# Can AI help us better predict the weather?

## Executive summary

Google DeepMind's research demonstrates how AI is fundamentally transforming global weather forecasting by moving beyond traditional Numerical Weather Prediction (NWP). The latest model, WeatherNext 3, is notable for taking raw satellite imagery and predicting raw station observations in a single, end-to-end architecture. The discussion highlights the shift from deterministic single-point forecasts to probabilistic forecasting, which provides a range of possible scenarios crucial for managing extreme events and informing critical infrastructure decisions like energy load forecasting and disaster preparedness.

## Key takeaways

- AI's Advantage in Forecasting: AI models learn statistical patterns from historical weather data, allowing them to capture large-scale structures (e.g., across an entire hurricane) that traditional, localized NWP models might miss. This ability to process global structure improves accuracy, especially for complex events.
- WeatherNext 3 Architecture: WeatherNext 3 represents a significant architectural leap by taking raw satellite imagery and predicting raw station observations simultaneously, consolidating multiple traditional stages into a single, highly accurate model.
- Probabilistic Forecasting: Modern AI models generate multiple possible future scenarios (spaghetti plots) rather than a single deterministic guess. This probabilistic approach is vital for decision-making during extreme or rare events, allowing forecasters to quantify uncertainty.
- Hurricane Melissa Case Study: The model's ability to predict the rapid intensification and specific trajectory of Hurricane Melissa, providing critical lead time (3+ days) for evacuation and preparation, demonstrated the operational value of the AI system.
- Impact on Critical Systems: Beyond general weather, the model can predict specific variables like wind, solar, and temperature with high resolution, enabling better predictions for electrical load forecasting and agricultural planning.

## Technical details

- Numerical Weather Prediction (NWP): The traditional method involves supercomputers running algorithms based on physics equations (fluid dynamics). It approximates solutions by simulating weather at a coarse scale, requiring detailed input variables to solve for future states.
- Graphcast and Gencast: Graphcast was an early AI model (Phase 3) that took the full global state of the weather to simulate forecasts out to 10 days. Gencast was a follow-on that introduced probabilistic forecasting, predicting many likely scenarios instead of just the average expected weather.
- WeatherNext 3 Inputs/Outputs: Unlike traditional methods, WeatherNext 3 takes raw satellite imagery and predicts raw station observations, operating in a single model architecture. It also provides higher temporal resolution (hourly) and includes new variables like wind and solar power estimates.
- Probabilistic Techniques: Two methods are used to achieve probabilistic outputs: 1) Diffusion Models (refining noisy images into varied weather states) and 2) Functional Generative Networks (injecting different input scenarios to change network weights, creating output diversity).

## Practical implications

- Improved early warning systems for Category 5 storms, providing crucial lead time for evacuations.
- Enhanced prediction of renewable energy supply (wind and solar) and electrical load forecasting, aiding grid stability.
- Informing agricultural decisions regarding planting, harvesting, and drought resistance based on long-term temperature and rainfall forecasts.

## Topics

AI, Machine Learning, Weather Forecasting, Numerical Modeling, Climate Change, Graph Neural Networks, Probabilistic Modeling, WeatherNext 3, Weather Lab

Source: https://www.youtube.com/watch?v=O_EWbnkjXdk