Guest Author
Mierk Schwabe
Physicist, DLR
Introduction
When my coauthors and I started working on our paper, Opportunities and challenges of quantum computing for climate modeling, we had a specific goal in mind: to take a critical look at how quantum computing could contribute to one of the most pressing scientific and societal challenges of our time, accurately predicting climate change also on a regional scale for better adaptation measures and technology assessments.
Despite significant progress in Earth System Models (ESMs), key challenges remain, for example in resolution, the representation of unresolved processes with so-called parameterizations, and the resources required for tuning the models. Classical machine learning is already demonstrating great potential to advance climate modelling. Tools like neural network-based subgrid-scale parameterizations incorporated into hybrid artificial intelligence (AI)–physics models are being tested and started to be deployed in operational systems.
So why quantum? Because we believe that quantum computing could potentially complement and enhance this progress, especially in areas where classical methods remain computationally constrained or where data is sparse. Keeping in mind the rapid progress of quantum computing and the urgency of the problem due to progressing climate change, we would like to stimulate more cross-disciplinary collaborations between climate and quantum scientists, preparing to take advantage of coming large-scale quantum computers.
This blog post shares what motivated us, what we think quantum computing can realistically offer climate science, and where we believe collaboration across disciplines will be most fruitful.
Data-driven climate modeling
In the past few years, there’s been a surge of interest in AI-driven climate modeling. Researchers are using machine learning to:
- Improve the representation of unresolved physics, such as cloud formation or turbulence,
- Emulate expensive model components (like radiation schemes),
- Calibrate the models,
- Perform fast surrogate simulations for scenario exploration and ensemble forecasts, and
- Downscale the output with machine learning tools for regional climate information.
These advances are not only accelerating parts of the workflow—they’re also improving model fidelity in some regimes. Deep learning tools can now capture nonlinear dynamics, maintain conservation properties, and in some cases even generalize to out-of-sample forcing conditions.
But there are limits. These models still rely heavily on massive data availability, demand huge training resources, and often lack interpretability. Their integration into full ESMs remains challenging, due to concerns about stability, generalization properties, and physical consistency.
That’s where we think quantum computing might come in—not to replace classical ML, but to complement it.
Where we see quantum making a difference?
Fig. 1: Overview of the climate modeling tasks and category of matching quantum computing algorithms. (Image of the Earth by NASA/Apollo 17, Figure from Schwabe et al, 2025)
In our paper, we identify four core areas of potential quantum contribution. Many ESM components depend on solving partial differential equations. In theory, quantum linear solvers offer exponential speedups under specific conditions. These solvers, or quantum machine learning based models such as qPINNs, may one day accelerate the calculations inside ESM cores. This could enable finer spatial resolution or faster ensemble generation.
We’re particularly intrigued by the potential of quantum machine learning (QML) to learn to represent subgrid-scale processes under data-limited conditions. Quantum neural networks could offer improved expressivity or better generalization capabilities, even using current noisy intermediate-scale quantum (NISQ) devices, complementing classical machine learning approaches.
ESMs have various free parameters, which ideally are recalibrated after each model change. This tuning involves searching high-dimensional, multimodal parameter spaces, which could be enhanced with quantum computing, eventually helping find good parameter configurations faster and with fewer computational resources.
And why this is so challenging?
Climate modelling is among the most complex real-world applications one could attempt to tackle with quantum tools. Fault-tolerant systems needed for many climate applications are years away. Today’s NISQ devices can only support small-scale QML or optimization prototypes, and even those require careful circuit design to avoid barren plateaus and excessive noise. Also, many theoretical quantum algorithms assume mathematical properties that don’t necessarily hold in real climate workflows. Furthermore, the coupling between state-of-the-art supercomputers and quantum devices which is needed for most applications is challenging.
One of our biggest takeaways: meaningful progress needs climate scientists and quantum researchers to work closely together. The models, the data, the constraints—they all have domain-specific structures that quantum teams need to understand deeply. Likewise, climate teams need to be part of algorithm design from the beginning.
What we recommend?
So where do we go from here? Quantum climate modelling is not just around the corner—but there are clear next steps:
- Use simple, well-defined toy problems from climate physics (e.g., Lorenz-96) to develop and test quantum algorithms in a targeted way.
- Build hybrid AI-quantum models, where quantum components work in tandem with a classical ML framework.
- Keep climate modeling in mind for hardware-software co-design and coupling HPC and quantum devices.
- Invest in co-training programs for scientists working across quantum computing, atmospheric physics, and AI.
Quantum computing has potential, but needs real-world, societally relevant testbeds, and climate modeling needs rapid progress together with high-resolution modeling and classical machine learning. There’s a real opportunity here to build meaningful, high-impact collaborations!
Read the paper
Mierk Schwabe, Lorenzo Pastori, Inés de Vega, Pierre Gentine, Luigi Iapichino, Valtteri Lahtinen, Martin Leib, Jeanette Miriam Lorenz, Veronika Eyring: Opportunities and challenges of quantum computing for climate modelling. Published in Environmental Data Science (2025; 4: e35) https://doi.org/10.1017/eds.2025.10010`