Evaluating the transferability potential of deep learning models for climate downscaling
Ayush Prasad, Paula Harder, Qidong Yang, Prasanna Sattegeri, Daniela Szwarcman, Campbell Watson, David Rolnick
TL;DR
This work addresses the limited generalizability of deep learning for climate downscaling by pre-training on multiple, diverse climate datasets and evaluating zero-shot transfer across spatial, variable, and product transfer tasks. It compares CNNs, Fourier Neural Operators, and CNN–ViT hybrids, showing that a CNN–ViT hybrid often provides the strongest zero-shot spatial and product transfer, while Fourier-based models excel in transferring to unseen variables. Across two-simulation scenarios, zero-shot transfer is limited, but fine-tuning on target-like data substantially improves accuracy, highlighting the value of domain adaptation. Overall, pre-training on diverse data combined with selective fine-tuning emerges as an effective strategy to enhance transferability in climate downscaling with deep learning.
Abstract
Climate downscaling, the process of generating high-resolution climate data from low-resolution simulations, is essential for understanding and adapting to climate change at regional and local scales. Deep learning approaches have proven useful in tackling this problem. However, existing studies usually focus on training models for one specific task, location and variable, which are therefore limited in their generalizability and transferability. In this paper, we evaluate the efficacy of training deep learning downscaling models on multiple diverse climate datasets to learn more robust and transferable representations. We evaluate the effectiveness of architectures zero-shot transferability using CNNs, Fourier Neural Operators (FNOs), and vision Transformers (ViTs). We assess the spatial, variable, and product transferability of downscaling models experimentally, to understand the generalizability of these different architecture types.