2019 paper, 'Seasonal Prediction of Indian Summer Monsoon in India: The Past, the Present and the Future' (Mausam). The IMD's regional forecasts remained inaccurate during this period. 'The forecast error was more than model error for years like 1994, 1997 and 1999,' Rao et al wrote. The power regression model was critically evaluated after it failed to predict the drought of 2002 that followed 14 good monsoons and was the worst since 1987. TWO NEW MODELS: In 2003, the IMD introduced two new models of monsoon prediction, with eight and 10 parameters. It also adopted a new two-stage forecast strategy. The first stage forecast was issued in mid-April, and an update or second stage forecast was issued by the end of June. The new models accurately predicted the 2003 monsoon, but failed to forecast the drought of 2004, sending the IMD back to the drawing board. The Department re-evaluated its models with two major objectives: '(a) a re-visit of the suitable and stable predictors, which have physical relationships with monsoon rainfall and (b) critical way of model development based upon identifying the optimum number of predictors and optimum model training period etc.,' according to the 2019 study. STATISTICAL FORECASTING SYSTEM: In 2007, the IMD came up with a Statistical Ensemble Forecasting System (SEFS) to support its two-stage forecast strategy, and further reduced the number of parameters in its models. A five-parameter model replaced the eight-parameter model for the first forecast in April, and a new six-parameter model replaced the 10-parameter model for the forecast update in June. The intention was to ensure there was no 'overfitting' of models, in which a model matches or memorises the training set so closely that it fails to make correct predictions based on new data. The Department also introduced the concept of ensemble forecasts. In this method, all possible forecasting models based on all the combinations of predictors are considered to create a single, more robust prediction. The new system helped the IMD improve its forecast significantly. The average absolute error (difference between forecast and actual rainfall) between 2007 and 2018 was 5.95% of the LPA (rainfall recorded over a particular region for a given interval) compared with the average absolute error of 7.94% of LPA between 1995 and 2006. Forecasts in recent years COUPLED DYNAMIC MODEL: The improvement in monsoon prediction was also due to the launch of the Monsoon Mission Coupled Forecasting System (MMCFS) in 2012. This was a coupled dynamic model, which could combine data from the ocean, atmosphere, and land to provide more accurate forecasts. The IMD used MMCFS along with the SEFS for its predictions. MULTI-MODEL ENSEMBLE: The accuracy of forecasts was further enhanced with the launch of a system based on a 'multi-model ensemble (MME)' in 2021. This new MME system used the coupled global climate models (CGCMs) from various global climate prediction and research centres, including India's own MMCFS model. Since the introduction of SEFS in 2007 and the MME approach in 2021, the IMD's operational forecasts for the monsoon have improved noticeably, the Ministry of Earth Sciences informed Parliament this February. BETTER FORECASTS, SCOPE FOR IMPROVEMENT: The absolute forecast error in all of India's seasonal rainfall reduced by about 21% during the years 2007-2024 compared with the same number of years between 1989 and 2006, Earth Sciences Minister Dr Jitendra Singh told Rajya Sabha. IMD's April forecasts, too, have become more accurate. The actual rainfall in the previous four years (2021-2024) deviated from the April forecast by 2.27 percentage points, well within the forecast range of 4%. However, there is still much scope for IMD to improve. In their paper, Rajeevan and Pattanaik pointed out that the Department should refine its dynamical models by improving systematic errors/ biases and teleconnectivity — significant relationships or links between weather phenomena — with global climate modes such as ENSO.
