When Cyclone Arnab began crossing the Andhra Pradesh coast near Yanam early on October 28, it closely followed forecasters' predictions. The India Meteorological Department's (IMD) bulletin the previous afternoon had placed the storm's position at 05:30 IST within about 10 km of where it was subsequently observed. It had also forecast a forenoon crossing near Yanam, with winds of 125–135 km/h, gusting to 150 km/h.
The European Centre for Medium-Range Weather Forecasts' (ECMWF) 51-member ensemble had assigned a 35% probability to landfall between the Godavari delta and Yanam, with a median landfall time of 07:11 IST.
-
Indian Markets Turning Corner After Two Tough Years, Says Sandeep Tandon -
NDTV Profit Ignite 2026: Experts See Opportunity In Equities, Stay Bullish For Long-Term -
Buy Right, Sit Tight Does Not Work for Individual Stocks, Says Dinshaw Irani -
'Worst Is Behind': Quant MF Sees Market Bottom, Flags 'Extreme Euphoria' In Midcaps -
India's Mutual Fund Industry Could Become A $1 Trillion Opportunity, Say Fund Managers -
'Foreign Investors Are Baraatis': Raamdeo Agrawal's Take On LTCG Tax -
Stay Invested Through Market Cycles: Raamdeo Agrawal at NDTV Profit Ignite -
Ramesh Damani Remains Bullish On India For Next 20-30 Years, Sees Bargains Amid Market Weakness -
CAS Came After A Long Consultation Process: SEBI Chairman Tuhin Kanta Pandey AT NDTV PROFIT IGNITE -
Aim To Cut FPI Onboarding Time From Mth To Five Days: SEBI Chairman Tuhin Kanta Pandey AT NDTV PROFI
From A Single Track To Range Of Possibilities
Manmeet Singh, Academic Visitor, Safexpress Centre for Data, Learning and Decision Sciences, Ashoka University and Sandeep Juneja, Professor of Computer Science and Director, Safexpress Centre for Data, Learning and Decision Sciences, Ashoka University said that an ensemble runs a forecasting model multiple times from slightly different starting conditions.
ECMWF's ensemble uses a physics-based model to generate 51 possible outcomes. Their spread helps forecasters estimate both the storm's likely path and the uncertainty around it. These forecasts also put numbers on local risks: for example, a 73% probability that the grid point nearest Visakhapatnam would receive at least 115.6 mm of rain in 24 hours.
AI models, trained on decades of weather data, can generate ensembles in minutes. This speed allows forecasters to explore many more possible outcomes.
AI Joins Forecast Desk
Google DeepMind's WeatherNext 3, launched on September 3, reads live satellite imagery directly and produces a new global forecast every hour, at resolutions as fine as 5 km, compared with its predecessor's 25 km grid. Google says its rainfall forecasts are up to 50% more accurate. At launch, the model topped Operational WeatherBench, Brightband's live scoreboard comparing AI and physics-based global forecasting models.
Generative models such as DeepMind's diffusion-based GenCast produce multiple plausible weather scenarios. This helps capture extreme events that can be obscured when predictions are averaged into a single forecast. In August, a WeatherNext model described in Nature predicted cyclone tracks, intensity and wind structure together, say , say Manmeet Singh and Sandeep Juneja.
ALSO READ | NASA Reveals What Cyclone Arnab Looked Like From Space As It Hit India
Pinpointing Rainfall Remains Difficult
Arnab also illustrates a persistent challenge: predicting exactly where the heaviest rain will fall. IMD warnings identify the risk of isolated extreme rainfall across a district, while a roughly 25 km model grid can smooth out local peaks.
Diffusion-based downscaling models aim to bridge this gap. NVIDIA's CorrDiff converts 25 km data into 2 km detail and is being used in Taiwan to capture typhoon features. Its sibling, StormCast, generates hourly storm-scale ensembles at 3 km resolution over the United States.
Together, these approaches could help forecasters better estimate where severe weather may strike, how intense it may be and how much uncertainty remains. Translating those probabilities into timely, accurate warnings is the next step, concludes Singh and Juneja.
ALSO READ | Cyclone Arnab Wreaks Havoc In Odisha, Over 55,000 Evacuated Across Nine Districts
Essential Business Intelligence, Sharp Market Insights, Practical Personal Finance Advice, Daily Fuel, Gold and Silver Prices and Latest Stories — On NDTV Profit.