KLIM-QML: Quantum Machine Learning for Improved Climate Modeling
As part of the DLR Quantum Computing Initiative, planqc and d-fine develop quantum machine learning models to enhance climate simulations. The KLIM-QML project aims to reduce forecast uncertainties—laying the foundation for more effective climate strategies in aviation, energy, and maritime sectors.
The Challenge
Climate models are critical tools for understanding and mitigating climate change. The main challenge addressed by KLIM-QML is that climate models in general need to use coarse grids for describing the atmosphere. Without this simplification, it would be impossible to propagate these models far enough into the future. However, it is known and well understood that these grids are too coarse, which introduces errors and uncertainties, especially in capturing small-scale processes such as cloud formation, turbulence, or regional weather effects.
Our Approach
Recently, machine learning (ML) models have been investigated to compensate for these shortcomings. The aim of KLIM-QML is to investigate whether quantum machine learning (QML) can do even better than classical ML.
The project investigates novel QML techniques for parameterizing sub-grid processes, applies hybrid quantum-classical algorithms to optimize the tuning of existing models, and explores quantum-inspired methods to compress massive climate datasets. All solutions are benchmarked against classical methods. By doing so, the consortium systematically evaluates where and when quantum computing could deliver tangible advantages for climate science.
Quantum neural networks can harness exponentially dense frequency spectra and therefore allow for a highly efficient representation of machine learning model functions. In the scope of KLIM-QML, we are investigating how such models and their variations can be utilized to help climate scientists to simulate weather and climate for a more sustainable future.
The Potential
Quantum-enhanced climate models could reduce forecast uncertainties, support more reliable risk assessments for extreme weather, and enable actionable strategies for emission reduction in aviation, energy, and shipping. Beyond climate, KLIM-QML could demonstrate how QML can be integrated into industrial workflows—offering a blueprint for other sectors facing data- and simulation-intensive challenges.
Quantum machine learning could become an important tool for improving the accuracy of climate models. With KLIM-QML, we are working with DLR and strong partners to explore where quantum methods can add real value and help make climate predictions more reliable for practical use.
Project Lead
-
Felix Herbort
Quantum Algorithm Engineer