UN 
Recent research highlights that conventional weather forecasting models still hold a significant edge over artificial intelligence (AI) systems when it comes to predicting extreme and record-breaking weather events. Despite the rapid advancements in AI technologies across multiple sectors, their application in meteorology faces critical limitations, particularly in forecasting unprecedented and highly variable weather phenomena.
Traditional numerical weather prediction models rely on well-established physical laws and extensive historical climate data to simulate atmospheric conditions. These models incorporate complex equations governing fluid dynamics, thermodynamics, and radiation, enabling meteorologists to generate reliable forecasts even for rare and extreme weather scenarios. In contrast, AI models primarily depend on pattern recognition from past data, which may not fully capture the underlying physical processes driving novel extreme events.
Experts emphasize that while AI has shown promise in enhancing short-term weather predictions and identifying subtle patterns in large datasets, its predictive skill diminishes when confronted with unprecedented conditions that fall outside its training data scope. This limitation is particularly concerning in the context of climate change, where the frequency and intensity of extreme weather events are evolving in ways that challenge traditional statistical assumptions.
Improving the resilience and accuracy of weather forecasts is critical for disaster preparedness and sustainable development, aligning closely with the United Nations Sustainable Development Goals (SDGs), particularly SDG 13 on Climate Action. Accurate early warnings can save lives, protect infrastructure, and enable better planning for vulnerable communities. Therefore, integrating AI as a complementary tool to enhance, rather than replace, traditional forecasting methods is currently viewed as the most effective strategy.
Researchers advocate for continued investment in high-resolution physical models and enhanced observational networks, alongside the development of hybrid approaches that combine AI’s data processing capabilities with the robust physics-based frameworks of traditional models. Such synergy could improve forecast lead times and precision, ultimately supporting global efforts to mitigate climate risks and adapt to a rapidly changing environment.
