CORDEX Task Force report on proposed actions, steps and timelines in preparation for CORDEX-CMIP7

CORDEX Task Force report on proposed actions, steps and timelines in preparation for CORDEX-CMIP7 2025−12−05 Jesus Fernandez1 , Lars Buntemeyer 2, Grigory Nikulin 3, Seth McGinnis 4, ClaasTeichmann 2, Tereza Cavazos5, SwenBrands1, Pierre Nabat6, Dominique Paquin7, Marcus Thatcher8, AlessandroDosio9, Melissa Bukovsky10, Jason P. Evans11 Affiliations 1 Instituto de Física de Cantabria (IFCA), CSIC-Universidad de Cantabria, Santander, Spain 2 Helmholtz-Zentrum Hereon, Hamburg, Germany 3 Swedish Meteorological and Hydrological Institute, Norrkoping, Sweden 4 NSF National Center for Atmospheric Research, Boulder, CO, USA 5 Center for Scientific Research and Higher Education of Ensenada (CICESE), Baja California, Mexico 6 CNRM, Université de Toulouse, Météo-France, CNRS, Toulouse, France 7 Ouranos, Canada 8 Commonwealth Scientific and Industrial Research Organisation (CSIRO), Melbourne, Australia 9 European Commission Joint Research Centre, Ispra, Italy 10 University of Wyoming, Laramie, WY, USA 11 Climate Change Research Centre, University of New South Wales, Sydney, Australia Acknowledgements S. McGinnis’s work was supported by the NSF National Center for Atmospheric Research, which is a major facility sponsored by the U.S. National Science Foundation under Cooperative Agreement No. 1852977. It was also supported by the US Department of Energy Regional and Global Climate Modelling program award DOE DE-SC0016605. S. Brands and J. Fernandez acknowledge support by the project ATLAS2 (PID2024−162703OB-I00) funded by MCIN/AEI/10.13039/501100011033. 1 Table of contents Table of contents 2 List of acronyms 3 Document version history 4 1. Introduction 5 2. What is new in CMIP7? 7 3. How have CORDEX modelling approaches evolved since CORDEX-CMIP6? 11 4. Lessons learned from CORDEX-CMIP6 14 5. Planning steps and specific actions 15 5.1 CMIP7 Data request (DR) 15 5.2 Experiment design 16 5.3 Driving GCM selection 21 5.4 Monitoring simulations 23 5.5 Data request 24 5.6 Archiving specifications 27 5.7 Model documentation 30 5.8 Quality Control 32 5.9 User guidelines 34 5.10 Data publication 35 5.11 Errata 39 5.12 Monitoring the science 40 6. Links to other task forces 42 6.1 CORDEX-CORE 42 6.2 Convection-permitting modelling 43 6.3 Ocean 45 6.4 Machine Learning 46 7. Timeline 47 References 49 Appendix 1: Survey on potential commitments for CORDEX-CMIP7 in time for AR7 52 Appendix 2: Survey on climate forcing data used in CORDEX-CMIP6 60 Appendix 3: Proposal for an upcoming Working Group on CORDEX Infrastructure 63 2 List of acronyms AFT CMIP7 Assessment Fast Track AR7 IPCC 7th Assessment Report AWS Amazon Web Services CDNOT CMIP Data Node Operations Team CEDA Centre for Environmental Data Analysis, UK CMIP Coupled Model Intercomparison Project CPM Convection-Permitting Modelling (e.g. as in TF-CPM) CPRCM Convection-permitting regional climate model CV Controlled Vocabulary DR Data Request DRS Data Reference Syntax ESD Empirical-statistical Downscaling ESGF Earth System Grid Federation ESGF-NG ESGF Next Generation ESM Earth System Model FPS CORDEX Flagship Pilot Study GCM Global Climate Model LBC Lateral Boundary Conditions LULC Land use and land cover LULCC LULC changes ML Machine Learning NCAR National Center for Atmospheric Research, USA OMDP CLIVAR Ocean Model Development Panel ORCM Ocean Regional Climate Model 3 POC CORDEX Point of Contact RCM Regional climate model REF CMIP Rapid Evaluation Framework REP Representative Emission Pathway SAT CORDEX Science Advisory Team SSP Shared Socioeconomic Pathway TF CORDEX Task Force VIA Vulnerability, impacts and adaptation Document version history Version Comment 2025−12−05 ● First public version ● Update dates and references ● Fix ES-Doc errata system availability ● Update “Data publication” section with the latest progress ● Update REF framework progress ● Add new TF developments: experiment design to GitHub, collection of CORDEX-CMIP6 non-ESGF servers, … ● Other minor edits 2025−08−04 Initial version shared with the SAT for comments 4 1. Introduction The latest cycle of the Coupled Model Intercomparison Project (CMIP7) is about to start producing a set of climate forcings and simulations to support scientific research and assessments by the climate community, climate services, governments, and the Seventh Assessment Report of the Intergovernmental Panel on Climate Change (IPCC AR7). Concurrently, CORDEX is in the midst of generating the CORDEX-CMIP6 ensemble, with several regional communities still underrepresented in this dataset. Recent efforts in developing protocols and technical documentation for CORDEX-CMIP6 have laid a foundation for the upcoming CORDEX-CMIP7. However, integrating the advancements in modelling strategies from both CMIP and CORDEX will be essential. One of the primary goals of the Task Force (TF) in preparation for CORDEX-CMIP7 was to evaluate whether CORDEX can and should adopt CMIP7 forcing in time to contribute to the next IPCC AR7 (publications should be accepted by the end of 2027). It is thus critical to balance the innovation and demand for CMIP7−forced data with the complexities of releasing two products (CORDEX-CMIP6 and CORDEX-CMIP7) simultaneously. This dual effort could potentially detract from the completeness of the CORDEX-CMIP6 ensemble. To align with the AR7 timeline, the CORDEX-CMIP7 protocols and technical documentation should be finalized by the time the first CMIP7 models become available for downscaling, expected in about one year (mid-2026). Regardless of the timeline, CORDEX-CMIP7 will inevitably take shape in the coming years, requiring the community to be adequately prepared. In this document, we explore the necessary actions and steps to ensure a smooth transition to a CMIP7−driven CORDEX ensemble. Achieving this will require collaboration among diverse experts and consultation with the broader CORDEX community. The number and type of models used by the CORDEX community have increased and, for the next CMIP cycle, the experimental design must account for Regional Climate Models (RCMs) with coupled and uncoupled components, a resolution increase to convection-permitting (few-km) scale, and the extensive use of Empirical Statistical Downscaling (ESD) and Emulators benefiting from new Machine Learning techniques. Therefore, close coordination with other ongoing CORDEX TFs will be necessary to develop the new experimental design. For the most technical parts, we advocate for the creation of a Working Group tasked with developing the required actions/steps and maintaining the documentation and infrastructure. With this working group in mind, we review the past efforts in preparing CORDEX-CMIP5 and the current CORDEX-CMIP6, providing links to a wealth of existing resources and discussing the new opportunities and challenges for CORDEX-CMIP7. We provide recommendations to move forward in a practical and scientifically sound way, taking into account the tight deadlines imposed by the IPCC cycle. 5 Apart from the review and discussion of the required steps in preparation for CORDEX-CMIP7, some specific actions have been already carried out by the TF: ● Publication of existing CORDEX documentation with persistent identifiers and versioning in Zenodo1. ● Migration and rewording of the initial CORDEX (-CMIP5) documentation from the IS-ENES project to the WCRP-CORDEX organization in Github2. ● Update of the CORDEX-CMIP6 Archiving Specifications for Dynamical Downscaling and the CORDEX experiment design for dynamical downscaling of CMIP6. ● Migration of the development of these documents to GitHub, for a public track of changes and discussion with the community. ● Collection of regional-scale climate processes and metrics for climate model evaluation, in support of the GCM selection step (see Section 5.3). ● Collection of CORDEX publications in a Zotero.org library, where it can be accessed for different purposes and maintained collaboratively. ● Collection of servers which already provide public access to CORDEX-CMIP6 data. ● Preparation of a form to collect projects supporting the CORDEX activities (see Section 5.12) ● Community survey (Appendix 1) on potential commitments to support CORDEX-CMIP7 in time for AR7. ● Community survey (Appendix 2) on the climate forcing data sets used in CORDEX-CMIP6 ● Preparation of a suitable call for members (Appendix 3) of the working group to handle the preparation of CORDEX-CMIP7 and beyond. The document is structured by introducing the novelties in CMIP7 (Section 2) and CORDEX (Section 3), along with an overall view of the lessons learned in the preparation of CORDEX-CMIP6 (Section 4). The steps in preparation of CORDEX-CMIP7 are introduced next (Section 5), with a common structure describing the planning steps, implementation in previous CMIP cycles, a discussion of the opportunities and challenges in CMIP7, and recommendations to be considered by the CORDEX Science Advisory Team (SAT). We finalise analysing the synergies with other CORDEX TFs (Section 6) and the recommended timeline of the different steps in relation to the CMIP and IPCC timelines known to date (Section 7). 2 https://github.com/WCRP-CORDEX/cordex-cmip5 1 https://zenodo.org/communities/wcrp-cordex 6 2. What is new in CMIP7? Recently, different CMIP working groups have published several articles with updated information for CMIP7. The overview paper by Dunne et al. (2025) presents the scientific rationale in line with WCRP 2019−2028 Science Objectives, the experimental framework for CMIP7, an updated version of the mandatory Diagnostic, Evaluation and Characterization of Klima (DECK), and the newly introduced Assessment Fast track (AFT) that will be a subset of CMIP7 experiments to support upcoming assessments, such as IPCC AR7 and CORDEX, among other users. As stated in Dunne et al. (2025), the CMIP vision has been always the coordination of community-based efforts to answer key and timely climate science questions and the facilitation for delivery of relevant multi-model simulations through shared infrastructure for the benefit of the physical understanding, vulnerability, impacts and adaptations (VIA) analysis, national and international climate assessments, and society at large. Some novel points for CMIP7 discussed in the article “An evolving Coupled Model Intercomparison Project phase 7 (CMIP7) and Fast Track in support of future climate assessment” (Dunne et al., 2025) are listed in the CMIP web page https://wcrp-cmip.org/cmip-phases/cmip7 and summarized here: ● CMIP7 includes a revised DECK based on the same experiments used in previous CMIP, but it has been expanded by adding a) the historical simulation, b) a small set of “fixed-SST” experiments to characterize effective radiative forcing, and c) an expanded protocol to facilitate participation with Earth System Models (ESMs) that close the carbon budget and are capable of running with interactive CO2 forced by emissions (including positive, zero, and negative scenarios) in addition to prescribed concentrations. This will allow for more complete description and characterization. ● CMIP7 also introduces the new AFT Experiments to support both the direct needs of the climate research community for synthesis and physical science assessment as well as downstream climate services applications. The subset of simulations will include near-term prediction and long-term projection experiments that will provide critical information to satisfy the needs for both short and long term planning and for the impacts, mitigation and adaptation communities; it will also contain high temporal resolution forcing for regionally tailored information through dynamical and statistical downscaling such as CORDEX. ● CMIP7 will have updated, expanded, and improved historical forcings (solar radiation, volcanic eruption representation, aerosols, etc.) extended to at least 2021 and new scenario experiments. 7 Another CMIP7 paper that is under review is the Scenario Model Intercomparison Project for CMIP7 (ScenarioMIP-CMIP7) by van Vuuren et al. (2025) that describes the scenario experimental framework. An important innovation is that most scenarios are intended to be run, if possible, in emission-driven mode, providing a better representation of the earth system uncertainty space. The scenarios cover the period 2025−2100 (AD) with long-term extensions up to 2500 (AD). One of the goals of ScenarioMIP is to produce scenarios that can be useful to the IPCC AR7. This means that studies forming the basis of the assessment and relying on the new scenarios outcomes need to start appearing in the peer-reviewed literature in the 2026−2027 time frame. The primary purpose of the ScenarioMIP is to provide emissions and land use pathways to drive ESMs. The scenarios should encompass a wide range of policy-relevant emission trajectories considered to be plausible/feasible. Up to now, ScenarioMIP experiments were driven by concentrations. As this does not account for uncertainty in the carbon cycle response to climate, the decision was made for CMIP7 to run most simulations preferably in emission driven mode (Sanderson et al., 2024). There will be six emission-driven scenarios of high priority (esm-scen7, with three high/medium and three low and very low scenarios) and three concentration-driven scenarios (scen7, with high to low concentrations), as described in Table 1 of van Vuuren et al. (2025). The three highest emission-driven scenarios are: ● High/Medium emission (esm-scen7−h) scenario is expected to result in forcings below SSP5−8.5. ● Medium emission (esm-scen7−m) scenario explores consequences of continuing current policies without modification. ● Medium-Low (esm-scen7−ml) scenario explores a delayed increase in mitigation efforts, short of the Paris temperature goals but achieving net-zero CO2 emissions by the end of the century. Van Vuureen et al. (2025) acknowledges the role of complex climate models and the possibility of reducing the computing burden by using emulators . However, at this point in time, it is envisioned that all ScenarioMIP scenarios will be run using ESMs. However, they do not exclude the possibility that, for specific projects, some climate variables and specific scenarios emulators can be used, The High-Resolution Model Intercomparison Project phase 1 (HighResMIP1; Haarsma et al., 2016) was a new project within the CMIP6. HighResMIP1 provided a protocol for model simulations (both atmosphere-only and coupled from 12 modeling centers) to be consistently performed and compared to each other and with available observations. Given the broad impacts of the HighResMIP1 outcomes, and extensive needs for improved 8 high-resolution climate data in climate change impact assessments, CMIP7 will have the HighResMIP2 to continue providing important insights in time for the next IPCC report. The article under review “High-Resolution Model Intercomparison Project phase 2 (HighResMIP2) towards CMIP7” by Roberts et al. (2025) describes the improvements and extension of the previous works to address new science questions, and to further advance our understanding of the role of horizontal resolution (and hence process representation) in state-of-the-art climate simulations. Here we summarize some relevant points from this article: ● There are already 10 modeling groups that have proposed to contribute to HighResMIP2 with couple model simulations using forcings from CMIP6 or CMIP7; the planned atmospheric resolution varies from 25 to 9 km and the midlatitude ocean resolution varies from 25 to 5 km, as shown in Table 1 of Roberts et al. (2025). These increments may produce more accurate representation of the ocean's mesoscale, including eddies and boundary currents (e.g., Chang et al., 2020); enhancements in atmosphere resolution may produce more realistic climate extremes and upscale feedbacks (Scaife et al., 2019). ● The choice of model physics and dynamics settings with resolution (such as the treatment of atmospheric deep convection or implementation of aerosol forcing) is left to the modelling groups. In contrast to the HiResMIP model simulations, the vast majority of models in CMIP6 had grid spacings of 100–200 km in the atmosphere and 50–100 km in the ocean. Our current understanding is that such resolutions (described as standard resolution) will likely remain typical in CMIP7 (Dunne et al., 2025), given historic rates of increase in model resolution (Hewitt et al., 2022). According to IPCC AR6 (2021; their Figure 1.19), CMIP increased from 37 standard models in CMIP5 to 50 in CMIP6. The information on the number of modeling centers participating in CMIP7 is anticipated to be finalized by late 2025. In CMIP, the Climate Model Output Rewriter (CMOR) is the main software library used to translate raw model output into a standards‑conformant form. It ensures CF conventions and CMIP controlled vocabularies, maps variables to the official tables, applies required naming and units, encodes dimensions and coordinates (including calendars and bounds), writes required global attributes and citations, and produces filenames and directory layouts consistent with the CMIP DRS (Data Reference Syntax). CMOR also enables validation (e.g., via PrePARE) to catch schema errors before publication, so that downstream services encounter predictable, machine‑readable datasets. The ESGF publication workflow depends on these outcomes. The publishers expect datasets that already conform to the metadata tables and CVs (Controlled Vocabulary), carry the mandatory attributes, and follow the DRS/file naming rules. In short, CMORization is what 9 the time of writing, the CMIP7 DR is in version 1.2.2, released on Jul 25th, 202510. This version is considered final and will only be subject to updates for fixes to specific issues raised via the Github repository. CORDEX collected input in a shared document to help coordinate the input to the Mural board. The process resulted in two separate opportunities “Dynamical Downscaling” (split into 7 variable groups) and “Empirical Statistical Downscaling and Emulators” (4 variable groups), which can be explored in the corresponding Airtables (DD, ESD). As compared to the CORDEX request for the CMIP6 DR, the CMIP7 DR includes many new variables for the atmospheric chemistry and the ocean and ocean biogeochemistry, which resulted from the contribution of the CORDEX Task Force on Regional ocean climate projections. The data request might still need careful checking. For example, v1.2.1 had some inconsistency in the lake area fraction (sftlaf) as compared to the definition in CORDEX, which was prior to the CMIP7 DR. We opened an issue and, in v1.2.2, they are already consistent. In Med-CORDEX, we found some inconsistencies in the cell_methods of ocean surface wave variables (new in CMIP7); this is still to be reported. After v1.2.2, variable definitions in CMIP will be handed over from the DR task team to the CV task team, which will develop a more general WCRP Variable Registry GitHub. The proposal for longer term governance of the Variable Registry is still pending at the time of writing. Any changes post v1.2.2 should take place in the WCRP Variable Registry GitHub11. 5.2 Experiment design The CORDEX experiment design is the minimum set of instructions to coordinate the simulations carried out in the different regional domains. It is laid out in a document where mandatory and recommended practices are established regarding, for example, the simulation periods to be covered, the GHG scenarios to be used as forcing, the minimal spatial resolution, etc. Regional communities can adopt the experiment design as is, or adapt it for their purposes; but always towards being more restrictive. That is, mandatory instructions should not be relaxed. As a result of this Task Force, past documents with the experiment design have been published in Zenodo to better preserve them and to make them easily findable via unique DOIs for each different version. In CORDEX-CMIP5, a very basic set of instructions was proposed as experiment design for dynamical downscaling. In fact, it was merged into a single document including also the archiving specifications (Section 5.6). A CORDEX-ESD Experiment 1 protocol was also envisaged, following the framework of the EU-COST action VALUE. 11 https://github.com/WCRP-CMIP/Variable-Registry 10 https://wcrp-cmip.org/cmip7−data-request-v1−2−2 16 For CORDEX-CMIP6, the experiment design for dynamical downscaling was first released in 2021. Prior to that, two drafts were shared with the community for “blind” comments, to focus the improvements on common pitfalls. This experiment design included a separate document for the treatment of the aerosol forcing, prepared by the FPS-Aerosol. The experiment design for ESD is still as a second draft, since Apr 2024. Flagship Pilot Studies (FPS) have their own, more targeted, experimental setups. Some examples are the FPS-CONV, FPS-CPTP, etc. FPSs have explored new research avenues (Convection-permitting modelling, Machine learning ESD tools, Urban climate modelling, etc) that will likely feed back to the general experimental design to be set up for CORDEX-CMIP7. Some of these are also under heavy coordination efforts at the moment in several Task Forces. Therefore, the experiment design step will necessarily need to be carried out in view of the outcome of these other efforts. In the following, we just discuss some relevant aspects to take into account and, where appropriate, point to other strategic activities (TFs, FPSs) for deeper discussion. We also take into account the CORDEX White Papers on the scientific challenges for dynamical downscaling and ESD. For CORDEX-CMIP7, given the tight deadlines to contribute to AR7 and other assessments, it will be hardly possible for the modelling groups to perform strong model developments and for the community to agree on major changes to the experimental setup. On the other hand, new modelling approaches demand a careful experiment design to integrate all approaches in a consistent manner. Therefore, we advocate for a two tiered experiment design, as reported to the IPCC on February 20th (Briefing on CORDEX and AR7), in which there is: ● An “Assessment Fast Track” stream (CORDEX-CMIP7−AFT), essentially using the same CORDEX-CMIP6 models and minimally adapted protocols to provide some CMIP7−based information in time for AR7 and other assessments. If ready, ML-based techniques could contribute to expanding this stream. This initiative will feature a compact set of experiments targeting climate science priorities relevant to AR7 (e.g. overshoot scenarios), with results anticipated tentatively by mid-2027. ● A “Standard” stream (CORDEX-CMIP7) over a longer time frame, where new model and climate forcing developments are included, representing a next generation downscaling. A first challenge in this setup is the compatibility of the two streams. As proposed above, CORDEX-CMIP7−AFT would be closer to CORDEX-CMIP6 in terms of protocols and specifications. This would also facilitate joint CORDEX-CMIP6/CORDEX-CMIP7−AFT analyses relevant for AR7. On the other hand, we need to ensure that CORDEX-CMIP7−AFT is compatible or part of CORDEX-CMIP7. But the latter will very likely need new specifications to fully adapt to the CMIP7 standards. Many CMIP7 technical details are still 17 unknown, so this issue will need further discussion in the future. In the following, several aspects of the experiment design are discussed, without attaching them specifically to a given stream. One of the main aspects covered by the experimental setup is the definition of standard domains and recommended and minimal resolutions (grid spacing). The minimal resolution has increased from 50 km in CORDEX-CMIP5 to 25 km in CORDEX-CMIP6. A natural recommendation for a minimal resolution in CORDEX-CMIP7 would be to set it to 12 km. The recommended resolution can also be about this 12−10 km grid spacing as grids of ~6 km resolution would lay in the so-called “gray zone”, and it would be better skipped in favour of CP simulations in the range of 2−4 km, but these would need intermediate domains, and fall in the realm of the TF-CPM. A minimal resolution of 12 km can still be challenging for large domains and/or modelling groups with less resources. To maximise participation, especially for CORDEX-CMIP7−AFT, the same setup as in CORDEX-CMIP6 could be kept (25 km minimum, 12.5 km recommended). Standard, continental domains have proved useful and regional communities are already organized around them. There are mechanisms already established to propose changes to those domains. Additionally, as CP simulations become standard in CORDEX, a new mechanism needs to be devised to propose and approve sub-continental domains. National domains play a key role at this scale as many CORDEX groups run on national funds with targeted interests. Attracting these downscaling efforts will increase the number of climate projections and modelling centres in CORDEX. Care should be taken in order not to exclude countries with less capacities, notably over large domains. The Africa domain has been the focus in both CORDEX-CMIP5 and CORDEX-CMIP6. In CORDEX-CMIP5, this was a success story, with many simulations contributed for this domain. In CORDEX-CMIP6, this focus region highlighted in the experiment design has been largely ignored by the community, and no simulation is currently available over Africa. The strategy to cover domains with few local resources and high vulnerability needs to be revisited. A tight integration of CORDEX-CORE into the general experiment design from the beginning could help in this regard. CP simulations at continental scale have been carried out in different centres: SAM-4 (NCAR-SAAG), AFR-4 (UKMO), CONUS-404 (NCAR), NAM-4 (ANL), SEA-8 (SINGV), etc. Unfortunately, such efforts did not meet the CORDEX protocol for different reasons (e.g. due to using a Pseudo Global Warming -PGWapproach). Future experimental setups could be relaxed or extended to accommodate some of these downscaling efforts, so CORDEX extends the coordination to a wider community. Such particular discussion belongs to the TF-CPM, but other potential extensions could be considered regarding the bias adjustment of lateral boundary conditions (LBC) prior to downscaling (Wamahiu et al., 2025), or the downscaling of time slices (e.g. GWLs). These pose their own new challenges 18 as different methodologies exist to bias adjust LBCs, and GWLs may limit downstream applications. Dynamical downscaling is still central to CORDEX, with most survey participants (Appendix 1) focusing on this approach. Still, statistical approaches have also a central and complementary role, which has never been fully integrated in the CORDEX framework despite several attempts. The diversity of methodologies (Perfect prognosis, Model output statistics, Bias adjustment, hybrid methods, RCM emulators) and output nature might have played a role. Challenges include (Gutierrez et al., 2022) multivariate aspects, the progress in intercomparison frameworks and, generally, the distillation of robust climate information not only from these methods, but from all lines of evidence. Some of these challenges are being tackled in the ongoing CORDEX TF for Machine Learning (TF-ML), but we are also planning a POC meeting (end of 2025) focused on the better integration of ESD methods in the CORDEX framework. The outcomes of these initiatives will hopefully bring a clearer joint framework for CORDEX-CMIP7. In particular, RCM emulators are promising tools to explore natural variability. This might give a chance to downscaled climate (decadal) prediction, which was already considered in the early days of CORDEX12, but disregarded due to the computational burden. Emulators, and ML in general, is also leading to a growing CORDEX community that will need to be integrated in the activities. Dynamical downscaling has also progressed to a point (see Section 3) where the recommendations to increase the effort towards including additional model components (ocean, sea ice, glaciers, urban areas, lakes, vegetation/agriculture, land hydrology, aerosols and chemistry) might be turned to mandatory, at least for certain domains, model resolutions or scenarios (e.g. modelling glacier recovery could be key in overshoot scenarios). In this respect, the work of different past and ongoing FPSs (Air-sea, Aerosol, LUCAS, URB-RCC, CPTP, etc.) will be key to inform the experiment design, as well as the activity of the ongoing TF-Ocean and TF-CPM (see also Section 6). Specific dynamical downscaling approaches such as spectral nudging13 and variable-resolution global models only forced by the SST14 might need to be more carefully considered to be integrated with the rest of downscaling approaches. Addressing the long-standing challenge of ensuring that forcing datasets are fully consistent with the driving CMIP GCMs should be a priority for CORDEX-CMIP7, if not for CORDEX-CMIP7−AFT. CMIP consistently uses input4MIPs15 datasets, but this is not the 15 https://input4mips-cvs.readthedocs.io/en/latest/dataset-overviews 14 https://github.com/WCRP-CORDEX/simulation-status/issues/78 13 https://github.com/orgs/WCRP-CORDEX/discussions/13 12 see e.g. section 6.5 in the Toulouse report (2009): https://cordex.org/wp-content/uploads/2018/03/cordex2009_toulouse_report.pdf 19 case in CORDEX (Appendix 2). Some forcings are easier to adopt (e.g. solar irradiance) than others (e.g. land use changes, emissions), which would need a strong development effort. Regarding the land use data sets, some recommendations from FPS-LUCAS (D. Rechid) are: ● Consider land use & land cover (LULC) changes in ERA-driven simulations. It would be important that all models use the same LULCC dataset. Ideally, there would be a regional expert group which could transfer the LUCAS method to their respective CORDEX region, so that datasets are also consistent between regions, but could consider specific regional LULCC characteristics. ● Use the same LULCC datasets for the historical GCM-driven simulations. ● For the GCM-driven future climate change simulations, so far only human-driven land use changes have been considered in RCMs, but not climate-driven land cover changes (at least within LUCAS). It first would require new methods to derive maps which consider both humanand climate-driven LULCC. However, this would also add additional uncertainties. Thus, the recommendation would be to at least consider human-driven land use changes, consistent with land use change scenarios e.g. from SSPs or REPs. Most CORDEX models are not able to deal with emissions and are not considering it in their near future development plans (Appendix 2). Also, GHG emission-driven GCMs will likely show strong variability in their future GHG evolution. This is a new source of uncertainty in CMIP7. In CORDEX-CMIP7, even if emissions-ready RCMs would be available, they will likely develop their own GHG concentration response, inconsistent with the driving GCM. In this situation, we would recommend, at least for CORDEX-CMIP7−AFT, to stick to concentration-driven GCMs to drive the RCMs, so using a consistent GHG forcing. The selection of priority CMIP7 REP scenarios to simulate will be a major decision which needs to be taken relatively quickly for CORDEX-CMIP7−AFT. Some aspects to consider are: ● CORDEX-CMIP7−AFT and CORDEX-CMIP6 will coexist in time and both should be analysed together. Therefore, the REP scenarios prioritized should either complement the CORDEX-CMIP6 SSP1−2.6 and SSP3−7.0 with an alternative future not considered in SSPs (e.g. an overshoot scenario) or align with the existing scenarios (e.g. the High REP scenario, if this leads to a forcing close to SSP3−7.0). ● Scientifically, scenarios with a strong forcing (High, High overshoot) are preferred, since they lead to a clearer signal and also reach a variety of climatic conditions which is key e.g. to train emulators, which can then be used to explore other scenarios. 20 ● The decision needs to consider different users' views and different needs (impacts, adaptation, mitigation communities), which lie beyond the CORDEX community and which typically demand realistic but also worst-case scenarios. From the survey (Appendix 1), the CORDEX modelling community has no strong commitments towards particular scenarios. A potential way forward is to have a workshop or a POC meeting dedicated to REP scenario selection. POCs would collect needs from their respective communities and key talks could show the details of the new scenarios and their implications. A final comment on the experiment design regarding the extent of the simulations. As we progress in the century, the year 2100 is closer, and a minimal extension to at least 2125 (100 years) should be considered, especially to better sample the evolution of overshoot scenarios. Even longer extensions could make sense in polar regions (e.g. up to 2300), but LBCs might not be available. Also, extensions back to GWL0 (1850−1900) could be considered to better characterize the regional warming since preindustrial levels. 5.3 Driving GCM selection The CORDEX community applies model selection for two main reasons. Firstly, GCM errors are well known to propagate through the downscaling methods and affect the suitability of regionalized, high-resolution model output data, thereby compromising all downstream applications. Secondly, the community cannot afford to downscale all available global climate model simulations and a limited number of model runs must therefore be chosen. In order to guide informed model selection, a number of model selection criteria must be defined and the models be evaluated correspondingly. In CORDEX-CMIP5, this important working step was not coordinated in a systematic way and a GCM “ensemble of opportunity” was used in most CORDEX regions. CORDEX-CMIP6 has seen a number of coordinated GCM selection efforts for Europe (Sobolowski et al. 2025), North America (Ashfaq et al. 2022, Goldenson et al. 2023, Matte et al. 2025), South and Southeast Asia (Nguyen et al. 2025), South America (Arias et al. 2025), and Australia (Di Virgilio et al. 2022, Grosse et al. 2023). For the example of EURO-CORDEX, the following set of selection criteria was established: 1. Data availability 2. Model Plausibility / Performance / Fidelity 3. Model Diversity / Independence 4. Spread of future outcomes / Climate Sensitivity 21 These criteria have been agreed on in this form or in a similar form in the other CORDEX regions. In support of the AR7 Assessment Fast Track, the CMIP Model Benchmarking Task Team (MB-TT) was created to provision a core set of diagnostics and metrics to compare GCMs against observations, currently on global or continental scale, on the basis of monthly data. The software developed to this end is the Rapid Evaluation Framework (REF, Hoffman et al. 2025), based on different existing frameworks (ESMvalTool, PMP, ILAMB, IOMB) that can be executed in a container, and will be run directly on the ESGF nodes in order to quickly evaluate newly available GCMs, beginning with those contributing the CMIP7 Assessment Fast Track. The v1 release of the REF (Hassler et al. 2025) launched on October 7, 2025 consists of approximately 30 diagnostics (this is the corresponding work-in-progress table) that were discussed in a dedicated Hackathon event in March 2025. The MB-TT is currently considering new processes (aka diagnostics) and metrics to be included in the REF. In this context, we had a series of meetings with the CORDEX POCs (starting on February 27, 2025) to ask for regional-scale evaluation processes and corresponding metrics relevant in their domains. These led to a collection initially published on Zenodo on May 14, 2025 that can be applied to select GCMs whenever needed (Brands et al. 2025). Thereafter, a substantial part of the metrics published in this collection was included into a comprehensive Python toolkit developed by Moetasim Ashfaq for the TF CORDEX-CORE-CMIP6, and the associated performance, independence, and spread in future outcome results have already been calculated. These ongoing GCM selection efforts were reported to the MB-TT and the broader GCM evaluation community at the Model selection – supporting downstream activities use of CMIP7 Assessment Fast Track workshop on October 7, 2025. They are expected to be considered in the extension phase of the REF that is currently under way. Meanwhile, several regional-scale diagnostics and metrics available from the REF dashboard can be used for providing “objective” GCM selection criteria whenever needed. However, the description and proper identification of some of the metrics used there is still unclear, a shortcoming expected to be resolved in the future. There is also a REF data request. The variables relevant for CORDEX not included there should be covered by the CORDEX opportunity sent to the CMIP7 data request (see Section 5.5). Based on this information, we can filter those GCMs that actually provide the data necessary for downscaling. The aim for CORDEX-CMIP7 is to explore and condense the aforementioned regional GCM selection efforts into a single community approach using a common, accepted methodology, taking into account the perspective of other CMIP and RIfS task teams and Model Intercomparison Projects. To come to a consensus in this regard, we are currently organizing a multi-community side session about GCM selection to be held at the CMIP Community Workshop 2026 in Kyoto. If there is a substantial amount of CMIP7−AFT 22 scenario simulations available by then, we should be able to select driving GCMs for CORDEX shortly after this Workshop. A dedicated, CORDEX-only workshop on driving GCM selection will likely be needed to make a final decision. 5.4 Monitoring simulations CORDEX simulation ensembles typically take a long time to complete (Figure 2) and they develop very differently in the different domains (Figure 3). Monitoring the simulation status and plans by the different modelling groups provides essential information to plan and balance the CORDEX grand-ensemble as it grows. In CORDEX-CMIP5, there was no central monitoring system and each regional community managed internally the status of their simulations, typically exchanged during annual meetings. For CORDEX-CMIP6, the CORDEX SAT made an effort to collect centrally the simulation plans; initially by means of a Google Doc table, but then moved to machine-readable files under GitHub16, so anyone can produce their own status plots (e.g. Figures 2 and 3) out of this central information. Some regional communities that kept their own simulation status (e.g. MED) are moving to this central resource. The information collected allows to display GCM-RCM-SSP combination matrices for the different domains, or even to keep track of specific experiments designed by the regional communities. Datatables quick filters are available in the simulation list to easily locate simulations by any of the fields requested, or to share unique URL searches; e.g. https://wcrp-cordex.github.io/simulation-status/CMIP6_downscaling_plans.html?searc h=completed%20evaluation shows completed evaluation simulations. 16 https://wcrp-cordex.github.io/simulation-status 23 Figure 2: Cumulative number of CORDEX simulations completed on different dates. The stacked bars show the number of simulations completed, running and planned as of 2025−07−04. The cumulative simulations for the future use information on expected completion dates from https://wcrp-cordex.github.io/simulation-status. Figure 3: Number of scenario simulations completed, running and planned by each of the regional CORDEX domains as of 2025−07−04, according to the information in https://wcrp-cordex.github.io/simulation-status. Of course, the information provided is only as good as the community keeps it updated with their simulation plans and status. The report of simulation plans is an entry card for model and institution registration, i.e. no model can be registered unless the simulations planned with it are known to the community. Unlike the CV registry, the simulation status collection is quite flexible and plans can be registered with tentative model identifiers or details to be decided (e.g. driving GCM or scenario). The idea is to facilitate the early emergence of as many CORDEX simulation plans as possible. Any detail can be updated at any time. The same repository is already holding plans and simulations for CORDEX-CMIP7 from an early stage. In fact, it already holds retrospective information files for CORDEX-CMIP5, using ESGF-derived information. A pending task is merging the information provided by the modelling groups with that of the ESGF as CORDEX-CMIP6 simulations start to be available there. 5.5 Data request The CORDEX data request (DR) is the set of variables requested from the modelling groups to be saved in a format consistent with the CORDEX archiving specifications (see next section). The data request consists of a list of climate variables, along with their output frequency and other attributes, such as the corresponding variable name, long name, 24 standard name (i.e. the physical parameter according to the CF conventions), the aggregation in time and space, or the dimensionality of the variables. The development of the data request involves both the scientific task of selecting the variables to be included and their priorities, and the technical question of how to actually store or present the data request to the users. For CORDEX-CMIP5, the DR was distributed as a PDF table17 maintained in a MS Excel file under the github organization of the EU-funded IS-ENES project18. As part of this Task Force, all CORDEX-CMIP5 documents have been collected under the WCRP-CORDEX organization in Github19 and updated to the current naming of this initial stage of CORDEX (CORDEX-CMIP5). This initial CORDEX DR was influenced by previous coordinated downscaling initiatives, such as the US NARCCAP program20 or the EU-funded ENSEMBLES project21, which already provided standardized lists of RCM variables based on early CMIP conventions. Apart from the general CORDEX DR for regional domains, other initiatives, such as FPSs, maintained their own variable lists based on the general one (e.g. the FPS-CONV variable list). For CORDEX-CMIP6, initially, an online Google spreadsheet was used (also served from the cordex.org website as a MS Excel xlsx file), but these documents are now obsolete and the data request is maintained as machine-readable CSV files in Github22 to track its development and allow automatic creation of consistent CMOR tables23 and MS Excel files. A tutorial document provided a summary of the DR, but it is now also superseded by the README file shown as the front page in the DR repository. The selection of variables in the CORDEX-CMIP6 was based on the CORDEX data download statistics from ESGF24 and through discussions with the CORDEX and impact modelling communities. In particular, the 17 CORE (mandatory) variables in the CORDEX-CMIP6 DR were selected as the most downloaded variables, plus the fixed fields for the terrain elevation (orog) and land-sea mask (sftlf). New variables and output frequencies arose from the dialogue with different modelling and user communities. The CORDEX-CMIP6 DR still needs fine tuning. It is set up as a default data request that can be adopted “as is” by the different regional communities or activities (e.g. FPS). However, different CORDEX communities can also create their own DR25, potentially with different 25 Under https://github.com/WCRP-CORDEX/data-request-table/tree/main/data-request 24 https://esgf-ui.cmcc.it/esgf-dashboard-ui/cordex.html 23 https://github.com/WCRP-CORDEX/cordex-cmip6−cmor-tables 22 https://github.com/WCRP-CORDEX/data-request-table 21 https://ensemblesrt3.dmi.dk/RCM_output_table_final.xls 20 https://www.narccap.ucar.edu/data/data-tables.html 19 https://github.com/WCRP-CORDEX/cordex-cmip5 18 https://github.com/IS-ENES-Data/cordex 17 https://wcrp-cordex.github.io/cordex-cmip5/CORDEX_variables_requirement_table.pdf 25 both machineand human-readability of the source files. One of the potential points to extend the EMD for CORDEX is regarding the input forcing data sets. In CMIP, it is taken for granted that the modelling centers use the input4MIPs data sets but, according to the survey run as part of this TF (see Appendix 2), this is not the case in CORDEX. Therefore, it would be advisable to include information regarding the forcing inputs used, as these can be key to interpret the results. Initiative MIP era Model documentation CORDEX CMIP5 Table in Zenodo maintained as Google Document EURO-CORDEX CMIP6 Table in Zenodo maintained as Google Spreadsheet Med-CORDEX CMIP5 System maintained by the HyMeX project with predefined form. CMIP6 Github-maintained website with free-text description plus machine-readable metadata preamble CORDEX East Asia CMIP5 Website: https://cordex-ea.climate.go.kr/cordex/models.do with free text model descriptions (common sections and extension) CMIP6 Table 2: Links to CORDEX models documentation for different initiatives. 5.8 Quality Control CORDEX Quality Control (QC) ensures that published datasets comply with CF conventions and CORDEX archiving specifications (see above). This compliance allows users to trust that the datasets are compatible with standard CF-compliant tools and analysis scripts. Adherence to CF conventions includes using only registered vocabularies for variables and coordinates—for example, the correct use of the standard_name attribute and valid area types. Ensuring alignment with standard controlled vocabularies (CVs) and conventions makes CORDEX datasets usable across a broad range of standard metric tools. CORDEX QC also includes specific checks for consistency within domain specifications. It is important to note that this quality control process does not assess scientific validity. Instead, it focuses on technical compliance, with minor exceptions such as verifying that values (e.g., temperature) fall within expected ranges. 32 In the CORDEX-CMIP5 phase, quality checks were primarily performed during the data publication process at the respective ESGF data nodes. Datasets that failed QC at these nodes were rejected and not published. The main tool used was QA-DKRZ, developed by DKRZ, which could also validate CORDEX-specific requirements such as grid definitions. However, this tool has since been deprecated. In addition to automated checks, CORDEX-CMIP5 included a manual QC process based on a checklist to ensure basic compliance. If a dataset needed to be retracted from the ESGF, this was typically documented through the Earth System Documentation (es-doc) service. However, despite the existing efforts, the CORDEX-CMIP5 datasets proved to be less compliant with controlled vocabularies (CVs) and archiving specifications than initially expected. This lack of consistency posed challenges for the automated processing and analysis of large ensemble datasets. These issues only became fully apparent once the data were shared and exchanged through the ESGF and evaluated across different modeling centers. A key reason appears to be the use of varying QA strategies at different ESGF data nodes, leading to inconsistent levels of quality control and validation. With the transition to CMIP6, QA-DKRZ was deprecated. As a result, several publicly available compliance checkers—primarily focused on CF compliance—were evaluated for use with CORDEX-CMIP6 datasets. These included: ● PrePARE ● CEDA cf-checker ● IOOS compliance-checker Although PrePARE became the primary tool for CMIP6 dataset validation, it is not suitable or flexible enough to be effectively applied to CORDEX datasets without significant effort. A more general-purpose controlled vocabulary (CV) checker may be included in the upcoming CMOR4 library, expected by the end of 2025. At present, no dedicated QC tool exists that can handle arbitrary controlled vocabularies. However, there is ongoing discussion around leveraging the IOOS compliance checker, which is anticipated to evolve in alignment with CF-1.11 standards. In parallel, a collaborative effort is underway to develop a plugin—cc-plugin-cc6—based on the compliance-checker framework. This plugin is designed to be flexible, supporting multiple MIP-specific CVs in addition to the CF compliance checks already provided by the compliance-checker. Currently, the EURO-CORDEX community has implemented the cc-plugin-cc6 as part of its QA process for the joint evaluation activities under CORDEX-CMIP6. In this effort, a large number of ensemble simulation datasets from various modeling groups are currently 33 collected at a dedicated data node at the Jülich Supercomputing Centre (JSC). This staging area allows for preliminary quality control and technical validation before the datasets are formally published to the ESGF via DKRZ. This workflow ensures that data undergoes a consistent QC process and helps identify and resolve potential compliance issues early, facilitating smoother and more reliable publication. It also supports more coordinated evaluation and intercomparison activities across the CORDEX-CMIP6 community. In the context of CMIP7, there is ongoing debate about whether quality assurance and quality control should be a prerequisite for publishing data to the ESGF (Earth System Grid Federation). This topic was prominently discussed during the WIP (WGCM Infrastructure Panel) meeting on February 25th, 2025. A key question is where and by whom the QA/QC process should be carried out. Many data publishers and ESGF node managers argue that QA should be performed at the modeling centers—before data is transferred to ESGF nodes—to avoid wasting bandwidth and storage space on non-compliant datasets. The urgency of some projects, such as the AR7 fast-track, complicates this issue, as they often proceed without waiting for full QA/QC validation. However, this raises concerns—what happens if a bug prevents a dataset from passing QA? In such cases, publication would be delayed until the issue is resolved and valid data is provided. To support improved QA/QC workflows, DKRZ plans to contribute to quality control efforts for CORDEX-CMIP6/UDAG and the upcoming CMIP7/CAP7. As part of a broader initiative, French colleagues are currently developing a general-purpose ESGF-QC application under a C3S project. This tool is intended to serve multiple projects, including CORDEX-CMIP6 and, eventually, CORDEX-CMIP7. The new QC tool is based on a fork of the IOOS compliance-checker and aims to offer flexible, extensible validation mechanisms for a variety of controlled vocabularies and project-specific requirements. DKRZ is actively contributing to its development to ensure it meets the specific needs of the CORDEX and CMIP communities. 5.9 User guidelines User guidelines are key as supporting material of the simulation data generated within CORDEX. Unlike technical model documentation or research papers, they provide basic information on model output data meaning and consider the whole ensemble, helping users understand the fitness of different approaches, models or sub-ensembles for particular purposes. Technical language is avoided to reach as wide an audience as 34 possible. User guidelines can be shaped as a document, but also as collections of Frequently Asked Questions (FAQ). In CORDEX-CMIP5, several guidelines were prepared by different regional communities (Table 3). The CORDEX FAQ at cordex.org focuses mainly on technical data access and processing questions, rather than data meaning and usability. Initiative User guidelines CORDEX https://cordex.org/faq/faq-data-and-access EURO-CORDEX Guidance for EURO-CORDEX climate projections data use (v1.1, Feb-2021) NA-CORDEX https://na-cordex.org/guidance-data-use.html … Table 3: Links to CORDEX models documentation for different initiatives. No guidelines are currently available specifically for CORDEX-CMIP6. However, there is a current effort, e.g. in Europe, to update the existing guidelines to the particularites of CMIP6 and the ensemble design in EURO-CORDEX-CMIP6. The general CORDEX FAQ section also needs maintenance, as some of the questions are outdated and point to resources (e.g. Terms of Use) for CORDEX-CMIP5. A new joint task team on “Responsible Data Use” has been setup by CMIP and RIfS31, seeking to reduce data misuse by developing recommendations for better means of documenting the limitations of model outputs for various key applications. They plan to gather existing information on downstream usage of model output and engage both modellers and users to develop optimal communication channels for appropriate use of climate model output data. This effort aligns well with the CORDEX needs in this matter and a follow up will be necessary to ensure that the CORDEX modelling methodologies are considered; both as a user of global model output and as a producer of regional model output. Swen Brands is the current CORDEX representative in this task team. 5.10 Data publication CORDEX data publication refers to the process of making model output, following the archiving specifications (Section 5.6), available publicly. This has been typically done via the Earth System Grid Federation (ESGF), a global network for sharing and accessing climate and Earth system data, supporting not only CORDEX, but also CMIP and other 31 https://www.wcrp-rifs.org/activities/working-groups/responsible-data-use 35 initiatives. CORDEX data publication entails technical work by the ESGF administrators to configure their nodes to support the CORDEX DRS. The configuration feeds from the CORDEX CVs and thus needs feedback between CORDEX and ESGF, and requires models (source_id’s) to be registered prior to publication. To date, only CORDEX-CMIP5 data have been published to ESGF. A global configuration file is available32, but local ESGF nodes might have modified versions to allow the publication of specific data sets. Apart from the Domain activity, data from other activities, such as CORDEX-Adjust or CORDEX FPS-CONV, have been published on ESGF. Conversely, there are CORDEX-CMIP5 data published out of the ESGF. Notably, Med-CORDEX data were never published to ESGF and relied on their own server33 and registration process to access the data. Data for other domains have also been replicated in other non-ESGF servers. For instance, NCAR keeps a copy of the data for the North American domain34, but reformatted, not following the CORDEX DRS. Some monthly EURO-CORDEX data are available on the cloud, via the Amazon Web Services (AWS) open data35, as are a subset of daily NA-CORDEX data in Zarr format36. This simplifies the access to portions of the data. The Copernicus Climate Data Store (CDS37) also holds a copy of some commonly used daily variables from CORDEX-CMIP5 and has gone through an additional quality control (Diez-Sierra et al., 2022). Data publication on ESGF for CORDEX-CMIP6 is still pending. Some data are available for specific domains via dedicated servers (e.g. in AUS; even regridded and bias adjusted data), but ESGF configuration to support publication of CORDEX-CMIP6 data is still in testing phase. As an interim solution, in the TF-CMIP7 we started collecting servers which already provide access to CORDEX-CMIP6 data. The ESGF itself is currently going through deep changes. ESGF 1.5 is the current stable version, but ESGF Next Generation (ESGF-NG) is expected to be available after December 2025, modernizing the system with a cloud-native architecture, improved scalability, and enhanced user workflows. The old system will need to be down shortly after. Under the current resource constraints, CORDEX-CMIP5 data will not be immediately migrated to ESGF-NG; this needs to be handled to maintain access. By the end of Aug. 2025, there was a meeting between TF-CMIP7 and ESGvoc (L. Troussellier) to discuss the requirements to ingest the CORDEX-CMIP6 CV into ESGvoc and enable publication on ESGF-NG. This work is now ongoing, along with the preparation of the rest of the steps to have CORDEX data published on ESGF-NG by early 2026. 37 https://cds.climate.copernicus.eu/datasets/projections-cordex-domains-single-levels 36 https://registry.opendata.aws/ncar-na-cordex 35 https://registry.opendata.aws/euro-cordex 34 https://rda.ucar.edu/datasets/d316009/dataaccess 33 https://www.medcordex.eu/medcordex.php 32 https://github.com/ESGF/config/blob/master/publisher-configs/ini/esg.cordex.ini 36 Regardless of the technical difficulties of ESGF node configuration, a major challenge for CORDEX is to locate ESGF nodes willing to store the output, which, for CORDEX-CMIP6, is estimated to be about 8 PB considering Core and Tier1 variables. The limited data storage capability of the CORDEX modelling groups is the most significant barrier found during CORDEX-CMIP6 (see Appendix 1). Some ESGF nodes are willing to store the output from specific domains or modelling groups, but this still leaves some CORDEX output with no destination node. In order to locate ESGF nodes willing to take CORDEX data, two different surveys have been carried out: one from this task force, distributed via the CORDEX POCs, and another one in collaboration with CMIP IPO, asking for the willingness to store CMIP and CORDEX data and distributed via de CDNOT (Petrie et al., 2021). The former collected contact point details and comments regarding the possibility to host CORDEX-CMIP6 data for each node. Contact information is not reproduced here38, but the few nodes replying and their comments are shown in Table 4. Dom. ESGF Node Contact point Comments AUS esgf.nci.org.au Ben Evans Hosting data (as the primary ESGF node) for the institutes that we have storage contracts with: all AUS-20i except IRD-MF EUR data.meteo.unican.es Antonio Cofiño At least UCAN data. Might store some others. EUR esg-dn1.nsc.liu.se Prashanth D. Rao Only HCLIM43−ALADIN simulations generated by SMHI EUR esgf-ictp.hpc.cineca.it Currently down due to Bologna flooding EUR esgf-node2.cmcc.it Will store CMCC data. Need to check whether other data could be taken EUR esgf.dwd.de No CORDEX data storage planned EUR esgf1.dkrz.de Stephan Kindermann In principle, only data from German modelling groups. EUR vesg.ipsl.upmc.fr G. Levavasseur (at least Med-CORDEX, EURO-CORDEX, Central America, Australasia) with French models (ALADIN, RCSM, AROME) WAS esg-cccr.tropmet.res.in Sandip Ingle The CCCR-IITM ESGF data node will be able to host the CORDEX-CMIP6 South Asia domain simulations 38 They are kept by the TF team and will be handed over to the upcoming WG. 37 Table 4: Information on active CORDEX ESGF nodes and plans to store CORDEX-CMIP6 as collected via the CORDEX POCs. The survey conducted by the CMIP IPO (Table 5) was responded to by a few more node administrators but the answers do not always match. Total storage allocated for CORDEX-CMIP6 among the respondents is 3.340 PB. One potential way forward is to contact again all respondents to both surveys, showing the summary of the expected storage, and requesting again the maximum storage that they can allocate for CORDEX-CMIP6 and CORDEX-CMIP7. Another pending activity is to estimate the storage covered by the comments in Table 4, as large fractions of the total storage could be already covered (e.g. the AUS simulations -1.3 PBare covered by the NCI node, even if no exact space allocation was provided in the CMIP IPO survey). Table 5: Excerpt of the survey conducted by the CMIP IPO, including only the responses declaring participation or plans in the CORDEX data archival. Columns are: Primary contact name | Node identity | Current plans for storage and retrieval of historical ESGF data | Plans for 2025 and beyond | Active projects associated with the node | Archival projects associated with the node | Storage dedicated to CORDEX-CMIP5 (PB) | Storage dedicated to CORDEX-CMIP6 (PB). Finally, in CORDEX-CMIP7 we could consider enhancing part of the data distribution from ESGF to cloud-based object storage, for example through services offered by the 38 European Open Science Cloud39 (EOSC) or commercial providers such as AWS through open data sponsorships as was done for a small subset of EURO-CORDEX datasets40 and NA-CORDEX41. Such a move would improve global accessibility, allow better co-location with cloud compute resources, and reduce the operational burden of maintaining a distributed network of ESGF mirrors. However, to take full advantage of the performance and scalability of cloud storage, data would likely need to be converted from traditional file-based NetCDF into the Zarr format. Zarr is designed for cloud object storage and enables efficient, parallel access to data over HTTP(S) or S3, which would support more flexible and scalable workflows for users. At the same time, adopting Zarr could help simplify some long-standing issues tied to the ESGF infrastructure, particularly the DRS directory structure, which is tightly coupled to traditional file-system-based storage. Because Zarr stores metadata within the dataset itself, it reduces the need for complex directory hierarchies and could make the overall system easier to maintain and extend. One caveat is that RCMs do not write directly to cloud storage, so at least initially data will still need to be stored on a traditional filesystem. For large datasets, Zarr data stores can consist of a huge number of files, which can be problematic with the number of filesystem inodes they take up. In practice, the workflow will likely need to still write the data to netCDF, CMORize, and then convert it to Zarr. 5.11 Errata Errata refers to a record of errors found in the data and/or metadata after publication, along with their current status indicating whether a corrected version exists and potential problems in using the data as they are. In principle, once data are published, they should not be unpublished, as they could have been used for analyses that could not otherwise be replicated. Up to now, CORDEX has no centralized errata reporting system and different regional communities have set up their own solutions. In CORDEX-CMIP5, in Europe, there is a Google spreadsheet listing the errors (the latest dates from 2021) and, more generally, this website which lists some other resources and instructs users to report problems via e-mail. For NAM, there are caveats42 (not exactly errata) and known issues43. On top of this, there were a few CORDEX problems reported in the ES-Doc errata service (now under https://errata.ipsl.fr, selecting project “CORDEX”), that were collected there by the C3S CDS. The ES-Doc developments are planned to continue under the framework of an ESGF Errata Service. 43 https://na-cordex.org/known-issues.html 42 https://na-cordex.org/caveats.html 41 https://registry.opendata.aws/ncar-na-cordex 40 https://registry.opendata.aws/euro-cordex 39 https://eosc.eu 39 Some problems considering compliance with CF conventions were reported in a github issue. These issues mainly came up when EURO-CORDEX datasets were intercompared and aggregated across different RCM models. In CORDEX-CMIP6, there is no common approach, either. Although, in this case, there are no data officially published, so we are still in time to create a centralized errata system. In Australasia, there is this website collecting the errors in a table. In Europe, an errata system is still pending. The use of GitHub issues is still in discussion; this approach is being followed in FPS-URB-RCC, where data problems can be discussed and errata reports can be generated (here a mock-up example). Also in FPS-CONV. In CORDEX-CMIP7, we should have a centralized errata system where users can find all potential problems in a single place. Plain text solutions facilitate long-term availability (unlike e.g. ES-Doc) and machine-readability facilitates the display of the information in various ways (e.g. selecting the errors for a particular regional domain, creating reports, etc.). It is especially important not to rely on dedicated infrastructures, developed for this task in specific projects and likely to be discontinued after the project ends. An errata system is something sufficiently simple to be maintained with very basic tools. The new ESGF Errata System to be created as a central hub to report and look up for data errors seems a promising way forward. Errata, documentation (Section 5.7) and guidelines (Section 5.9) should be accessible from the ESGF, where data lives, so users can easily find how to interpret the data they are downloading and the errors they may contain. Another problem is that once data are downloaded, users rarely return to ESGF to find this information. Or, if a user finds the data in the filesystem, there is no way to know from where it was downloaded. Therefore, we might also consider adding in the archiving specifications (Section 5.6) global metadata to all files, pointing to the errata, documentation and guidelines on how to use the entire ensemble (not just the individual file). 5.12 Monitoring the science All of the above steps refer to CORDEX coordinated output data production, but this is just a means to perform coordinated downscaling science. Science monitoring refers to tracking the science enabled by CORDEX. This usually takes the shape of scientific publications, but also scientific projects and other science outputs (more informal analyses, applications in different platforms, etc.). Since its inception, there is a central collection of scientific publications under cordex.org. Several CORDEX initiatives (regional communities, FPS, …) have duplicated this effort trying to collect publications in various ways (Table 6), in order to have a list of the publications relevant for their domain. Typically, publications are collected in websites or 40 documents, not really appropriate for further analysis or even collection. These are typically also easily outdated (Table 6), even though these need to be collected for the annual domain reports . Initiative Publication list Updated CORDEX https://cordex.org/publications/peer-reviewed-publications 2024 AFR https://www.csag.uct.ac.za/cordex-africa/cordex-africa-publications 2020 AUS http://cordex-australasia.wikidot.com/publications 2011 EUR https://euro-cordex.net/060380/index.php.en 2018 MED https://www.medcordex.eu/publications.php 2025 MENA https://mena-cordex.cyi.ac.cy/index.php/publications 2023 NAM https://na-cordex.org/bibliography.html 2023 SEA https://www.ukm.my/seaclid-cordex/Publications/Publications_2021.html 2021 FPS-CPTP http://rcg.gvc.gu.se/cordex_fps_cptp/Publications.html 2024 FPS-CONV https://wcrp-cordex.github.io/cordex-fps-conv/publications.html 2025 Table 6: Links to CORDEX publication lists for different initiatives. The last column shows the date of the latest publication reported. Another option would be to use a proper bibliographic tool, such as the open source Zotero.org, which can expose the publications under different collections (e.g. per domain) and has an API to retrieve the data programmatically. A central collection of publications in such a tool would allow to track the CORDEX science by reporting the publications only once. The curation of the publications could be centralized in the regional POCs, who could gather into a common tool the info they are already collecting. And the other way around, help could be offered to the POCs to sync the information back into their regional websites in their preferred way. A similar attempt is in place in CMIP644. We created an example with this proposal45, which is already feeding automatically the publication lists of the FPS-CONV and Med-CORDEX46. It would be up to the regional communities to decide what is considered a CORDEX publication for their domain. Different categories could be created to monitor separately e.g. informal analyses, model 46 https://med-cordex.github.io/publications/publications.html 45 https://www.zotero.org/groups/5816477/wcrp-cordex/library 44 https://wcrp-cmip.org/publication-library 41 Figure 4: Schematic representation of the timeline for steps in preparation for CORDEX–CMIP7–AFT to meet the deadlines of IPCC AR7. The bars representing the TF-proposed steps span the development and maintenance phases of each step, while the CORDEX stages show preparation, simulation, and science phases (see legend). Activities with flat ends are expected to finish at the corresponding date, while arrow-headed ones may extend beyond that point and should preferably be ready by then. The red line marks the time of writing: left of this line, dates are accurate; right of the line, dates are estimates. 48 References Baño-Medina, J., Manzanas, R., Cimadevilla, E., Fernández, J., González-Abad, J., Cofiño, A. S., and Gutiérrez, J. M.: Downscaling multi-model climate projection ensembles with Deep Learning (DeepESD): contribution to CORDEX EUR-44, Geoscientific Model Development, 15, 6747–6758, https://doi.org/10.5194/gmd-15−6747−2022 , 2022. Chang, P., Zhang, S., Danabasoglu, G., Yeager, S. G., Fu, H., Wang, H., Castruccio, F. S., Chen, Y., Edwards, J., Fu, D., Jia, Y., Laurindo, L. C., Liu, X., Rosenbloom, N., Small, R. J., Xu, G., Zeng, Y., Zhang, Q., Bacmeister, J., Bailey, D. A., Duan, X., DuVivier, A. K., Li, D., Li, Y., Neale, R., Stössel, A., Wang, L., Zhuang, Y., Baker, A., Bates, S., Dennis, J., Diao, X., Gan, B., Gopal, A., Jia, D., Jing, Z., Ma, X., Saravanan, R., Strand, W. G., Tao, J., Yang, H., Wang, X., Wei, Z., and Wu, L.: An Unprecedented Set of High-Resolution Earth System Simulations for Understanding Multiscale Interactions in Climate Variability and Change, J. Adv. Model. Earth Sy., 12, e2020MS002298, https://doi.org/10.1029/2020MS002298, 2020. Doury, A., Somot, S., Gadat, S., Ribes, A., and Corre, L.: Regional climate model emulator based on deep learning: concept and first evaluation of a novel hybrid downscaling approach, Clim Dyn, 60, 1751–1779, https://doi.org/10.1007/s00382−022−06343−9, 2023. Dunne, J. P., Hewitt, H. T., Arblaster, J., Bonou, F., Boucher, O., Cavazos, T., Durack, P. J., Hassler, B., Juckes, M., Miyakawa, T., Mizielinski, M., Naik, V., Nicholls, Z., O’Rourke, E., Pincus, R., Sanderson, B. M., Simpson, I. R., and Taylor, K. E.: An evolving Coupled Model Intercomparison Project phase 7 (CMIP7) and Fast Track in support of future climate assessment, GMD, 18, 6671–6700, https://doi.org/10.5194/gmd-18−6671−2025, 2025. Giorgi, F., Coppola, E., Jacob, D., Teichmann, C., Omar, S. A., Ashfaq, M., Ban, N., Bülow, K., Bukovsky, M., Buntemeyer, L., Cavazos, T., Ciarlo`, J., Rocha, R. P. da, Das, S., Sante, F. di, Evans, J. P., Gao, X., Giuliani, G., Glazer, R. H., Hoffmann, P., Im, E.-S., Langendijk, G., Lierhammer, L., Llopart, M., Mueller, S., Luna-Nino, R., Nogherotto, R., Pichelli, E., Raffaele, F., Reboita, M., Rechid, D., Remedio, A., Remke, T., Sawadogo, W., Sieck, K., Torres-Alavez, J. A., and Weber, T.: The CORDEX-CORE EXP-I Initiative: Description and Highlight Results from the Initial Analysis, Bulletin of the American Meteorological Society, 103, E293–E310, https://doi.org/10.1175/BAMS-D-21−0119.1, 2022. Grose, M. R., Narsey, S., Trancoso, R., Mackallah, C., Delage, F., Dowdy, A., Di Virgilio, G., Watterson, I., Dobrohotoff, P., Rashid, H. A., Rauniyar, S., Henley, B., Thatcher, M., Syktus, J., Abramowitz, G., Evans, J. P., Su, C.-H., and Takbash, A.: A CMIP6−based multi-model downscaling ensemble to underpin climate change services in Australia, Climate Services, 30, 100368, https://doi.org/10.1016/j.cliser.2023.100368, 2023. 49 Gutiérrez, J. M., Cavazos, T., Evans, J., Nikulin, G., Somot, S., Maraun, D., Benestad, R. E., Hewitson, B., and Bettolli, M. L.: The Future scientific challenges for CORDEX: Empirical Statistical Downscaling (ESD), 2022. https://cordex.org/wp-content/uploads/2022/08/White-Paper-ESD.pdf Gutowski Jr., W. J., Giorgi, F., Timbal, B., Frigon, A., Jacob, D., Kang, H.-S., Raghavan, K., Lee, B., Lennard, C., Nikulin, G., O’Rourke, E., Rixen, M., Solman, S., Stephenson, T., and Tangang, F.: WCRP COordinated Regional Downscaling EXperiment (CORDEX): a diagnostic MIP for CMIP6, Geosci. Model Dev., 9, 4087–4095, https://doi.org/10.5194/gmd-9−4087−2016, 2016. Haarsma, R. J., Roberts, M. J., Vidale, P. L., Senior, C. A., Bellucci, A., Bao, Q., Chang, P., Corti, S., Fučkar, N. S., Guemas, V., von Hardenberg, J., Hazeleger, W., Kodama, C., Koenigk, T., Leung, L. R., Lu, J., Luo, J.-J., Mao, J., Mizielinski, M. S., Mizuta, R., Nobre, P., Satoh, M., Scoccimarro, E., Semmler, T., Small, J., and von Storch, J.-S.: High Resolution Model Intercomparison Project (HighResMIP v1.0) for CMIP6, Geosci. Model Dev., 9, 4185–4208, https://doi.org/10.5194/gmd-9−4185−2016, 2016. IPCC: Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change, edited by: Mason-Delmotte, V., Zhai, P., Pirani, A., Connors, S. L., Pean, C., Berger, S., Caud, N., Chen, Y., Goldfarb, L., Gomis, M. I., Huang, M., Leitzell, K., Lonnoy, E., Matthews, J. B. R., Maycock, K., Waterfield, T., Yelekci, O., Yu, R., and Zhou, B., Cambridge University Press, https://doi.org/10.1017/9781009157896, 2021. Juckes, M., Taylor, K. E., Durack, P. J., Lawrence, B., Mizielinski, M. S., Pamment, A., Peterschmitt, J.-Y., Rixen, M., and Sénési, S.: The CMIP6 Data Request (DREQ, version 01.00.31), Geosci. Model Dev., 13, 201–224, https://doi.org/10.5194/gmd-13−201−2020, 2020. Kopp, G. and Lean, J. L.: A new, lower value of total solar irradiance: Evidence and climate significance, Geophysical Research Letters, 38, https://doi.org/10.1029/2010GL045777, 2011. Roberts et al., (2025). High-Resolution Model Intercomparison Project phase 2 (HighResMIP2) towards CMIP7. https://doi.org/10.5194/gmd-18−1307−2025. Ruane, A. C., Pascoe, C. L., Teichmann, C., Brayshaw, D. J., Buontempo, C., Diouf, I., Fernandez, J., Gonzalez, P. L. M., Hassler, B., Hernaman, V., Im, U., Iovino, D., Juckes, M., Lake, I. L., Lam, T., Lin, X., Mao, J., Nazarian, N., Parey, S., Roy, I., Tseng, W.-L., Turner, B., Wiebe, A., Zhao, L., and Zurell, D.: CMIP7 Data Request: Impacts and Adaptation Priorities and Opportunities, EGUsphere, 1–61, https://doi.org/10.5194/egusphere-2025−3408, 2025. Ruti, P. M., Somot, S., Giorgi, F., Dubois, C., Flaounas, E., Obermann, A., Dell’Aquila, A., Pisacane, G., Harzallah, A., Lombardi, E., Ahrens, B., Akhtar, N., Alias, A., Arsouze, T., Aznar, R., Bastin, S., Bartholy, J., Béranger, K., Beuvier, J., Bouffies-Cloché, S., Brauch, J., Cabos, W., Calmanti, S., Calvet, J.-C., Carillo, A., Conte, D., Coppola, E., Djurdjevic, V., Drobinski, P., Elizalde-Arellano, A., Gaertner, M., Galàn, P., Gallardo, C., Gualdi, S., Goncalves, M., Jorba, O., Jordà, G., L’Heveder, B., Lebeaupin-Brossier, C., Li, L., Liguori, G., Lionello, P., Maciàs, D., Nabat, P., Önol, B., Raikovic, B., Ramage, K., Sevault, F., Sannino, G., Struglia, M. V., Sanna, A., 50 Torma, C., and Vervatis, V.: Med-CORDEX Initiative for Mediterranean Climate Studies, Bulletin of the American Meteorological Society, 97, 1187–1208, https://doi.org/10.1175/bams-d-14−00176.1, 2016. Sanderson, B. M., Booth, B. B. B., Dunne, J., Eyring, V., Fisher, R. A., Friedlingstein, P., Gidden, M. J., Hajima, T., Jones, C. D., Jones, C., King, A., Koven, C. D., Lawrence, D. M., Lowe, J., Mengis, N., Peters, G. P., Rogelj, J., Smith, C., Snyder, A. C., Simpson, I. R., Swann, A. L. S., Tebaldi, C., Ilyina, T., Schleussner, C.-F., Seferian, R., Samset, B. H., van Vuuren, D., and Zaehle, S.: The need for carbon emissions-driven climate projections in CMIP7, EGUsphere, https://doi.org/10.5194/egusphere-2023−2127, 2024. Scaife, A. A., Camp, J., Comer, R., Davis, P., Dunstone, N., Gordon, M., MacLachlan, C., Martin, N., Nie, Y., Ren, H.-L., Roberts, M., Robinson, W., Smith, D., and Vidale, P. L.: Does increased atmospheric resolution improve seasonal climate predictions?, Atmos. Sci. Lett., 20, e922, https://doi.org/10.1002/asl.922, 2019. Solman, S., Jacob, D., Frigon, A., Teichmann, C., and Rixen, M.: The future scientific challenges for CORDEX, 2021. https://cordex.org/wp-content/uploads/2022/08/The-future-of-CORDEX-MAY-17−2021_2.pdf van Vuuren, D., O’Neill, B., Tebaldi, C., Chini, L., Friedlingstein, P., Hasegawa, T., Riahi, K., Sanderson, B., Govindasamy, B., Bauer, N., Eyring, V., Fall, C., Frieler, K., Gidden, M., Gohar, L., Jones, A., King, A., Knutti, R., Kriegler, E., Lawrence, P., Lennard, C., Lowe, J., Mathison, C., Mehmood, S., Prado, L., Zhang, Q., Rose, S., Ruane, A., Schleussner, C.-F., Seferian, R., Sillmann, J., Smith, C., Sörensson, A., Panickal, S., Tachiiri, K., Vaughan, N., Vishwanathan, S., Yokohata, T., and Ziehn, T.: The Scenario Model Intercomparison Project for CMIP7 (ScenarioMIP-CMIP7), EGUsphere, 1–38, https://doi.org/10.5194/egusphere-2024−3765, 2025. 51 Appendix 1: Survey on potential commitments for CORDEX-CMIP7 in time for AR7 This survey was launched on April 15th, 2025, as part of the work in the CORDEX Task Force (TF) in preparation for CORDEX-CMIP7. The questions collect prior knowledge of existing national or institutional commitments to downscale the CMIP7 Fast Track in time for AR7 or other assessments/adaptation plans in the short term. It also covers some details on model development plans and longer term commitment for CORDEX-CMIP7, beyond the IPCC AR7 deadlines. All data collected are treated confidentially, so here we share only aggregated statistics. In any case, the answers from the modelling groups are non-binding and for exploratory purposes only. Actual, detailed CORDEX-CMIP7 plans will be collected in the future (see Section 5.4). The survey collects some initial information for future reference (full name, contact email, institution) and then some multiple-choice and free text questions collect detailed information which is summarized next for the 24 responses received as of Jul 30th, 2025. No discussion of the results is included in this Appendix, but key figures are replicated and discussed in the main document where appropriate. Questions Downscaling model type Select just one. If multiple models are used in your institution/group, please, fill out one survey per model. ESD (Empirical-statistical downscaling) RCM-emulator ARCM (prescribed-SST RCM) AORCM (ocean-coupled RCM) AGCM (prescribed-SST GCM, including variable-resolution models) AOGCM (ocean-coupled GCM, including variable-resolution models) Other... 52 What were the main challenges you faced during the downscaling of CMIP6 data? Please select up to three of the most significant barriers. We did not participate in CORDEX-CMIP6 Insufficient computing resources Limited data storage capacity CMORization process Lack of funding Insufficient personnel Ongoing model development Other... 53 Do you have a newer version of your downscaling model (as compared to CORDEX-CMIP6) ready to be used? Yes No If yes, please, comment briefly how has your model evolved Free text. Answers (8): ● Inclusion of regional carbon cycle and double moment cloud microphysics. ● New urban parameterization ● new dynamical core 54 ● Resolution change through New TCO Grid ● Now able to activate additional components, such as the urban model ● We have changed the RCM from RegCM4 to the coupled model, COAWST. ● Ocean coupling, improvements in model setup and parameterizations ● Various improvements consistent with the Hadley Centre GCM configuration. Are you planning to downscale CMIP7 Fast Track simulations in time for AR7 (downscaled data available by mid-2027) over any of the CORDEX domains? Yes Yes, but only if there’s a coordinated effort in the domain Depends on funding, which is likely to be available Depends on funding, which is not likely to be available No Other … Are you planning to downscale CMIP7 simulations at some point (after AR7) over any of the CORDEX domains? Yes Yes, but only if there’s a coordinated effort in the domain Depends on funding, which is likely to be available Depends on funding, which is not likely to be available No Other … 55 Which domain(s) would you be planning to downscale? Check all that apply. Do you have an in-house or preferred CMIP7 GCM to provide input for your simulations? 56 If yes, which one(s)? Do you have a commitment to downscale a particular scenario or scenario family? 57 This document has been written as a guideline for such WGCI, with plenty of details for each step, including past actions and strategies and many links to further information. It has been delivered way ahead of the TF appointed deadline to make room for the WG call while the TF is still active, so we transition smoothly to the new team. We offer ourselves for a candidate pre-selection and to make a draft proposal of the WG composition to the SAT. One step in preparation for CORDEX-CMIP7 that could hardly be handled by the WGCI just proposed above is the experiment design (Section 5.2). This requires very experienced CORDEX representatives with expertise in different fields. Time-availability would also be a constraint, but this alternative WG could go on as an executive science body between the SAT and the community, to actually carry out the developments, assisted by the more technical WGCI. If POCs are encouraged to be part of this WG, it would also serve the purpose of enhancing cross-domain interaction. PoCs are also much welcome in the WGCI, so the regional communities would ideally have representatives in the science advisory, executive and technical bodies. Alternatively, the experiment design could be assigned to a TT dedicated to this task with a given deadline and requiring members with deeper CORDEX expertise in different fields. All these WGs and TTs should be active in a very short time in order to meet the tight deadlines of CORDEX-CMIP7−AFT (see Section 7). 64