Downscaling multi-model climate projection ensembles with deep learning (DeepESD): contribution to CORDEX EUR-44

Abstract

Abstract. Deep learning (DL) has recently emerged as an innovative tool to downscale climate variables from large-scale atmospheric fields under the perfect-prognosis (PP) approach. Different convolutional neural networks (CNNs) have been applied under present-day conditions with promising results, but little is known about their suitability for extrapolating future climate change conditions. Here, we analyze this problem from a multi-model perspective, developing and evaluating an ensemble of CNN-based downscaled projections (hereafter DeepESD) for temperature and precipitation over the European EUR-44i (0.5∘) domain, based on eight global circulation models (GCMs) from the Coupled Model Intercomparison Project Phase 5 (CMIP5). To our knowledge, this is the first time that CNNs have been used to produce downscaled multi-model ensembles based on the perfect-prognosis approach, allowing us to quantify inter-model uncertainty in climate change signals. The results are compared with those corresponding to an EUR-44 ensemble of regional climate models (RCMs) showing that DeepESD reduces distributional biases in the historical period. Moreover, the resulting climate change signals are broadly comparable to those obtained with the RCMs, with similar spatial structures. As for the uncertainty of the climate change signal (measured on the basis of inter-model spread), DeepESD preserves the uncertainty for temperature and results in a reduced uncertainty for precipitation. To facilitate further studies of this downscaling approach, we follow FAIR principles and make publicly available the code (a Jupyter notebook) and the DeepESD dataset. In particular, DeepESD is published at the Earth System Grid Federation (ESGF), as the first continental-wide PP dataset contributing to CORDEX (EUR-44).

Downscaling Multi-Model Climate Projection Ensembles with Deep Learning (DeepESD): Contribution to CORDEX EUR-44

“…Code and data availability. To promote transparency and reproducibility of our results, we provide the data (DOI: https://doi.org/10.5281/zenodo.6823421, Baño-Medina et al, 2022a) and the companion Jupyter notebook (DOI: https://doi.org/10.5281/zenodo.6828303, Baño-Medina et al, 2022b), explaining how DeepESD has been produced. This notebook is based on the R software and builds on the climate4R framework, a set of libraries specifically designed for climate data access and post-processing (Iturbide et al, 2019).…”

Using Explainability to Inform Statistical Downscaling Based on Deep Learning Beyond Standard Validation Approaches

“…Taking into account the PP assumption that GCM predictors have to be realistically simulated (when compared with the ERA-Interim reanalysis in this case), following (Baño-Medina et al, 2022) we perform a signal-preserving adjustment of the monthly mean and variance of the GCM predictors to increase the distributional similarity with their counterpart reanalysis predictor fields.…”

“…However, there is no best model outperforming the others in all regions for all scores and, overall, there is no reason to discard any of these models. To test the extrapolation capability of the CNN models under future climate change conditions (when applied to predictors from GCM projections; see Section 2.1), we follow previous work and use the "raw" GCM projections as pseudo-reality (Vrac et al, 2007;Baño-Medina et al, 2022). We divide the future scenario into three different periods (2006-2040, 2041-2070 and 2071-2100) and compute the delta change between the future and historical scenarios for the GCM and CNN models.…”

“…Different studies have examined the performance of these methods for downscaling in both present (Pan et al, 2019;Baño-Medina et al, 2020;Sun & Lan, 2021) and future climate conditions (Baño-Medina et al, 2021, with successful results. Moreover, these advances have led to the production of continental-wide projections following the CORDEX protocol constituting the first continental-wide CORDEX contribution with PP statistical downscaling methods (Baño-Medina et al, 2022). However, the complexity and black-box nature of CNNs raise some concerns about their widespread application to generate actionable downscaled results.…”

Using Machine Learning to Cut the Cost of Dynamical Downscaling

“…Compared to DD, the implementation of SD is fast and far less computationally intensive (Wang, Liu, et al., 2021). However, SD models can perform poorly under extrapolation to future climates as few methods account for nonstationary relationships between predictors and predictands under climate change (Hernanz et al., 2022; Hewitson et al., 2014; Lanzante et al., 2018; Salvi et al., 2016; Schoof, 2013), although there are some exceptions (Baño‐Medina et al., 2022; Pichuka & Maity, 2018). Despite the fairly low cost of SD, it is only possible to implement it in regions where fine‐scale observational data are available for training, and typically little is known about its ability to perform downscaling in regions outside of the training domain (Wang, Tian, et al., 2021).…”