From alert to action: Why India's weather warnings still miss the last mile

Why does a modern warning ecosystem still fail to protect people on the ground? Experts detail deeper problems in India's weather alert system

Uttar Pradesh recently saw severe storms that killed at least 117 people. A viral video from Bareilly showed a man being flung into the air after powerful winds ripped through the structure he was clinging to.

 

The incident raised a sharper question: If India’s weather forecasting and early warning systems have improved, why did so many people still die in one state?

 

Experts say the answer lies in the gap between warning and action.

Forecasts are improving, but what about protection?

India’s weather and disaster warning network is led by the India Meteorological Department, the National Disaster Management Authority, the Indian National Centre for Ocean Information Services, the Central Water Commission, state disaster management authorities and district administrations.

 

This system is supported by doppler weather radars, satellites, lightning detection systems, automatic weather stations, AI-backed forecasting models and geo-targeted mobile alerts.

 

The government has often cited India’s cyclone forecasting as a success story, with deaths falling sharply over the decades because of better forecasts, evacuation planning and disaster preparedness. But experts say that success has not been replicated equally for fast-moving storms, lightning and localised extreme weather.

The 'last-mile' gap

“The break is rarely at one single point,” said Dikshu C Kukreja, environmentalist, climate change expert and Managing Principal at CP Kukreja Architects. “India has made significant progress in forecasting, radar coverage and digital alerts, but the real weakness often lies in the last mile, where a warning must become a clear, trusted and actionable instruction.”

 

Kukreja said many deaths during the Uttar Pradesh storm were linked not only to strong winds but also to collapsing roofs, falling trees, lightning strikes and damaged electrical infrastructure.

 

“That shows that the gap is often between ‘information received’ and ‘protective action taken’,” he said.

 

Anupam Shrey, founder of climate risk management platform Plutas.ai, said the larger failure was in last-mile communication and local response.

 

“The last-mile communication gap is the binding constraint, not the absence of plans, but the absence of accountability,” he said.

 

According to him, frontline workers at block and panchayat levels often receive “no structured training, no communication scripts, and no defined role” when a red alert is issued.

Why storms remain difficult to predict

Experts said thunderstorms and lightning events are far harder to manage than cyclones. Cyclones are slower systems that can often be tracked days in advance. Thunderstorms and dust storms can form quickly, intensify within minutes and shift direction unpredictably.

 

“Predicting a fast-moving thunderstorm at the scale of a village or neighbourhood remains extremely difficult,” Kukreja said. “Such systems can form, intensify and shift direction within short time windows.”

 

Even when the broader forecast is accurate, the impact can vary sharply within a few kilometres depending on terrain, building quality, settlement density and the number of people outdoors.

 

“The future of forecasting must therefore move from weather prediction to impact prediction,” Kukreja said.

 

Shrey said district-level alerts are often too broad to help individuals make quick decisions.

 

“A district-level red alert covers hundreds of square kilometres. It tells a farmer nothing about whether the risk is 2 km away or directly overhead,” he said.

 

Agendra Kumar, Managing Director at GIS software and location intelligence platform Esri India, said India’s monitoring infrastructure also remains unevenly distributed.

 

“Rich weather and seismic monitoring infrastructure tends to cluster around urban and economically significant areas,” Kumar said. “Remote, rural, and tribal geographies, often the most exposed, are systematically under monitored.”

Warnings must change behaviour

India sends disaster alerts through SMS, apps and telecom-based broadcasts. But experts say the number of alerts sent is not the same as lives saved.

 

“A warning system cannot be judged only by the number of alerts issued,” Kukreja said. “The real test is not whether the message was delivered, but whether it changed behaviour on the ground.”

 

Shrey said repeated broad warnings can also create alert fatigue.

 

“The mortality numbers make this hard to dismiss,” he said while discussing alert fatigue. “Volume of warnings and reduction in casualties are clearly not the same thing.”

 

Kumar said warnings often fail when people distrust the source, have no safe shelter nearby or have ignored earlier false alarms.

 

“Communities that receive accurate, timely, spatially precise warnings still may not evacuate if they distrust the source, have nowhere safe to go, or have ignored false alarms before,” he said.

Weak infrastructure makes storms deadlier

Experts said many deaths are caused not by weather alone, but by the vulnerability of homes, roads, power lines and public spaces.

 

“A large part of the tragedy is linked to vulnerability in the built environment,” Kukreja said. “Weak roofs, informal housing, exposed electrical lines, unplanned settlements, poor-quality public shelters, unsafe trees and fragile utility infrastructure convert a weather event into a human disaster.”

 

Shrey said farmers, outdoor labourers and daily wage workers remain among the most exposed because they often lack access to safe shelter.

 

“No forecast, however accurate, protects someone whose roof cannot withstand the event,” he said.

 

“The system is improving at the top while vulnerability is deepening at the bottom,” Shrey said. “Climate change is widening that gap faster than institutions can close it.”

Can technology help?

Experts said GIS and GeoAI tools could help authorities map storm risks more precisely and identify vulnerable populations before disaster strikes.

 

“GeoAI can synthesise census microdata, building footprint classifications, road connectivity and population density into live risk indices,” Kumar said.

 

Such tools, he said, could help officials move from general alerts to location-specific warnings backed by local evacuation plans, vernacular messaging and community-level preparedness.

 

The video of a man being lifted into the air in Bareilly captured the force of the storm. But the larger warning from Uttar Pradesh is that India’s disaster systems cannot stop at forecasting danger from above. They must also ensure that people on the ground know exactly what to do, where to go and how to survive when the warning arrives.