José Manuel Gutiérrez

Deep learning (DL) models are increasingly used for climate downscaling, yet the impact of observational uncertainty on their performance remains poorly understood. Here we assess this effect by training ensembles of DL downscaling models using four independent gridded observational data sets over continental Spain that differ in source, methodology, and spatial coverage. Each ensemble includes multiple training realizations of the same model to capture training variability. The models are then applied to downscale global climate model temperature projections. Results show that both intra‐model (across realizations) and inter‐dataset (across observational products) variability substantially influence the downscaled climate change signal. Differences among data sets dominate this uncertainty, especially for extremes (annual 99th percentile) at regional scales, leading to warming discrepancies of up to 2.5°C. However, these discrepancies are smaller for high‐quality regional data sets built from dense observation networks and rigorous quality control, whereas global observational data sets, known to misrepresent local climate features (Bedia et al., 2013, https://doi.org/10.1016/j.gloplacha.2013.04.005 ), exert a larger impact on downscaled projections. These findings highlight that the use of well‐constructed regional data sets can substantially reduce uncertainty and improve the reliability of deep learning–based downscaling. Explicitly incorporating observational uncertainty into training and evaluation frameworks remains essential to ensure robust and credible regional climate projections.

Spain's geographical position and climate characteristics make it highly vulnerable to climate change–driven projected aridity change. This study assesses the projected evolution of climatic aridity across mainland Spain, the Balearic Islands, and the Canary Islands throughout the 21st century. Using the UNEP Aridity Index ( AI = P / ETo ), we analyse an ensemble of high‐resolution CMIP6 simulations under four Shared Socioeconomic Pathways (SSPs) and four Global Warming Levels (GWLs), relative to the 1985–2014 baseline. The CMIP6 models exhibit a systematic dry bias, characterised by underestimated precipitation ( P ) and overestimated reference evapotranspiration ( ETo ) over the Spanish mainland, leading to artificially low historical AI values and excessively steep ETo trends. To mitigate this bias, we apply a delta‐change ( Δ AI ) method, which corrects the climatological baseline but does not eliminate process‐level uncertainties. Despite these limitations, the ensemble shows a consistent directional signal of increasing aridity across all regions and scenarios. The magnitude of projected aridity change scales with global warming and emissions intensity, reflecting a clear dose–response pattern. This trend is largely attributed to rising ETo , although its physical realism remains uncertain given the models' underestimation of precipitation and the use of empirical models to estimate ETo . Under high‐emission scenarios (SSP3‐7.0 and SSP5‐8.5), arid and semi‐arid zones expand across southern Spain, whilst the Canary Islands experience a pronounced shift towards hyper‐arid conditions. Overall, the results indicate that Spain is likely to face a progressive and spatially heterogeneous increase in climatic aridity, though the mechanistic basis of this change remains uncertain. These findings highlight both the directional robustness of projected aridity change and the need for more physically grounded model evaluation before fully translating these projections into adaptation strategies.

Deep learning (DL) models are increasingly used for climate downscaling, yet the impact of observational uncertainty on their performance remains poorly understood. Here we assess this effect by training ensembles of DL downscaling models using four independent gridded observational datasets over continental Spain that differ in source, methodology, and spatial coverage. Each ensemble includes multiple training realizations of the same model to capture training variability. The models are then applied to downscale global climate model temperature projections. Results show that both intra-model (across realizations) and inter-dataset (across observational products) variability substantially influence the downscaled climate change signal. Differences among datasets dominate this uncertainty, especially for extremes (annual 99th percentile) at regional scales, leading to warming discrepancies of up to 2.5 Â{degree sign}C. However, these discrepancies are smaller for high-quality regional datasets built from dense observation networks and rigorous quality control, whereas global observational datasets, known to misrepresent local climate features \cite{bedia_dangers_2013} , exert a larger impact on downscaled projections. These findings highlight that the use of well-constructed regional datasets can substantially reduce uncertainty and improve the reliability of deep learning—based downscaling. Explicitly incorporating observational uncertainty into training and evaluation frameworks remains essential to ensure robust and credible regional climate projections.

The Sixth Assessment Report (AR6) of the IPCC highlights the regionally varying impacts of climate change on both mean and extreme conditions, with significant socio-economic impacts. A key innovation of AR6 is the Interactive Atlas, which provides spatial and temporal analysis of over 20 Climatic Impact-Drivers (CIDs), integrating global and regional projection datasets from the CMIP and CORDEX model intercomparison projects. The Interactive Atlas allows users to explore, compare and download products relevant to regional climate change risk assessment, while adhering to FAIR principles to ensure data accessibility and reusability. This paper describes additional work undertaken to further support these principles, acknowledging the challenges of processing large-scale datasets. We present a cloud-enabled data laboratory—DataLab— developed using R and Python frameworks for advanced analysis and visualization. The DataLab seamlessly integrates the gridded dataset underpinning the Interactive Atlas with computational resources, combining traditional remote data access services and modern cloud-based solutions to balance cost-effectiveness with technological advances. This comprehensive resource supports regional climate change assessment and could be used to inform national adaptation plans and strengthen climate change research and policy development.

Pierce’s Disease (PD) is a vector-borne disease caused by the bacterium Xylella fastidiosa , which poses a significant threat to grapevines worldwide. Despite its importance, the risk of future PD establishment in Europe remains unclear due to previous incomplete methodologies followed by conflicting results. Here we present a comprehensive approach considering the compound effect of climate change on the pathosystem. Within the general trend of progressively increasing PD risk, we identified the +3ºC scenario as a turning point for potential spreading beyond Mediterranean regions, representing a serious risk for French and Italian viticulture. Our innovative methodology reveals PD risk as a multi-factor multi-scale process, showing contrasting spatial patterns and different risk velocities across regions, as well as a high timing uncertainty. By overcoming previous limitations, our findings contribute to a better understanding of the potential spread of PD in Europe, supporting informed decision-making for disease management and prevention.

Assessing the regional responses to different Global Warming Levels (GWLs; e.g. + 1.5, 2, 3 and 4 ºC) is one of the most important challenges in climate change sciences since the Paris Agreement goal of keeping global temperature increase well below 2 °C with respect to the pre-industrial period. Regional responses to global warming were typically analyzed using global projections from Global Climate Models (GCMs) and, more recently, using higher resolution Regional Climate Models (RCMs) over limited regions. For instance, the IPCC AR6 WGI Atlas provides results of the regional response to different GWLs for several climate variables from both GCMs and RCMs. These results are calculated under the assumption that the regional signal to global warming is consistent between the GCMs and the nested RCMs. In the present study we investigate the above assumption by evaluating the consistency of regional responses to global warming from global (CMIP5) and regional (CORDEX) projections. The dataset aggregated over the new IPCC reference regions, available from the IPCC AR6 WGI Atlas repository, is analyzed here for temperature and precipitation. The existing relationships between the regional climate change signals and global warming are compared for both CMIP5 and CORDEX. Our results show significant linear scaling relationships between regional changes and global warming for most of the regions. CORDEX and CMIP5 show remarkably similar scaling relationships and similar robustness in the emergence of the climate change signal for most of the regions. These results support the use of regional climate models in the context of global warming level studies.

The collaboration between the Coordinated Regional Climate Downscaling Experiment (CORDEX) and the Earth System Grid Federation (ESGF) provides open access to an unprecedented ensemble of regional climate model (RCM) simulations, across the 14 CORDEX continental-scale domains, with global coverage. These simulations have been used as a new line of evidence to assess regional climate projections in the latest contribution of the Working Group I (WGI) to the IPCC Sixth Assessment Report (AR6), particularly in the regional chapters and the Atlas. Here, we present the work done in the framework of the Copernicus Climate Change Service (C3S) to ­assemble a consistent worldwide CORDEX grand ensemble, aligned with the deadlines and ­activities of IPCC AR6. This work addressed the uneven and heterogeneous availability of CORDEX ESGF data by supporting publication in CORDEX domains with few archived simulations and performing quality control. It also addressed the lack of comprehensive documentation by compiling information from all contributing regional models, allowing for an informed use of data. In addition to presenting the worldwide CORDEX dataset, we assess here its consistency for precipitation and temperature by comparing climate change signals in regions with overlapping CORDEX domains, obtaining overall coincident regional climate change signals. The C3S CORDEX dataset has been used for the assessment of regional climate change in the IPCC AR6 (and for the interactive Atlas) and is available through the Copernicus Climate Data Store (CDS).

The Sixth Assessment Report (AR6) of the Intergovernmental Panel on Climate Change (IPCC) has adopted the FAIR Guiding Principles. We present the Atlas chapter of Working Group I (WGI) as a test case. We describe the application of the FAIR principles in the Atlas, the challenges faced during its implementation, and those that remain for the future. We introduce the open source repository resulting from this process, including coding (e.g., annotated Jupyter notebooks), data provenance, and some aggregated datasets used in some figures in the Atlas chapter and its interactive companion (the Interactive Atlas), open to scrutiny by the scientific community and the general public. We describe the informal pilot review conducted on this repository to gather recommendations that led to significant improvements. Finally, a working example illustrates the re-use of the repository resources to produce customized regional information, extending the Interactive Atlas products and running the code interactively in a web browser using Jupyter notebooks.