Quantum Computing for Climate Modelling

The rapid maturation of quantum hardware signals that the climate modelling community must position itself to exploit these emerging capabilities, taking advantage of the rapid progress to improve the projection of the changing climate for better climate mitigation and adaptation options. Earth system models (ESMs) are constrained by three bottlenecks: (i) the costly numerical integration of partial differential equations (PDEs) that describe atmospheric and oceanic dynamics, (ii) the reliance on physics‑based subgrid-scale parameterizations to capture the influence of complex processes such as convection and cloud formation, and (iii) the labor‑intensive, often manual calibration of the free parameters belonging to these parameterizations.

In collaboration with the industry contractors d‑fine and planqc, we present three examples of potential quantum-enhancements for climate modelling: quantum physics-informed neural networks (qPINNs) for PDE solving, quantum‑assisted model calibration, and quantum machine learning (QML)‑based parameterizations. First, qPINNs use variational quantum circuits to solve PDEs taking into account measurements, boundary conditions etc. Second, quantum computing potentially enhances the calibration loop: trained quantum emulators could replace expensive model runs. Third, we showcase how QML-based parameterizations could replace physics‑based subgrid schemes by training on high‑resolution cloud‑resolving simulation data, robustly learning physically plausible relationships.

Finally, we propose a roadmap towards taking advantage of the rapid progress of quantum hardware, including scaling studies on HPC‑quantum hybrid architectures, thereby establishing a pathway for quantum‑augmented ESMs.