FastNet

Introduction

Weather forecasts play a vital role in our daily lives, influencing a multitude of day-to-day decisions, from planning our commutes to determining outdoor activities. Weather forecasts also aid in mitigating risks associated with natural disasters and in optimising resources for various industrial sectors including agriculture, transportation, and energy. As climate change accelerates, we are experiencing more frequent and intense weather extremes across the world, displacing communities and damaging infrastructure. The time has come to adopt artificial intelligence (AI) as a force for good and to reimagine the development of numerical weather prediction to help increase society’s resilience to weather extremes.

The collaboration between the Met Office and The Alan Turing Institute is helping maintain the UK’s position at the cutting edge of weather prediction and deploying AI for public good.

Researchers are developing FastNet a new AI model, known as a graph neural network, to forecast weather patterns. This will allow them to test the accuracy of their new model against existing NWP weather forecasting methods . FastNet draws together the Met Office AI projects and Turing expertise to drive forward development for data-driven approaches to weather forecasting. Met Office experts have been able to provide crucial training in meteorology and the metrics used to evaluate weather models, while researchers from the Turing have shared their expertise in AI model design and optimisations for scaling-up on cloud platforms.

This partnership has galvanised a new cross-disciplinary team, and the exchange of knowledge and skills is accelerating innovation and driving forward the field of weather prediction.

Explaining the science

Weather phenomena occur across various scales, from local short-lived disturbances to extensive global atmospheric circulation patterns, and across various weather variables including temperature, rainfall, pressure and humidity. The complex and non-linear interplay between these weather scales and variables not only contributes to the daily weather we experience, but also the ability of meteorologists to predict it ahead of time.

The challenge of forecasting our weather should certainly not be underestimated. As datasets continue to expand and become ever more varied, and with the increasing need for more granular and skilful predictions, the time is now upon us to look to use new advancements in AI and machine learning to transform weather forecasting and the way we develop and run numerical weather prediction (NWP) models.

Machine learning algorithms are often trained on existing datasets such as ERA5 reanalysis which are themselves made up from vast observational data - including various satellites, ship measurements, and surface sensors – to automatically recognise complex patterns in weather phenomena. These AI weather models can already compete with the current generation of NWP models and, for some applications, exceed their performance at a fraction of the computational cost.

Project aims

The Met Office and Turing teams will work together to realise and maximise the potential of AI and ML to complement and enhance the forecasting excellence being achieved by physics-based modelling.

Applications

The ultimate goal of this collaboration is to operationalise the FastNet model so that the Met Office can use the optimal blend of physics-based and ML-based modelling for UK weather prediction. This means that AI could be used alongside physics-based models and play an important role in delivering the daily forecasts that we all rely on. We believe that using AI alongside physics-based numerical models provides the most robust way forward in a changing climate.

Building on initial success with FastNet in global predictions, the partnership will next develop high-resolution regional forecasts for the UK using AI methods. By ingesting past UK weather forecasts into the FastNet training process, we will further train and refine it to produce detailed regional UK weather forecasts for operational use.

As AI continues to evolve and integrate into our lives, the outlook for the future of weather forecasting looks brighter than ever.

Animation comparing ERA5 (which uses a blend of observations and NWP model to produce the best estimate of historical reality) with a prediction from our FastNet machine learning model.