Like many industries, artificial intelligence (AI) and machine learning (ML) will revolutionise weather forecasting in the coming years.
Scientific progress, including the development of solely AI-based weather forecasting models, has been continuing at pace in recent months. Organisations, including the Met Office, are publishing research, assessing opportunities and limitations, as well as developing and enhancing innovative AI-based models that can complement existing physics-based methods.
This includes the Met Office’s work with the Alan Turing Institute on the development of FastNet, a ML-based model using an approach known as a graph neural network, to predict weather patterns. Further updates on the ongoing performance of FastNet are expected later this year.
One key need for the weather industry is to ensure that benefits from AI can be fully realised in a safe and reliable way. So how exactly does that process take shape? And what’s the ongoing role for humans in an increasingly data-driven meteorological world?
Expertise paired with AI
With the science of AI and meteorology taking shape in the scientific community, the focus for the Met Office, as the UK’s national forecaster, is to ensure that potential improvements from AI forecasts can be implemented responsibly and transparently, underpinned by world-leading science in a way that doesn’t undermine existing confidence in forecasts and warnings.
That’s why, even at this developmental stage, the Met Office is drawing on the expertise of teams working directly with industry to explore how specific products and services can be fit for the next generation of forecasts for the industries they serve, including aviation, energy and the public.
Steve Ramsdale, Principal Operational Meteorologist at the Met Office, is helping shape how AI and ML are integrated into forecasting tools over the coming years.
Reflecting on current developments, Steve said: “While most of the public will know about the Met Office from our everyday forecasts and warnings that help people to stay safe, there is a wealth of products and services that we offer to industry, government and businesses that are also going to be influenced by use of AI in the coming years.
“We’re actively looking at what we offer and how human-AI interaction will continue to enhance our services, whether that’s for aviation, transport networks or the public.”
Forecasting the future
Because diverse industries and businesses are seeking different things from weather forecasts, bespoke products are likely to be enhanced, or even developed entirely, using advancements with AI.
Steve continued: “We’re planning to engage with people from across different work areas because we need that diversity of opinion on what the next generation of forecasts look like for different industries and what the AI-human interaction should be.
“We’re building AI systems in an intelligent manner. The simple thing would be to just throw AI-driven data out there and see what the results are and how people use it. But we’re ensuring that we make the most of the coming technological enhancements by putting experts from across industry at the centre of developments.”
By using expertise from across the breadth of Met Office services, we’re ensuring that the next stage of implementing AI into workflows will be as beneficial as possible, enabling Met Office expertise to add an extra layer of credibility and authority to forecasts.
Future work in the routine implementation of AI into weather forecasts will need involvement from across industry, science and technology to realise the most ambitious aims and will continue as scientific research continues at pace at the Met Office, academia and in industry.
Humans as decision-makers
One of the challenges of implementing AI solutions in different industries is that raw data is rarely useful for organisations.
Meteorologists will likely continue to play a key role in assessing model outputs, looking at biases and corrections and applying expert meteorological knowledge as needed. To do this effectively for the next generation of forecasts, Met Office staff are training in the use of data science and AI to ensure ongoing expertise can be properly applied in the next generation of weather forecasts.
Steve added: “In the age of ever-evolving forecasts and products, what’s going to be increasingly important is the expertise and trust that people have in the Met Office and our ability to extract knowledge from different outputs and empower decision-makers.”
Find out more about our work with the Alan Turing Institute.
Check the latest update on FastNet.
