Abstract
Understanding and estimating regional climate change under different anthropogenic emission scenarios is pivotal for informing societal adaptation and mitigation measures. However, the high computational complexity of state-of-the-art climate models remains a central bottleneck in this endeavour. Here we introduce a machine learning approach, which utilises a unique dataset of existing climate model simulations to learn relationships between short-term and long-term temperature responses to different climate forcing scenarios. This approach not only has the potential to accelerate climate change projections by reducing the costs of scenario computations, but also helps uncover early indicators of modelled long-term climate responses, which is of relevance to climate change detection, predictability, and attribution. Our results highlight challenges and opportunities for data-driven climate modelling, especially concerning the incorporation of even larger model datasets in the future. We therefore encourage extensive data sharing among research institutes to build ever more powerful climate response emulators, and thus to enable faster climate change projections.
Highlights
To achieve long-term climate change mitigation and adaptation goals, such as limiting global warming to 1.5 or 2 °C, there must be a global effort to decide and act upon effective but realistic emission pathways[1]
Exploring more detailed relationships between emissions and multiregional climate responses still requires the application of Global Climate Models (GCMs) that allow the behaviour of the climate to be simulated under various conditions[10,11,12] on decadal to multi-centennial timescales
We evaluate the performance of the two different machine learning approaches where smaller scale patterns are more learning methods (Ridge, Gaussian Process Regression (GPR)) by benchmarking them against a traditional pattern scaling approach[36,39], often used for estimating future patterns of climate change[40,41,42]
Summary
To achieve long-term climate change mitigation and adaptation goals, such as limiting global warming to 1.5 or 2 °C, there must be a global effort to decide and act upon effective but realistic emission pathways[1]. Our statistical model approximates the behaviour of the full GCM for a specific target climate variable of interest; here we choose surface temperature at each GCM grid cell, a central variable of interest in climate science and impact studies This model is trained on simulations from the full global climate model (supervised learning35), in order to predict the long-term surface temperature response of the GCM from the short-term temperature responses to perturbations (Fig. 1c). We choose Ridge regression and GPR, because these two methods handle well the limited sample size (number of simulations available) for training, which limits how effectively the number of free parameters for other approaches such as deep learning, including convolutional neural networks, could be Fig. 2 Distribution of predicted grid-point scale surface temperaconstrained. We highlight that gridscale error metrics need to be interpreted with care because they can present misleading results, for higher resolution
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
