Some leading South Asian scientists have taken the initiative to use Artificial Intelligence (AI) models that integrate satellite imagery, weather forecasts, historical flood data, and topography to develop systems capable of predicting floods up to 72 hours in advance in parts of South Asia, with a special focus on Sri Lanka. These systems analyse multiple datasets and have demonstrated promising accuracy—up to 85 percent for major climate events in some river basins—particularly in rainfall and extreme weather forecasting.

The scientists are confident that AI can forecast extreme rainfall volumes, cyclone behaviour, dam inflows, and downstream flooding by combining meteorological, hydrological, and satellite data. Such capabilities could have significantly reduced fatalities in recent disasters such as Cyclone Ditwah, which caused widespread devastation in Sri Lanka, Indonesia, Thailand, Malaysia, and parts of India last month.

Dr Nalinda Somasiri, Associate Professor and Associate Dean (Data Science & Computer Science), York St John University – London Campus, together with his team—including Dr Swathi Ganesan, Associate Dean; Dr Soonleh Ling; Dr Anu Bala; Dr Rebecca Balasundaram; Sangita Pokhrel; and Dr Rashmi Siddalingappa—has taken the initiative to launch a GeoAI programme. This initiative brings together governments, universities, and technical experts to use AI combined with GIS (geospatial data) to improve early warnings, formulate danger maps, enable real-time hazard monitoring, and enhance coordination across agencies.

Cyclone Ditwah: A Wake-Up Call

Cyclone Ditwah’s devastating impact on Sri Lanka has revealed a truth the nation can no longer ignore, Dr Somasiri stated. Climate-driven disasters are becoming more frequent, more destructive, and more unpredictable. Flash floods swallowed communities overnight, while landslides in the central highlands buried roads, homes, and entire families. Thousands were displaced, hundreds lost their lives, and critical infrastructure was crippled.

This initiative is aligned with the United Nations’ Early Warnings for All programme. It focuses on building AI-enabled early-warning systems, climate intelligence, and resilient infrastructure planning, particularly for countries facing intensifying climate pressures.

Cyclone Ditwah has shown that Sri Lanka urgently needs AI-enabled capabilities across the prediction, response, and rebuilding phases. Dr Somasiri identified five dimensions where AI can revolutionise Sri Lanka’s disaster management.

AI, GIS and Real-Time Risk Mapping

GIS integrated with machine learning enables pattern recognition from maps and satellite imagery, helping automate the detection of changes in land conditions, water levels, and risk zones. AI and remote sensing are being combined to map landslide susceptibility and provide early warnings.

Real-time monitoring through satellites, drones, and large-scale data processing allows the identification of early hazard indicators such as rising floodwaters, soil saturation, slope instability, and vegetation changes that may signal wildfire risks or increasing vulnerability. These systems help authorities visualise evolving threats and act before disasters unfold.

Dr Somasiri is of the view that AI can combine elevation models, soil types, slope gradients, land use patterns, plantation coverage, rainfall intensity, and historical landslide data. Machine-learning models can then generate dynamic landslide risk indices updated every few hours.

Preparing for Unprecedented Disasters

Modern disasters often exceed historical precedents. Generative AI can create synthetic extreme-event scenarios, including cyclone tracks more intense than Ditwah, simulated rainfall patterns never previously recorded, cascading landslides triggered by prolonged rainfall, and dam stress-test scenarios. These simulations allow disaster agencies, local authorities, and the military to rehearse and prepare for “future disasters we have never seen,” he said.

Disaster messaging must evolve from generic alerts to personalised warnings, geographically precise risk assessments, multilingual instructions, and impact-based guidance. For example, people could receive alerts stating: “Your area may experience one metre of floodwater; move to higher ground within two hours.”

To enable this, AI systems can integrate real-time rainfall data, reservoir telemetry, satellite updates, public SOS messages, road accessibility status, hospital capacity, and forecast uncertainties. The result is clear, actionable, human-friendly alerts delivered via mobile phones, television channels, radio networks, and public sirens.

This AI-powered early-warning system mirrors the kind of platform Dr Somasiri helped design at Motorola, where situational-awareness software, real-time intelligence feeds, and rapid communication converge to save lives.

Post-Disaster Recovery, Resilience Planning

Post-disaster mapping can be conducted using AI-driven satellite image analysis to identify collapsed buildings, map inundated areas with 90 percent accuracy, detect blocked roads and washed-away bridges, estimate agricultural losses, and prioritise areas requiring urgent relief. Such AI-supported systems can replace weeks of manual mapping with near-instant assessments.

AI can also assist governments in identifying underserved communities and addressing critical recovery questions: Which households require urgent resettlement? Where should temporary shelters be expanded? Which bridges, hospitals, and schools must be rebuilt first? What infrastructure upgrades will reduce future risk?

Through vulnerability analytics, AI can highlight populations more likely to struggle with recovery, including plantation communities, low-income families, and settlements located on flood plains. India has already allocated a US$450 million fund for Sri Lanka’s post-cyclone recovery. The joint committee established by India and Sri Lanka to manage this fund will be able to implement AI-based disaster warning systems under Sri Lanka’s digitalisation programme, which is being supported by India.

(The author, a former Sri Lankan diplomat, is a political and strategic affairs commentator. Views expressed are personal. He can be reached at sugeeswara@gmail.com.)