Computer models that use artificial intelligence (AI) cannot forecast record-breaking weather as well as traditional climate models, according to a new study.
It is well established that AI climate models have surpassed traditional, physics-based climate models for some aspects of weather forecasting.
However, new research published in Science Advances finds that AI models still “underperform” in forecasting record-breaking extreme weather events.
The authors tested how well both AI and traditional weather models could simulate thousands of record-breaking hot, cold and windy events that were recorded in 2018 and 2020.
They find that AI models underestimate both the frequency and intensity of record-breaking events.
A study author tells Carbon Brief that the analysis is a “warning shot” against replacing traditional models with AI models for weather forecasting “too quickly”.
AI weather forecasts
Extreme weather events, such as floods, heatwaves and storms, drive hundreds of billions of dollars in damages every year through the destruction of cropland, impacts on infrastructure and the loss of human life.
Many governments have developed early warning systems to prepare the general public and mobilise disaster response teams for imminent extreme weather events. These systems have been shown to minimise damages and save lives.
For decades, scientists have used numerical weather prediction models to simulate the weather days, or weeks, in advance.
These models rely on a series of complex equations that reproduce processes in the atmosphere and ocean. The equations are rooted in fundamental laws of physics, based on decades of research by climate scientists. As a result, these models are referred to as “physics-based” models.
However, AI-based climate models are gaining popularity as an alternative for weather forecasting.
Instead of using physics, these models use a statistical approach. Scientists present AI models with a large batch of historical weather data, known as training data, which teaches the model to recognise patterns and make predictions.
To produce a new forecast, the AI model draws on this bank of knowledge and follows the patterns that it knows.
There are many advantages to AI weather forecasts. For example, they use less computing power than physics-based models, because they do not have to run thousands of mathematical equations.
Furthermore, many AI models have been found to perform better than traditional physics-based models at weather forecasts.
However, these models also have drawbacks.
Study author Prof Sebastian Engelke, a professor at the research institute for statistics and information science at the University of Geneva, tells Carbon Brief that AI models “depend strongly on the training data” and are “relatively constrained to the range of this dataset”.
In other words, AI models struggle to simulate brand new weather patterns, instead tending forecast events of a similar strength to those seen before. As a result, it is unclear whether AI models can simulate unprecedented, record-breaking extreme events that, by definition, have never been seen before.
Record-breaking extremes
Extreme weather events are becoming more intense and frequent as the climate warms. Record-shattering extremes – those that break existing records by large margins – are also becoming more regular.
For example, during a 2021 heatwave in north-western US and Canada, local temperature records were broken by up to 5C. According to one study, the heatwave would have been “impossible” without human-caused climate change.
The new study explores how accurately AI and physics-based models can forecast such record-breaking extremes.
First, the authors identified every heat, cold and wind event in 2018 and 2020 that broke a record previously set between 1979 and 2017. (They chose these years due to data availability.) The authors use ERA5 reanalysis data to identify these records.
This produced a large sample size of record-breaking events. For the year 2020, the authors identified around 160,000 heat, 33,000 cold and 53,000 wind records, spread across different seasons and world regions.
For their traditional, physics-based model, the authors selected the High RESolution forecast model from the Integrated Forecasting System of the European Centre for Medium-Range Weather Forecasts. This is “widely considered as the leading physics-based numerical weather prediction model”, according to the paper.
They also selected three “leading” AI weather models – the GraphCast model from Google Deepmind, Pangu-Weather developed by Huawei Cloud and the Fuxi model, developed by a team from Shanghai.
The authors then assessed how accurately each model could forecast the extremes observed in the year 2020.
Dr Zhongwei Zhang is the lead author on the study and a researcher at Karlsruhe Institute of Technology. He tells Carbon Brief that many AI weather forecast models were built for “general weather conditions”, as they use all historical weather data to train the models. Meanwhile, forecasting extremes is considered a “secondary task” by the models.
The authors explored a range of different “lead times” – in other words, how far into the future the model is forecasting. For example, a lead time of two days could mean the model uses the weather conditions at midnight on 1 January to simulate weather conditions at midnight on 3 January.
The plot below shows how accurately the models forecasted all extreme events (left) and heat extremes (right) under different lead times. This is measured using “root mean square error” – a metric of how accurate a model is, where a lower value indicates lower error and higher accuracy.
The chart on the left shows how two of the AI models (blue and green) performed better than the physics-based model (black) when forecasting all weather across the year 2020.
However, the chart on the right illustrates how the physics-based model (black) performed better than all three AI models (blue, red and green) when it came to forecasting heat extremes.

The authors note that the performance gap between AI and physics-based models is widest for lower lead times, indicating that AI models have greater difficulty making predictions in the near future.
They find similar results for cold and wind records.
In addition, the authors find that AI models generally “underpredict” temperature during heat records and “overpredict” during cold records.
The study finds that the larger the margin that the record is broken by, the less well the AI model predicts the intensity of the event.
‘Warning shot’
Study author Prof Erich Fischer is a climate scientist at ETH Zurich and a Carbon Brief contributing editor. He tells Carbon Brief that the result is “not unexpected”.
He adds that the analysis is a “warning shot” against replacing traditional models with AI models for weather forecasting “too quickly”.
The analysis, he continues, is a “warning shot” against replacing traditional models with AI models for weather forecasting “too quickly”.
AI models are likely to continue to improve, but scientists should “not yet” fully replace traditional forecasting models with AI ones, according to Fischer.
He explains that accurate forecasts are “most needed” in the runup to potential record-breaking extremes, because they are the trigger for early warning systems that help minimise damages caused by extreme weather.
Leonardo Olivetti is a PhD student at Uppsala University, who has published work on AI weather forecasting and was not involved in the study.
He tells Carbon Brief that “many other studies” have identified issues with using AI models for “extremes”, but this paper is novel for its specific focus on extremes.
Olivetti notes that AI models are already used alongside physics-based models at “some of the major weather forecasting centres around the world”. However, the study results suggest “caution against relying too heavily on these [AI] models”, he says.
Prof Martin Schultz, a professor in computational earth system science at the University of Cologne who was not involved in the study, tells Carbon Brief that the results of the analysis are “very interesting, but not too surprising”.
He adds that the study “justifies the continued use of classical numerical weather models in operational forecasts, in spite of their tremendous computational costs”.
Advances in forecasting
The field of AI weather forecasting is evolving rapidly.
Olivetti notes that the three AI models tested in the study are an “older generation” of AI models. In the last two years, newer “probabilistic” forecast models have emerged that “claim to better capture extremes”, he explains.
The three AI models used in the analysis are “deterministic”, meaning that they only simulate one possible future outcome.
In contrast, study author Engelke tells Carbon Brief that probabilistic models “create several possible future states of the weather” and are therefore more likely to capture record-breaking extremes.
Engelke says it is “important” to evaluate the newer generation of models for their ability to forecast weather extremes.
He adds that this paper has set out a “protocol” for testing the ability of AI models to predict unprecedented extreme events, which he hopes other researchers will go on to use.
The study says that another “promising direction” for future research is to develop models that combine aspects of traditional, physics-based weather forecasts with AI models.
Engelke says this approach would be “best of both worlds”, as it would combine the ability of physics-based models to simulate record-breaking weather with the computational efficiency of AI models.
Dr Kyle Hilburn, a research scientist at Colorado State University, notes that the study does not address extreme rainfall, which he says “presents challenges for both modelling and observing”. This, he says, is an “important” area for future research.
The post Traditional models still ‘outperform AI’ for extreme weather forecasts appeared first on Carbon Brief.
Traditional models still ‘outperform AI’ for extreme weather forecasts
Climate Change
Minerals for the planet or for war and AI? COP31 must make clear what we prioritise
Elisa Morgera is the UN Special Rapporteur on Climate Change and Human Rights.
For years, governments and international financial institutions have justified the mining boom by arguing that the energy transition cannot happen without critical minerals.
Increasingly, however, they are securing those minerals for something else: weapons and artificial intelligence (AI).
At COP31, governments must make clear what a sustainable and fair energy transition requires and reject demand for minerals that fuels war and other climate-harming uses.
Independent research shows that mineral demand is increasingly driven by AI and militarisation. Yet, both undermine climate action.
Major armed conflicts are linked to fossil fuel expansion, generate emissions that are unaccounted for under the Paris Agreement, and deliver record profits to fossil fuel and petrochemical companies through price spikes and disrupted access to fuel and food.
Meanwhile, AI infrastructure risks undoing the progress of the renewable energy transition by driving fossil fuel expansion to meet data centres’ vast new energy demand.
UN Secretary-General António Guterres and the head of UN Climate Change Simon Stiell have both warned of the climate harm of unchecked, fossil fuel-powered AI and data centre growth.
In addition, AI equipment and infrastructure require huge amounts of copper – one of the minerals most in demand for the energy transition. A recent academic study found that copper makes up roughly 82-83% of the mineral mass in data centres.
None of this demand serves the energy transition. Instead, the climate case for mining is being used to legitimise uses that undermine a sustainable future.
Mining still relies heavily on fossil fuels and often harms the environment, so it must be limited to extracting the resources strictly necessary for the energy transition. Otherwise, states breach their binding obligations to protect the environment and human rights, as the International Court of Justice (ICJ) has made clear.
National security is the new mining fast track
Campaigners have long warned that when minerals become a matter of national security, governments are more tempted to deregulate and fast-track mining projects. Communities pay the price.
Extraction without adequate consultation is a global pattern linked to killings of land and environmental defenders. More than half of transition mineral projects are located on or near Indigenous lands, and allegations of abuses linked to mining transition minerals surged 73% in 2025 alone.
Security and climate are not separate problems. Both are rooted in the fossil fuel-based global economy, now joined by the unchecked growth of AI and data centres.
Who should benefit from minerals?
Not every demand for minerals can or should be met. Minerals are finite and must be used to address the worsening climate crisis as a priority.
The COP31 presidency’s push for electrification will have significant implications for mineral demand. At the same time, electrification doesn’t necessarily help to phase out fossil fuels. Governments must be clear about how minerals are being used and prioritise uses that genuinely advance the energy transition.
- Climate action comes first: Mineral demand for the wind and solar projects that power communities and essential public services such as local food systems should be met first. Priority uses should include community-led wind and solar projects and renewable power for public transport.
- No support for climate-harming activities: Mineral demand projections for the energy transition should exclude the military-industrial complex and data centres, and governments should reconsider public spending on militarisation. Mineral demand for electrification should also be scrutinised to avoid deepening fossil fuel dependence.
- Cut demand before expanding supply: High-consuming countries should use less primary material through efficiency, longer-lasting products and by reducing car dependence, rather than exporting waste and harm to the Global South.
- Prioritise energy needs: Those without access to energy must be served first. Overproduction and overconsumption must be curbed, otherwise the transition will neither mitigate climate change nor reduce poverty.
- No mining without consent: No project should receive financing or permits without the Free, Prior and Informed Consent of Indigenous Peoples and genuine consultations with affected communities.
- No-go zones: States should establish no-go mining zones in ecologically sensitive areas, on land and at sea, and ensure that mining does not cause significant harm to biodiversity.
What can COP31 do?
Last year, governments agreed to develop a Just Transition Mechanism, but explicit provisions on transition mineral governance were dropped. At COP31, negotiators are reconsidering the issue. I urge negotiators in Antalya to:
- Reflect states’ obligations underpinning the ICJ Advisory Opinion and the UN Secretary-General’s Critical Energy Transition Minerals Panel principles in the Just Transition Work Programme.
- Keep mineral demand for the military and digital infrastructure separate from demand that serves climate protection.
- Require traceability and end-use disclosure, so it is publicly known whether a mineral ends up in a battery, a missile or a data centre.
- Respect the human rights of Indigenous Peoples, peasants and environmental human rights defenders, including the right to say no, protection from reprisals and access to remedy.
At COP31, transition minerals should be unequivocally linked to everyone’s human right to a safe climate and a healthy environment.
The post Minerals for the planet or for war and AI? COP31 must make clear what we prioritise appeared first on Climate Home News.
Minerals for the planet or for war and AI? COP31 must make clear what we prioritise
Climate Change
Countries take up industry push to loosen UN carbon market rules at COP31
Dozens of countries – from big emerging economies to the world’s poorest nations – are calling for the UN’s new carbon market to reassess climate integrity rules, in a push that campaigners and researchers have blamed on industry lobbying.
At COP31 talks on the Paris Agreement Crediting Mechanism (PACM) this weekend, the Like-Minded Developing Countries – a group that includes Saudi Arabia, India and China – the Least Developed Countries (LDCs) and landlocked developing countries raised concerns about the requirements project developers must meet, according to sources familiar with the closed-door discussions.
They warned about the cumulative weight of those requirements developed under Article 6 of the Paris Agreement as barriers to implementation. Several countries, including Singapore and some Latin American states, said rules must keep projects financially viable, while South Korea asked for proportionate requirements, the sources said.
Their demands closely track arguments from carbon market industry groups, which claim PACM’s conservative approach to counting emission reductions is so strict that few projects will ever come to fruition.
Industry and NGOs lobby to weaken UN carbon credit rules in “coordinated” push
In a letter published ahead of COP31, the lobby groups asked governments to “raise implementation concerns” and to request that the Supervisory Body – the UN body that oversees the market – assess the “combined impact” of its rules on the feasibility of projects.
The letter was signed by the International Emissions Trading Association (IETA), a business group whose members include fossil fuel giants Shell, BP, Saudi Aramco and Chevron, along with the Project Developer Forum and the Wildlife Conservation Society, which develops forest conservation programmes.
Draft text reflects industry demands
Those demands are reflected in the first draft of the COP31 decision on the mechanism, published late on Saturday. To guard against over-crediting, PACM’s rules deliberately err on the side of issuing fewer credits when emissions cuts from projects are uncertain.
The document asks the Supervisory Body to look at how these cautious assumptions add up across its rules and what that means for the number of credits, project costs and prices. The aim, it says, is to avoid “excessive compounding of conservative adjustments”.
Other parts of the draft tell the body not to make its rules too cautious or too rigid. That includes efforts to guard against the risk that carbon captured by a project is later released. Most of these passages add that environmental integrity must be maintained.
Risk of another ‘failure’
“Weakening the rules is not climate ambition – it is a license to pollute,” said Erika Lennon, a senior attorney at the Center for International Environmental Law (CIEL). “As we are on the brink of permanently breaching 1.5C, we cannot afford to turn the carbon market into another loophole for polluters,” she added.
Axel Michaelowa, research director at Perspectives Climate Research, said that if attempts to weaken the rules succeed, it could lead to another failure of an international carbon market. Demand for credits rests on trust, he said, and the slightest doubt over integrity erodes buyers’ willingness to pay.
Calls to revise requirements “must be supported by scientific evidence, not unsupported claims of under-crediting – especially when these assumptions come from private actors whose revenues depend on the volume of transacted credits rather than real mitigation outcomes,” Michaelowa said.
The EU, Switzerland and Senegal are among a handful of countries pushing back against demands to re-assess the market rules, according to the sources, and are opposing requests to weigh up the profitability of projects or to lower standards.







