Artificial intelligence for linking extreme weather forecasting and health impact prediction

Extreme weather events increasingly threaten human health, yet current weather-health systems often struggle to translate hazard-based forecasts into actionable, impact-oriented guidance. This gap arises primarily from spatiotemporal misalignment between meteorological and health data, reliance on static hazard thresholds, and limited integration of probabilistic risk assessment. Here, we propose a conceptual framework that leverages artificial intelligence (AI) to bridge extreme weather forecasting and health impact prediction. The AI-enabled system aligns heterogeneous data sources, captures nonlinear and lagged exposure-response relationships, and propagates uncertainty throughout the prediction chain. We illustrate the framework using heatwaves, wildfires, and extreme precipitation and outline key priorities for operational implementation, including model interpretability, privacy-preserving architectures, and institutional governance. Integrating interpretable and uncertainty-aware health impact prediction into operational forecasting systems will be essential to enabling anticipatory public health action in a future of intensifying weather extremes.

Introduction

Extreme weather events are increasing in both frequency and intensity under global warming, posing escalating risks to human health worldwide.1 Integrating meteorological forecasting with public health preparedness has therefore become a central component of climate risk reduction and an important pillar of national resilience strategies.2 International initiatives led by the World Meteorological Organization (WMO) and World Health Organization (WHO) have advanced weather-health early-warning capabilities, and many countries have established operational systems linking meteorological information with public health interventions.3

Despite these advances, the integration of extreme weather forecasting with public health response remains limited. Most existing systems rely on static hazard thresholds and reactive protocols, making limited use of high-resolution forecasts, probabilistic uncertainty, or dynamically evolving exposure and vulnerability conditions. This misalignment between hazard prediction and health impact assessment constrains the ability to anticipate risks and implement timely and targeted interventions.

Artificial intelligence (AI) offers a transformative opportunity to bridge the gap.4,5 Existing studies have largely focused either on improving hazard prediction or on refining epidemiological exposure-response relationships, with few efforts to integrate both within operational forecasting systems. Here, we propose a conceptual framework for an end-to-end AI-enabled weather-health prediction system. This framework aligns heterogeneous meteorological and health data, captures nonlinear and lagged exposure-response relationships, propagates uncertainty, and supports decision-relevant risk assessment. By positioning AI as a coordinating layer between meteorological forecasts and public health response, this perspective outlines a roadmap for anticipatory and impact-based health risk management.