Search

Deep learning (DL) methods have recently garnered attention from the climate change community for being an innovative approach to downscaling climate variables from Earth system and global climate models (ESGCMs) with horizontal resolutions still too coarse to represent regional- to local-scale phenomena. In the context of the Coupled Model Intercomparison Project phase 6 (CMIP6), ESGCM simulations were conducted for the Sixth Assessment Report (AR6) of the Intergovernmental Panel on Climate Change (IPCC) at resolutions ranging from 0.70 to 3.75 ∘. Here, four convolutional neural network (CNN) architectures were evaluated for their ability to downscale, to a resolution of 0.1 ∘ , seven CMIP6 ESGCMs over the Iberian Peninsula – a known climate change hotspot, due to its increased vulnerability to projected future warming and drying conditions. The study is divided into three stages: (1) evaluating the performance of the four CNN architectures in predicting mean, minimum, and maximum temperatures, as well as daily precipitation, trained using ERA5 data and compared with the Iberia01 observational dataset; (2) downscaling the CMIP6 ESGCMs using the trained CNN architectures and further evaluating the ensemble against Iberia01; and (3) constructing a multi-model ensemble of CNN-based downscaled projections for temperature and precipitation over the Iberian Peninsula at 0.1 ∘ resolution throughout the 21st century under four Shared Socioeconomic Pathway (SSP) scenarios. Upon validation and satisfactory performance evaluation, the DL downscaled projections demonstrate overall agreement with the CMIP6 ESGCM ensemble in magnitude for temperature projections and sign for the projected temperature and precipitation changes. Moreover, the advantages of using a high-resolution DL downscaled ensemble of ESGCM climate projections are evident, offering substantial added value in representing regional climate change over Iberia. Notably, a clear warming trend is observed in Iberia, consistent with previous studies in this area, with projected temperature increases ranging from 2 to 6 ∘ C, depending on the climate scenario. Regarding precipitation, robust projected decreases are observed in western and southwestern Iberia, particularly after 2040. These results may offer a new tool for providing regional climate change information for adaptation strategies based on CMIP6 ESGCMs prior to the next phase of the European branch of the Coordinated Regional Climate Downscaling Experiment (EURO-CORDEX) experiments. [ABSTRACT FROM AUTHOR]

Deep learning (DL) methods have recently garnered attention from the climate change community, as an innovative approach for downscaling climate variables from Earth System and Global Climate Models (ESGCMs) with horizontal resolutions still too coarse to represent regional-to-local-scale phenomena. In the context of the Coupled Model Intercomparison Project phase 6 (CMIP6), ESGCMs simulations were conducted for the Sixth Assessment Report (AR6) of the Intergovernmental Panel on Climate Change (IPCC), at resolutions ranging from 0.70° to 3.75°. Here, four Convolutional Neural Network (CNN) architectures were evaluated for their ability to downscale, to a resolution of 0.1°, seven CMIP6 ESGCMs over the Iberian Peninsula - a known climate change hotspot, due to its increased vulnerability to projected future warming and drying conditions. The study is divided into three stages: (1) evaluating the performance of the four CNN architectures in predicting mean, minimum, and maximum temperatures, as well as daily precipitation, trained using ERA5 data, and compared with the Iberia01 observational dataset; (2) downscaling the CMIP6 ESGCMs using the trained CNN architectures and further evaluating the ensemble against Iberia01; and (3) constructing a multi-model ensemble of CNNbased downscaled projections for temperature and precipitation over the Iberian Peninsula at 0.1° resolution throughout the 21st century, under four Shared Socioeconomic Pathway (SSP) scenarios. Upon validation and satisfactory performance evaluation, the DL downscaled projections demonstrate overall agreement with the CMIP6 ESGCM ensemble in terms of temperature and precipitation projections. Moreover, the advantages of using a high-resolution DL downscaled ensemble of ESGCM climate projections are evident, offering substantial added value in representing regional climate change over Iberia. Notably, a clear warming trend is observed, consistent with previous studies in this area, with projected temperature increases ranging from 2°C to 6°C depending on the climate scenario. Regarding precipitation, robust projected decreases are observed in western and southwestern Iberia, particularly after 2040. These results may offer a new tool for providing regional climate change information for adaptation strategies based on CMIP6 ESGCMs prior to the next phase of the European branch from the Coordinated Regional Climate Downscaling Experiment (EURO-CORDEX) experiments. [ABSTRACT FROM AUTHOR]

As cities encompass most of the global population, it is crucial to understand the effects of climate change in an urban context to develop tailored adaptation and mitigation strategies. Physically-based numerical models are computationally demanding and present scale limitations that complicate the representation of cities and their effects on the regional-to-local climate and vice-versa. Therefore, often, methodologies (e.g. statistical downscaling) are sought to complement the output of numerical models for urban climate studies. Deep Learning (DL) is a growing technology that has become a universal presence in society and the scientific community, geosciences included, showing promising results all around. In this study, we applied DL models, namely Convolutional Neural Networks (CNNs), to downscale land surface temperature (LST) and predict 2-m maximum and minimum temperatures (T2m max and T2m min , respectively) over Paris between 2004 and 2022, and compared the results with ERA5, the most recent atmospheric reanalysis of the European Centre for Medium Range Weather Forecasts. Several experiments featuring different sets of ERA5 predictors were used as input data to the DL models. Afterwards, the quality of the DL models in representing the urban heat island (UHI) over Paris was assessed. Our results showed substantial improvements in LST, T2m and UHI downscaling with DL (using a small number of predictors) in comparison to ERA5. Particularly, the best-performing DL experiments presented nighttime UHI and daytime SUHI biases (RMSE) below 0.80 °C and 0.50 °C (2.8 °C and 2.3 °C), respectively. This study supports the potential of using DL as a downscaling technique to help improve the representation of temperature extremes in an urban context in the historical period. • Deep Learning (DL) accurately portrays the urban heat island in Paris. • Downscaling land surface temperature with DL shows solid results using a small set of predictors. • Spatialization of 2-m temperature presents similar performance regardless of the predictors used to train the DL models. • Downscaling with DL show overall improvements in comparison with ERA5. [ABSTRACT FROM AUTHOR]