Weather forecasting

Weather forecasters rely on the physics of fluid dynamics and atmospheric thermodynamics to model and predict the behaviour of the atmosphere and future weather. Numerical weather prediction models that are based on these principles are used, with increasing sophistication, to simulate the evolution of the atmosphere and generate forecasts. These simulations require huge amounts of computer power (Figure 1).

Increasingly, artificial intelligence (AI) is being used to explore ways of improving the accuracy of weather forecasts. One strategy is to train AI using vast datasets of previous weather patterns to identify the most likely outcomes based on current measurements and previous evolutions of weather patterns – in other words, using pattern recognition to bypass much of the physics (Figure 2).