Please use this identifier to cite or link to this item:
On the limitations of deep learning for statistical downscaling of climate change projections : the transferability and the extrapolation issues
Title: On the limitations of deep learning for statistical downscaling of climate change projections : the transferability and the extrapolation issues
Authors: Hernanz Lázaro, Alfonso ORCID Autor AEMETCorrea, Carlos ORCID Autor AEMETSánchez Perrino, Juan CarlosAutor AEMETPrieto-Rico, IgnacioRodríguez Guisado, EstebanAutor AEMETDomínguez Alonso, Marta ORCID Autor AEMETRodríguez Camino, Ernesto ORCID Autor AEMET
Keywords: Convolutional neural networks; Deep learning; Emulators; Evaluation; EURO-CORDEX; Pseudo-reality; Extrapolation; Statistical downscaling; Regional climate models
Issue Date: 2023
Publisher: Royal Meteorological Society; John Wiley & Sons
Citation: Atmospheric Science Letters. 2023, e1195
Publisher version:
Abstract: Convolutional neural networks (CNNs) have become one of the state-of-the-art techniques for downscaling climate projections. They are being applied under Perfect-Prognosis (trained in a historical period with observations) and hybrid approaches (as Regional Climate Models (RCMs) emulators), with satisfactory results. Nevertheless, two important aspects have not been, to our knowledge, properly assessed yet: (1) their performance as emulators for other Earth System Models (ESMs) different to the one used for training, and (2) their performance under extrapolation, that is, when applied outside of their calibration range. In this study, we use UNET, a popular CNN, to assess these two aspects through two pseudo-reality experiments, and we compare it with simpler emulators: an interpolation and a linear regression. The RCA4 regional model, with 0.11° resolution over a complex domain centered in the Pyrenees, and driven by the CNRM-CM5 global model is used to train the emulators. Two frameworks are followed for the training: predictors are taken (1) from the upscaled RCM and (2) from the ESM. In both frameworks, the performance of the UNET when applied for other ESMs different to the one used for training is considerably worse, indicating poor generalization. For the linear method a similar deterioration is seen, so this limitation does not seem method specific but inherent to the task. For the second experiment, the emulators are trained in present and evaluated in future, under extrapolation. While averaged aspects such as the mean values are well simulated in future, significant biases (up to 5°C) appear when assessing warm extremes. These biases are larger by UNET than those produced by the linear method. This limitation suggests that, for variables such as temperature, with a marked signal of change and a strong linear relationship with predictors, simple linear methods might be more appropriate than the sophisticated deep learning techniques.
ISSN: 1530-261X
Appears in Collections:Artículos científicos 2023-2026

Files in This Item:
  File Description SizeFormat 
2,44 MBAdobe PDFThumbnail
Show full item record

Items in Arcimis are protected by Creative Commons License, unless otherwise indicated.

Arcimis Repository
Nota Legal Contacto y sugerencias