A global CORDEX-based dataset delineating urban areas and their surroundings to assess climate change on megacities

A global CORDEX-based dataset delineating urban areas and their surroundings to assess climate change on megacities Javier Diez-Sierra1, Yaiza Quintana1, Gaby S. Langendijk2, Josipa Milovac1, Matthias Demuzere3, Rita Nogherotto4,5, Joni-Pekka Pietikäinen6, Diana Rechid6, Natalia Zazulie5,7, Silvina A. Solman8, Jesús Fernández1 Preprint of the work submitted to Scientific Data https://doi.org/10.5281/zenodo.17367975 1. Instituto de Física de Cantabria (IFCA), CSIC-Universidad de Cantabria, Santander, Spain 2. Climate Adaptation and Disaster Risk Department, Deltares, PO Box 177, 2600 MH Delft, the Netherlands 3. B-Kode VOF, Ghent, Belgium 4. The Institute of Atmospheric Sciences and Climate (CNR-ISAC), Italy 5. The Abdus Salam International Centre for Theoretical Physics (ICTP), Trieste, Italy 6. Climate Service Center Germany (GERICS), Helmholtz-Zentrum Hereon, Fischertwiete 1, 20095 Hamburg, Germany 7. Departamento de Ciencias de la Atmósfera y los Océanos, Facultad de Ciencias Exactas y Naturales, Universidad de Buenos Aires, Consejo Nacional de Investigaciones Científicas y Técnicas, Buenos Aires, Argentina 8. Departamento de Ciencias de la Atmósfera y los Océanos, Facultad de Ciencias Exactas y Naturales, Universidad de Buenos Aires, Centro de Investigaciones del Mar y la Atmósfera, Consejo Nacional de Investigaciones Científicas y Técnicas, Instituto Franco-Argentino de Estudios sobre el Clima y sus Impactos (IRL 3351 IFAECI/CNRS-IRD-UBA), Buenos Aires, Argentina Abstract We present a global dataset of urban areas and their rural surroundings, developed within the framework of the CORDEX Flagship Pilot Study on Urban Environments and Regional Climate Change. The dataset is derived from model-specific urban fraction variables and additional static inputs. Urban and rural surrounding areas are delineated using Regional Climate Model (RCM) simulations from the global CORDEX-CORE and European EURO-CORDEX initiatives, focusing on a representative set of megacities worldwide. The analysis was conducted at horizontal resolutions of 25 km globally and 12.5 km for Europe. To facilitate future applications, we provide a Python-based workflow that can be extended for the analysis of additional cities and RCMs, including tools for evaluating the urban climate island effect. The dataset and tools, available via Zenodo and GitHub, offer a consistent and reproducible approach for assessing urban climate change in current and upcoming regional climate projections. This constitutes the first global RCM-based database of urban/rural areas, providing a foundation for future high-resolution model data analysis efforts, such as studies using convection-permitting simulations. Background & Summary Understanding how climate conditions vary within urban areas has important implications for stakeholders developing adaptation strategies 1. The Urban Heat Island (UHI) effect refers to the phenomenon whereby urban areas experience significantly higher temperatures than their surrounding suburban and rural regions 2,3. This effect can lead to several negative consequences, including increased heat stress and health risks, higher energy consumption, decreased labor productivity and elevated air pollution levels, among others4–8. In addition to temperature, dense urban environments can influence other climate variables. For instance, precipitation may increase over and downwind of highly urbanized areas 9–12; relative humidity is often lower in urban areas compared to rural locations 13–16; and wind speeds can be significantly altered in cities relative to their surroundings 17–20. From a climate perspective, the UHI effect amplifies the frequency, duration, and intensity of heat waves beyond what is typically driven by climate change alone 1. Consequently, understanding the combined impact of urban climate, urbanisation and climate change on cities worldwide is particularly relevant, especially given that a significant portion of the global population resides in urban areas which is expected to increase in the coming decades 21. Assessing climate change at regional or global scales requires the use of numerical climate models 22,23. Two main modeling approaches have emerged for understanding and analyzing the urban climate 24. First, mesoand microscale urban climate models resolve climate processes at the street-to-city scales (1 m to 1 km). These models simulate short-duration weather events, potentially under climate change conditions, for instance either by adding a specified temperature increase to the model 10,25,26, or using boundary conditions from Global Climate Models (GCM) or Regional Climate Models (RCM) 10, or through statistical downscaling methods 27–29. However, due to limitations in domain size, these models often struggle to fully capture the dynamic interactions between the city and its regional surroundings, as well as the combined effects of climate change and urbanization over climatological timescales 30,31. Second, RCMs have undergone an increase in grid resolution in recent years, going down to the kilometer scale (1-4 km), which allows them to resolve smaller scale processes and features of the Earth’s surface 32,33. These so-called convection-permitting RCMs (CPRCM) represent a larger proportion of grid boxes categorized as “urban”, often with higher urban fraction values. They show strong potential for simulating urban climates over long timescales, from decades to even a century 34,35. The representation of urban areas in RCMs varies in complexity 36, ranging from simple bulk urban parameterizations 37, to single-layer urban canopy models 38, to more advanced multilayer models such as the Building Effect Parameterization/Building Energy Model (BEP/BEM) 39. In parallel with the development of CPRCMs, several modeling centres have started to develop hectometric-scale weather and climate models for operational use 40. Unfortunately, the enormous computational cost inherent to the CPRCMs limits their application to relatively small geographic areas, making it particularly challenging to perform global-scale analyses 41,42. On the other hand, the coarse spatial resolution (~100 km) of long-term GCM ensemble simulations often prevents these models from capturing urban areas 16,30,43. Fortunately, RCM projections from initiatives such as the COordinated Regional Downscaling EXperiment – COmmon Regional Experiment (CORDEX-CORE) 44–48 and EURO-CORDEX 49–52 provide an opportunity to analyze urban climate change at finer horizontal resolutions: 25 km over all continental CORDEX domains and 12.5 km over the European domain. Despite these advancements, simulations from the CORDEX experiments still face notable limitations: urban schemes are often deactivated 30, spatial resolution remains insufficient to fully capture urban processes, the representation of urban phenomena is often incomplete or too simplistic 43,53, and land use changes including urbanisation or greening of cities are not considered to date. Nevertheless, CORDEX projections remain the only globally available source of regional climate change information based on dynamical downscaling at the time, and they have been used as a reference for the regional analyses in the Intergovernmental Panel on Climate (IPCC)’s Sixth Assessment Report 54. CORDEX-CORE simulations enable the climate research community to evaluate their limitations in representing urban environments across the globe and provide a critical foundation for improving future modeling frameworks 55. Looking ahead, new CORDEX simulations nested into Coupled Model Intercomparison Project Phase 6 (CMIP6) projections 56 and particularly the next generation of RCM ensembles at convection-permitting resolutions 32,57–59, along with advanced downscaling techniques based on deep learning 28, present a promising framework for future analyses. In this context, the CORDEX Flagship Pilot Study (FPS) on URBan environments and Regional Climate Change (URB-RCC) was launched in May 2021, with the aim of investigating the impact of cities on regional climates and vice versa 24. One of the key objectives of the FPS URB-RCC is to assess the capability of existing CORDEX-CORE simulations to represent urban climates. To this end, a selection of representative cities worldwide was made based on various criteria, ensuring a diverse and globally-relevant sample of urban areas 55. Langendijk et al. (2025) 55 show that, although limited, CORDEX-CORE models are capable of reproducing an urban imprint across various megacities. Their findings indicate that the models’ ability to simulate the UHI effect improved significantly when more sophisticated urban schemes were implemented and when the spatial resolution was increased from 25 km to 12.5 km, as in CORDEX-EUR-11. From the perspective of current and emerging initiatives aimed at providing regional climate change simulations, it is essential to establish a robust and consistent methodology for defining urban areas and their rural surroundings. This methodology must be applicable across different spatial resolutions and adaptable to the specific geographic characteristics of each city, following a land-based approach. Consequently, the methodology relies on a set of input-model static variables to characterize urban and rural areas within the RCM spatial framework, such as urban fraction (i.e., the percentage of the grid cell occupied by urban area), orography, and the land-sea mask, although additional variables could also be incorporated to assess the effect of urbanisation and land use changes. Such an approach is necessary because the representation of urban land cover in numerical climate models often diverges from administrative city boundaries due to the coarse horizontal resolution and the use of varying underlying land cover datasets. This article presents: (1) a database of urban areas and their rural surroundings for CORDEX-CORE (and CORDEX-EUR-11 for the European domain), covering the cities selected within the framework of the CORDEX FPS URB-RCC, which represents the first global resource for evaluating urban climate change based on two different RCMs; and (2) a Python-based workflow for delineating urban areas and their rural surroundings, which includes additional functionalities for analyzing and visualizing the UHI effect. These tools are specifically designed to be applicable to numerical climate model outputs. The overall objective of this work is to establish a collaborative and consistent framework for assessing urban climate change using reproducible methodologies. Methods Input data CORDEX simulations CORDEX (https://cordex.org) represents the first global initiative, established under the auspices of the World Climate Research Programme (WCRP), to coordinate high-resolution regional climate projections within a unified experimental framework 46,60. CORDEX provides spatially detailed climate change projections from a large ensemble of RCMs applied over large continental areas, at horizontal grid spacing ranging from 12 to 50 km. In this study, CORDEX-CORE simulations 45 are used to define urban areas and their rural surroundings areas for selected cities, as part of the first phase of CORDEX FPS URB-RCC (see the Section “Methods”). CORDEX-CORE simulations constitute, to date, the only ensemble of RCMs provided under a common protocol that cover nearly all continental areas of the world (across nine CORDEX domains) at a horizontal resolution of 25 km. The remaining CORDEX-CMIP5-based simulations were performed at coarser resolutions, typically 50 km for most domains, except for Europe, where a higher resolution of 12.5 km is available. The simulations include two RCMs, RegCM and REMO, at a spatial resolution of 25 km (0.22º). To complement this, CORDEX-EUR-11 simulations, with the same RCMs (i.e. REMO and RegCM) but at a higher resolution of 12.5 km (0.11º), are employed for the European cities. The availability of data at both 25 km and 12.5 km resolutions enables a comparative assessment of the added value of increased horizontal resolution in capturing urban climate features. This study focuses exclusively on simulations of the evaluation scenario nested to ERA-Interim reanalysis 61. Further information on the specific RCM versions used for each domain is provided in Langendijk et al. (2025) 55. The regional climate models RegCM (several variants of RegCM4) and REMO (REMO2015) represent urban areas differently. The REMO version (REMO2015) treats urban surfaces as purely sealed/impervious areas, whereas RegCM employs the CLM Urban (CLMU) model, a single-layer urban canopy model in which the urban fraction is further decomposed into three classes (see Langendijk et al., 2025 55 for a detailed description of the RCMs). For consistency, we aim to align the representation of urban areas across models. In REMO, urban areas are represented as rock surfaces simulating the impervious characteristics of cities, whereas RegCM includes an urban land unit within each grid cell that encompasses both impervious and pervious areas. The different urban densities in each land unit have different impervious area fractions. To enable comparison with REMO, the total impervious surface within each RegCM grid cell is extracted (see Langendijk et al., 2025 55). This study provides both the rural/urban database derived from the original urban area fraction variable (sfturf) and from the impervious area fraction (sftimf). All analyses presented in this article are based on the sftimf variable; however we refer to it generically as the “urban fraction” (UF) to avoid confusion, as this is the commonly used term for the variable. Selected cities A subset of 41 cities, representing a diverse and heterogeneous sample of urban areas worldwide, was selected following the work done in the CORDEX FPS URB-RCC and forms the basis of the global dataset of urban areas and their rural surroundings developed in this study (see Figure 5). Given the relatively coarse resolution of CORDEX-CORE (25 km), only large urban areas were included. The selection criteria consider city size, geographic characteristics (e.g., coastal, inland, mountainous, or regions with complex terrain), global balance across CORDEX domains, climate characteristics, and climate impact. A detailed description of the selection criteria and the selected cities is provided in Langendijk et al. (2025)55. Note that for RegCM a 40% cut-off value applied to the urban fraction, as well as some CORDEX domains (i.e. NAM-22 and EAS-22) do not include urban areas. This cut-off implies that cells with less than 40 % urban coverage are not classified as urban, effectively excluding most cities. An exception is the EUR-11 domain, where no cut-off value is used. Static (time-invariant) variables Urban or impervious fraction, orography, and land area fraction are the static (time-invariant) variables used in this work as input data to delineate urban and surrounding areas for cities around the world. Orography and land area fraction are part of the mandatory core set of model output variables defined in the CORDEX-CMIP5 downscaling protocols and are therefore publicly available through the Earth System Grid Federation (ESGF; https://esgf-metagrid.cloud.dkrz.de/search)62. In contrast, the urban fraction variable was neither designated as a core model output, nor included in Tier 1 (core or mandatory) or Tier 2 (optional or additional), and is only available upon request from the modeling centers. Fortunately, in the upcoming CORDEX-CMIP6 experiment, the urban fraction (sfturf) variable is classified as Tier 2, enhancing the FAIR principles 63 and enabling future analyses of urban climate with a larger ensemble of models. In this work, we use urban and impervious area fractions from Langendijk et al. (2025)64, who collected (and post processed) them from the REMO and RegCM CORDEX-CMIP5 modeling centers and made them publicly available on Zenodo (https://doi.org/10.5281/zenodo.15700267) 64. This is the first time that such urban fraction data (and derived impervious data for RegCM) have been made publicly available for CORDEX RCM data at both global and EURO-CORDEX scales. The dataset, provided in NetCDF (Network Common Data Form) format, complies with both the CORDEX archive specifications 65 and the Climate and Forecast (CF) metadata conventions. This dataset, together with the workflow presented in this study, enables the urban climate research community to investigate urban climate under future climate conditions using CORDEX simulations based on a minimal ensemble of two RCMs within a consistent and comparable framework. Algorithm for delineating urban areas and their surroundings Most studies analyze the UHI effect using satellite-derived land surface temperature (LST) and land use and land cover (LULC) data 66,67. However, urban representation in numerical climate models often differs from administrative city boundaries due to their typically coarse horizontal resolution and the simplification of LULC categories into a limited set of types interpretable by the models. Additionally, land surface representations vary among RCMs, adding further complexity to their interpretation and intercomparison 68. A common approach for delineating urban areas is the City Clustering Algorithm (CCA), developed by Rozenfeld et al. (2008)69, which predicts city growth based on population data. This method has been widely applied in UHI studies because it effectively captures the spatial extent of urban areas. However, since population data are not used as a parameter in climate models, this algorithm can be applied using land-use data instead 70–72. The CCA utilizes a parameter to define the maximum distance at which grid cells are considered connected and belong to the same urban cluster 73. Previous studies utilizing climate model outputs typically define a city as the grid cell within the model that is nearest to its center 74,75. To define rural surrounding areas, a common method involves generating consecutive layers of cell-width buffers around the urban cluster. The most widely used approach is the Boundary Generation Algorithm (BGA)73, which iteratively expands a rural buffer around the city until it reaches an area approximately equal to that of the urban region. Simpler methods often focus on individual cities rather than applying a consistent domain-wide methodology. In such cases, urban areas are identified within a predefined region using LULC-based thresholds, while the remaining grid cells are classified as rural –either explicitly or based on distance-based approaches 35,76–79. Although approaches such as the CCA and BGA algorithms are widely applied for delineating urban and rural areas, certain parameters (e.g., the maximum distance at which urban grid cells are considered connected in CCA, or the definition of potential areas for rural expansion in BGA) can significantly affect the results, particularly given the coarse horizontal resolution of RCMs, which often necessitates city-specific adjustments to achieve an accurate representation of urban areas. Algorithm description The methodology proposed in this study relies on three static variables commonly available in most RCM outputs: urban fraction (sfturf or sftimf), orography (orog), and land area fraction (sftlf). The algorithm can be briefly described as follows. A minimum threshold for the UF determines the grid cells representing the city in the model. Potential rural surrounding areas are then determined based on three main criteria: (1) grid cells must have UF values below a specified threshold; (2) large water bodies (lakes, oceans, and rivers) are excluded via a minimum land area fraction threshold; and (3) grid cells with an elevation difference above a threshold with respect to the urban area are excluded to avoid the effect of altitude on temperature (i.e., adiabatic lapse rate). Grid cells complying with these criteria are selected as candidate rural surroundings areas. The final rural surrounding area is obtained from this candidate grid cells through an iterative morphological dilation process expanding outward from the urban cells. Iterations stop when the number of rural cells reaches a predefined ratio relative to the number of urban cells. Along with the static variables (UF, orog and sftlf), the algorithm uses several parameters (see Table 1) to determine which grid cells are classified as urban or rural. First, the location of the city of interest (“lon_city” and “lat_city”) and the study area boundaries of a larger area surrounding the city (“lon_lim” and “lat_lim”) must be defined in geographic coordinates. Only grid cells within the predefined study area are eligible to be selected as either urban or rural. Then, the UF threshold (“urban_th”) determines which grid cells are classified as urban. Cells with UF values greater than this threshold are considered urban cells. Urban areas not connected to the city’s core can be excluded using the “min_city_size” parameter. This filters out urban clusters, classifying them as neither urban nor rural, and retains the main urban cluster nearest to the coordinates defined by “lon_city” and “lat_city”. A threshold for the urban fraction in the surroundings (“urban_sur_th”) is used to create a buffer zone around the urban area. Cells with UF values between “urban_sur_th” and “urban_th” may be influenced by the urban climate and, therefore, are excluded from the rural mask. This parameter is particularly relevant for high-resolution climate models, where relatively high values of “urban_th” can be used and, thus, significant urban fractions might affect the rural surroundings. To define potentially rural cells, the algorithm uses orography and land area fraction variables to apply additional filters. The parameter “orog_diff” is used to exclude surrounding mountainous areas where elevation difference relative to the minimum or maximum urban cells exceeds a user-defined threshold (in metres). Similarly, large water bodies, such as lakes, oceans and large rivers, are excluded using a user-defined threshold on the land area fraction (“sftlf_th”). Note that this parameter also affects the delineation of urban areas. Once candidate rural cells are identified, an iterative morphological dilation process is applied to grow the surrounding area outward from the urban core. The ratio of rural to urban cells is controlled by the “ratio_r2u” parameter. The morphological dilation function is implemented using the scikit-image Python package 80. This function assigns to a pixel the maximum value found over all pixel values within its surrounding local neighborhood. The neighborhood is defined by a footprint, which is a binary mask (a small matrix of 0s and 1s) that specifies the shape and size of the neighborhood by indicating which neighboring pixels are included in the operation. Pixels corresponding to 1s in the footprint are considered part of the neighborhood, while those with 0s are excluded. Two types of footprints are implemented: a cross-shaped footprint that considers 4-connected neighbors (cells sharing the edge), and a square-shaped footprint that includes both edgeand corner-connected neighbors (8-connected). In each iteration, the cross-shaped footprint is applied first. If no new rural cells are added, the square footprint is then applied. This dual-step approach is necessary because, in some cities, the cross-shaped footprint fails to expand the masks when using coarse-resolution data. In every iteration, grid cells excluded due to elevation, water bodies, or being classified as urban are ignored. The process terminates once the number of rural cells reaches the desired rural-to-urban ratio specified by the “ratio_r2u” parameter. Table 1. Description of the hyperparameters implemented in the algorithm. Hyperparameter Description “lon_city” and “lat_city” Longitude and latitude of the city center. “lon_lim” and “lat_lim” Geographic boundaries of the study area (incl. city surroundings) relative to the city center (“lon_city” and “lat_city”). Grid cells outside these limits (lon_city±lon_lim and lat_city±lat_lim) are excluded from the analysis. “urban_th” Urban fraction threshold (%). Grid cells with urban fraction values above this threshold are classified as urban cells. “urban_sur_th” Urban surrounding threshold (%). Grid cells with urban fraction values below this threshold are candidates for rural surroundings. Defaults to “urban_th”. “orog_diff” Maximum elevation difference (in meter) relative to the range (max-min) urban cell elevations (urban min elev. - orog_diff < rural elev. < urban max elev + orog_diff). Pixels exceeding this difference are excluded. “sftlf_th” Minimum land area fraction (%) required to include a grid cell in the analysis. “min_city_size” Minimum size (in number of edge-connected cells) for urban clusters to be retained. Urban clusters are excluded, except for the main cluster nearest to “lon_city” and “lat_city”, which is always retained. “ratio_r2u” Ratio of rural to urban grid cells. The iterative dilation process stops once this ratio is achieved. Hyperparameters selection criteria The algorithm presented here includes several hyperparameters that must be adjusted on a case-by-case basis to ensure optimal performance. Most of these hyperparameters, such as “urban_th”, “urban_sur_th”, “min_city_size” and “ratio_r2u”, are designed to accommodate different horizontal resolutions, ranging from cases where a city is represented by only a few grid cells typically for coarse spatial resolutions to others where dozens of grid cells represent the city, for instance at fine spatial resolutions. This makes the algorithm suitable for analysing RCM data across spatial scales, and it has been validated for resolutions ranging from 50 - 2 km. Other hyperparameters, such as “orog_diff” and “sftlf_th”, depend on the specific geographic characteristics of each city. The city-specific hyperparameters used to generate the urban/rural mask database for CORDEX-CORE and CORDEX-EUR-11 are provided in Table 2. Table 2. Hyperparameters used to generate the global dataset of urban areas and their rural surroundings. Some hyperparameters are common across cities and spatial resolutions, while others are defined ad hoc for each city and are included in the accompanying YAML file available in the GitHub repository referenced in the “Code Availability” section. Hyperparameter CORDEX-CORE CORDEX-EUR-11 lon_city and lat_city See YAML file lon_lim and lat_lim Typically, “lon_lim” = 1 and “lat_lim” = 1, but some cities require higher limits urban_th 10 % 40% urban_sur_th None 10% orog_diff Typically, “orog_diff” = 100 m, but some cities require higher limits Typically, “sftlf_th” = 70 %, but some cities require lower values to include any urban cell (see YAML file) sftlf_th min_city_size See YAML file ratio_r2u 2 The representation of urban environments is highly sensitive to both the horizontal resolution of the data and the urban fraction threshold applied. Commonly, studies use an urban fraction threshold (“urban_th”) between 10% and 30% 35,81. A sensitivity analysis of “urban_th” was conducted on the sample of cities to determine an appropriate value for the CORDEX-CORE models at a 25 km resolution 55. For “urban_th” > 10%, some cities either disappear or are represented by only a single urban grid cell (e.g., Mexico City). Consequently, due to the relatively coarse resolution of the CORDEX-CORE dataset, a UF threshold of 10% was selected for identifying urban grid cells. This choice is consistent with thresholds used in other regional climate modeling studies at similar horizontal resolutions, such as Daniel et al. (2019). For CORDEX-EUR-11, which has a higher horizontal resolution (four times as many grid cells as CORDEX-CORE), a higher UF threshold of 40% was applied. This finer resolution also allows the use of the “urban_sur_th” parameter to exclude cells with intermediate UF values that may still be influenced by urban environments. For CORDEX-EUR-11 cities, ”urban_sur_th” was set to 10%, thereby excluding cells with UF values between 10% and 40% from the rural surroundings, excluding suburban areas and smaller settlements around cities. Oceans, larger lakes, and major rivers were excluded by applying a land area fraction threshold (“sftlf_th”) of 70%. To account for temperature lapse rate effects, surrounding grid cells with an elevation difference of more than 100 meters from the maximum and minimum elevation of urban cells were also excluded (orog_diff = 100). These two parameters (“sftlf_th” and “orog_diff”) were adjusted in certain cases based on the geographic characteristics of individual cities, though we aimed to keep them as consistent as possible to ensure comparability. Finally, the ratio of rural to urban cells (“ratio_r2u”) was set to 2 for both CORDEX-CORE and CORDEX-EUR-11. The selected hyperparameters for each city are specified in the GitHub repository referenced in the “Code Availability” section. These hyperparameters were used to generate the dataset presented in this study. Data Records Dataset of urban areas and their surrounding reference rural regions A dataset of urban areas and their reference rural surroundings has been generated, using the input data (https://doi.org/10.5281/zenodo.15700267) 64 and the algorithm outlined in the Section “Methods”, and published on Zenodo (https://doi.org/10.5281/zenodo.17257489) 82, for the RCMs and cities listed in Langendijk et al. (2025) 55. The dataset consists of a series of NetCDF files –for each combination of RCM (REMO or RegCM), input data (sftimf or sfturf) and city– containing grid-point values of 0, 1, or NaN, representing rural, urban, or unclassified cells, respectively. For the European cities, two separate files were generated for CORDEX-CORE and CORDEX-EUR-11 at a 25 and 12.5 km of horizontal resolution, respectively. The native map projections of each RCM are preserved, and the NetCDF files also include the hyperparameters used in the algorithm (see the Section “Methods”). Filenames and NetCDF metadata are formatted to follow the CORDEX archive specifications 65 and the CF metadata conventions. Each filename includes the following fields, separated by underscores: “urmask” (Urban/RuralMASK) variable including input data (sftimf or sfturf depending on the input data used), CORDEX domain including city, driving GCM, experiment, ensemble member, RCM institution, RCM model name, and frequency. For example, the filename for REMO, CORDEX-EUR-11, and London is: urmask-sftimf_EUR-11-London_ECMWF-ERAINT_evaluation_r1i1p1_GERICS_REMO2015_fx.nc. Figure 1 shows an example showing a snapshot of the contents of this example, including the hyperparameters used during its generation. Figure 1. Urban/rural areas for the city of London using the REMO model from CORDEX-EUR-11. Shaded colors represent the values of the variable ‘urmask’, where 1 (yellow cells) indicates grid cells classified as urban, 0 (purple cells) corresponds to rural cells, and NaN denotes areas not classified as either urban or rural. On the right, the dimensions, coordinates, and attributes of the “urmask” variable are displayed. Technical Validation An in-depth technical validation was carried out for the data records of each city and RCM combination. The analyses include: (1) urban and rural mask representations; (2) an assessment of the UHI intensity; (3) summary information about the number of grid points classified as urban and rural; and (4) a sensitivity analysis evaluating the impact of interpolation and bias adjustment of the raw data (e.g. temperature) on the resulting UHI intensity estimates. The GHS-UCDB dataset83 offers a globally consistent and harmonized representation of urban centers, making it a suitable reference for validating the urban extent represented in the RCMs. It defines “Urban Centres” as polygons based on population and build-up area thresholds, using data from the Global Human Settlement Layer (GHSL) combined with other open datasets. These centers are mapped on a uniform 1x1 km global grid and include various thematic attributes across multiple time periods. While this study uses the GHS-UCDB 2019 version, we acknowledge that a more recent release of GHS-UCDB is available 84 which may be integrated in future versions of the dataset. Jupyter Notebook indicating how hyperparameters configuration effects UHI intensity is included on GitHub: https://github.com/FPS-URB-RCC/urclimask/blob/main/notebooks/paris_across_CORDEX_resolutions.ipynb. Acknowledgements MD is supported by the European Union’s HORIZON Research and Innovation Actions under grant agreement No 101137851, project CARMINE (Climate-Resilient Development Pathways in Metropolitan Regions of Europe, https://www.carmine-project.eu/). JM was funded by the Ministry for the Ecological Transition and the Demographic Challenge (MITECO) and the European Commission NextGenerationEU (Regulation EU 2020/2094), through CSIC's Interdisciplinary Thematic Platform Clima (PTI-Clima). DR was supported by the European Unions HORIZON project FOCAL - Efficient Exploration of Climate Data Locally - under grant agreement No. 101137787. JF acknowledges support from the European Union’s HORIZON Research and Innovation Actions under grant agreement No 101081555 (IMPETUS4CHANGE). JF acknowledges support from the project ATLAS2 (PID2024-162703OB-I00) funded by MCIN/AEI/10.13039/501100011033. GSL and JF acknowledge support from project PROTECT (PID2023-149997OA-I00), funded by MICIU/AEI/10.13039/501100011033 and by ERDF/EU. Author Contributions JD-S contributed to the conceptualization, investigation, formal analysis, software development, validation and writing. GSL, JM, MD and JF contributed to the conceptualization, investigation, formal analysis and writing. YQ contributed to the formal analysis, software development, validation and writing. J-PP, DR, RN, NZ and SAS contributed to writing & discussion. Competing Interests The authors declare that they have no competing interests. References 1. Bader, D. A. et al. Urban climate science. In Climate Change and Cities: Second Assessment Report of the Urban Climate Change Research Network (ARC3.2). https://www.giss.nasa.gov/pubs/abs/ba04500c.html (2018). 2. Oke, T. R. Boundary Layer Climates. (Routledge, London, 1987). doi:10.4324/9780203407219. 3. Rizwan, A. M., Dennis, L. Y. C. & Liu, C. A review on the generation, determination and mitigation of Urban Heat Island. J. Environ. Sci. 20, 120–128 (2008). 4. Flouris, A., Azzi, M., Graczyk, H., Nafradi, B. & Scott, N. Heat at Work: Implications for Safety and Health. A Global Review of the Science, Policy and Practice. ILO. https://www.ilo.org/sites/default/files/2024-07/ILO_OSH_Heatstress-R16.pdf (2024). 5. Ioannou, L. G. et al. The impact of workplace heat and cold on work time loss. J. Occup. Environ. Med. 10.1097/JOM.0000000000003332 (2023) doi:10.1097/JOM.0000000000003332. 6. Kura, B., Verma, S., Ajdari, E. & Iyer, A. Growing Public Health Concerns from Poor Urban Air Quality: Strategies for Sustainable Urban Living. Comput. Water Energy Environ. Eng. 2, 1–9 (2013). 7. Patz, J. A., Campbell-Lendrum, D., Holloway, T. & Foley, J. A. Impact of regional climate change on human health. Nature 438, 310–317 (2005). 8. Yang, Q. et al. A global urban heat island intensity dataset: Generation, comparison, and analysis. Remote Sens. Environ. 312, 114343 (2024). 9. Burian, S. J. & Shepherd, J. M. Effect of urbanization on the diurnal rainfall pattern in Houston. Hydrol. Process. 19, 1089–1103 (2005). 10. Doan, Q.-V. et al. Increased Risk of Extreme Precipitation Over an Urban Agglomeration With Future Global Warming. Earths Future 10, e2021EF002563 (2022). 11. Han, J.-Y., Baik, J.-J. & Lee, H. Urban impacts on precipitation. Asia-Pac. J. Atmospheric Sci. 50, 17–30 (2014). 12. Shepherd, J. M. Evidence of urban-induced precipitation variability in arid climate regimes. J. Arid Environ. 67, 607–628 (2006). 13. Holmer, B. & Eliasson, I. Urban–rural vapour pressure differences and their role in the development of urban heat islands. Int. J. Climatol. 19, 989–1009 (1999). 14. Kuttler, W., Weber, S., Schonnefeld, J. & Hesselschwerdt, A. Urban/rural atmospheric water vapour pressure differences and urban moisture excess in Krefeld, Germany. Int. J. Climatol. 27, 2005–2015 (2007). 15. Langendijk, G. S., Rechid, D. & Jacob, D. Urban Areas and Urban–Rural Contrasts under Climate Change: What Does the EURO-CORDEX Ensemble Tell Us?—Investigating near Surface Humidity in Berlin and Its Surroundings. Atmosphere 10, 730 (2019). 16. Zhao, L. et al. Global multi-model projections of local urban climates. Nat. Clim. Change 11, 152–157 (2021). 17. Baidar, S., Bonin, T., Choukulkar, A., Brewer, A. & Hardesty, M. Observation of the Urban Wind Island Effect. EPJ Web Conf. 237, 06009 (2020). 18. Childs, P. P. & Raman, S. Observations and Numerical Simulations of Urban Heat Island and Sea Breeze Circulations over New York City. Pure Appl. Geophys. 162, 1955–1980 (2005). 19. Droste, A. M., Steeneveld, G. J. & Holtslag, A. A. M. Introducing the urban wind island effect. Environ. Res. Lett. 13, 094007 (2018). 20. Lee, D. O. The influence of atmospheric stability and the urban heat island on urban-rural wind speed differences. Atmospheric Environ. 1967 13, 1175–1180 (1979). 21. UN. United Nations, Department of Economic and Social Affairs, Population Division (2019). World Urbanization Prospects: The 2018 Revision (ST/ESA/SER.A/420). New York: United Nations. https://population.un.org/wup/assets/WUP2018-Report.pdf (2018). 22. Chen, D. et al. Framing, context, and methods. in Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change (eds Masson-Delmotte, V. et al.) 147–286 (Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, 2021). doi:10.1017/9781009157896.001. 23. Eyring, V. et al. Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization. Geosci. Model Dev. 9, 1937–1958 (2016). 24. Langendijk, G. S. et al. Towards better understanding the urban environment and its interactions with regional climate change - The WCRP CORDEX Flagship Pilot Study URB-RCC. Urban Clim. 58, 102165 (2024). 25. Doan, V. Q. & Kusaka, H. Projections of urban climate in the 2050s in a fast-growing city in Southeast Asia: The greater Ho Chi Minh City metropolitan area, Vietnam. Int. J. Climatol. 38, 4155–4171 (2018). 26. Gu, Y., Kusaka, H. & Doan, Q.-V. An advection fog event response to future climate forcing in the 2030s–2080s: a case study for Shanghai. Front. Earth Sci. 17, 527–546 (2023). 27. Bushenkova, A., Soares, P. M. M., Johannsen, F. & Lima, D. C. A. Towards an Improved Representation of the Urban Heat Island Effect: A Multi-Scale Application of Xgboost for Madrid. SSRN Scholarly Paper at https://doi.org/10.2139/ssrn.4729233 (2024). 28. Johannsen, F., Soares, P. M. M. & Langendijk, G. S. On the deep learning approach for improving the representation of urban climate: The Paris urban heat island and temperature extremes. Urban Clim. 56, 102039 (2024). 29. Le Roy, B., Lemonsu, A. & Schoetter, R. A statistical–dynamical downscaling methodology for the urban heat island applied to the EURO-CORDEX ensemble. Clim. Dyn. 56, 2487–2508 (2021). 30. Hamdi, R. et al. The State-of-the-Art of Urban Climate Change Modeling and Observations. Earth Syst. Environ. 4, 631–646 (2020). 31. Masson, V., Lemonsu, A., Hidalgo, J. & Voogt, J. Urban Climates and Climate Change. Annu. Rev. Environ. Resour. 45, 411–444 (2020). 32. Coppola, E. et al. A first-of-its-kind multi-model convection permitting ensemble for investigating convective phenomena over Europe and the Mediterranean. Clim. Dyn. 55, 3–34 (2020). 33. Hundhausen, M., Feldmann, H., Laube, N. & Pinto, J. G. Future heat extremes and impacts in a convection-permitting climate ensemble over Germany. Nat. Hazards Earth Syst. Sci. 23, 2873–2893 (2023). 34. Grimmond, C. S. B. et al. The International Urban Energy Balance Models Comparison Project: First Results from Phase 1. J. Appl. Meteorol. Climatol. 49, 1268–1292 (2010). 35. Langendijk, G. S., Rechid, D., Sieck, K. & Jacob, D. Added value of convection-permitting simulations for understanding future urban humidity extremes: case studies for Berlin and its surroundings. Weather Clim. Extrem. 33, 100367 (2021). 36. Lipson, M. J. et al. Evaluation of 30 urban land surface models in the Urban-PLUMBER project: Phase 1 results. Q. J. R. Meteorol. Soc. 150, 126–169 (2024). 37. Taha, H. Modifying a Mesoscale Meteorological Model to Better Incorporate Urban Heat Storage: A Bulk-Parameterization Approach. J. Appl. Meteorol. Climatol. 38, 466–473 (1999). 38. Kusaka, H., Kondo, H., Kikegawa, Y. & Kimura, F. A Simple Single-Layer Urban Canopy Model For Atmospheric Models: Comparison With Multi-Layer And Slab Models. Bound.-Layer Meteorol. 101, 329–358 (2001). 39. Salamanca, F., Krpo, A., Martilli, A. & Clappier, A. A new building energy model coupled with an urban canopy parameterization for urban climate simulations—part I. formulation, verification, and sensitivity analysis of the model. Theor. Appl. Climatol. 99, 331–344 (2010). 40. Lean, H. W. et al. The hectometric modelling challenge: Gaps in the current state of the art and ways forward towards the implementation of 100-m scale weather and climate models. Q. J. R. Meteorol. Soc. 150, 4671–4708 (2024). 41. Fuhrer, O. et al. Near-global climate simulation at 1 km resolution: establishing a performance baseline on 4888 GPUs with COSMO 5.0. Geosci. Model Dev. 11, 1665–1681 (2018). 42. Schär, C. et al. Kilometer-Scale Climate Models: Prospects and Challenges. Bull. Am. Meteorol. Soc. 101, E567–E587 (2020). 43. Sharma, A., Wuebbles, D. J. & Kotamarthi, R. The Need for Urban-Resolving Climate Modeling Across Scales. AGU Adv. 2, e2020AV000271 (2021). 44. Coppola, E. et al. Climate hazard indices projections based on CORDEX-CORE, CMIP5 and CMIP6 ensemble. Clim. Dyn. 57, 1293–1383 (2021). 45. Giorgi, F. et al. The CORDEX-CORE EXP-I Initiative: Description and Highlight Results from the Initial Analysis. Bull. Am. Meteorol. Soc. 103, E293–E310 (2022). 46. Gutowski Jr., W. J. et al. WCRP COordinated Regional Downscaling EXperiment (CORDEX): a diagnostic MIP for CMIP6. Geosci. Model Dev. 9, 4087–4095 (2016). 47. Remedio, A. R. et al. Evaluation of New CORDEX Simulations Using an Updated Köppen–Trewartha Climate Classification. Atmosphere 10, 726 (2019). 48. Teichmann, C. et al. Assessing mean climate change signals in the global CORDEX-CORE ensemble. Clim. Dyn. 57, 1269–1292 (2021). 49. Coppola, E. et al. Assessment of the European Climate Projections as Simulated by the Large EURO-CORDEX Regional and Global Climate Model Ensemble. J. Geophys. Res. Atmospheres 126, e2019JD032356 (2021). 50. Jacob, D. et al. Regional climate downscaling over Europe: perspectives from the EURO-CORDEX community. Reg. Environ. Change 20, 51 (2020). 51. Jacob, D. et al. EURO-CORDEX: new high-resolution climate change projections for European impact research. Reg. Environ. Change 14, 563–578 (2014). 52. Vautard, R. et al. Evaluation of the Large EURO-CORDEX Regional Climate Model Ensemble. J. Geophys. Res. Atmospheres 126, e2019JD032344 (2021). 53. Masson, V. et al. City-descriptive input data for urban climate models: Model requirements, data sources and challenges. Urban Clim. 31, 100536 (2020). 54. Diez-Sierra, J. et al. The Worldwide C3S CORDEX Grand Ensemble: A Major Contribution to Assess Regional Climate Change in the IPCC AR6 Atlas. Bull. Am. Meteorol. Soc. 103, E2804–E2826 (2022). 55. Langendijk, G. S. et al. Representation of global mega-cities and their urban heat island in CORDEX-CORE regional climate model simulations. Preprint at https://zenodo.org/records/15691322 (2025). 56. Katragkou, E. et al. Delivering an Improved Framework for the New Generation of CMIP6-Driven EURO-CORDEX Regional Climate Simulations. Bull. Am. Meteorol. Soc. 105, E962–E974 (2024). 57. Ban, N. et al. The first multi-model ensemble of regional climate simulations at kilometer-scale resolution, part I: evaluation of precipitation. Clim. Dyn. 57, 275–302 (2021). 58. Pichelli, E. et al. The first multi-model ensemble of regional climate simulations at kilometer-scale resolution part 2: historical and future simulations of precipitation. Clim. Dyn. 56, 3581–3602 (2021). 59. Soares, P. M. M. et al. The added value of km-scale simulations to describe temperature over complex orography: the CORDEX FPS-Convection multi-model ensemble runs over the Alps. Clim. Dyn. 62, 4491–4514 (2024). 60. Giorgi, F. & Gutowski, W. J. Regional dynamical downscaling and the CORDEX initiative. Annu. Rev. Environ. Resour. 40, 467–490 (2015). 61. Dee, D. P. et al. The ERA-Interim reanalysis: configuration and performance of the data assimilation system. Q. J. R. Meteorol. Soc. 137, 553–597 (2011). 62. Juckes, M. et al. The CORDEX archive in ESGF: a global archive for regional data. in EGU General Assembly Conference Abstracts vol. 15 11043 (2013). 63. Iturbide, M. et al. Implementation of FAIR principles in the IPCC: the WGI AR6 Atlas repository. Sci. Data 9, 629 (2022). 64. Langendijk, G. S. et al. CORDEX-CORE urban and impervious surface area dataset. Zenodo https://doi.org/10.5281/zenodo.15700267 (2025). 65. Christensen, O. B., Gutowski, W. J., Nikulin, G. & Legutke, S. CORDEX Archive Design. 2020. (2020). 66. Chakraborty, T. et al. Large disagreements in estimates of urban land across scales and their implications. Nat. Commun. 15, 9165 (2024). 67. Voogt, J. A. & Oke, T. R. Thermal remote sensing of urban climates. Remote Sens. Environ. 86, 370–384 (2003). 68. Hoffmann, P. et al. High-resolution land use and land cover dataset for regional climate modelling: historical and future changes in Europe. Earth Syst. Sci. Data 15, 3819–3852 (2023). 69. Rozenfeld, H. D. et al. Laws of population growth. Proc. Natl. Acad. Sci. U. S. A. 105, 18702–18707 (2008). 70. Chakraborty, T. & Lee, X. A simplified urban-extent algorithm to characterize surface urban heat islands on a global scale and examine vegetation control on their spatiotemporal variability. Int. J. Appl. Earth Obs. Geoinformation 74, 269–280 (2019). 71. Peng, S. et al. Surface Urban Heat Island Across 419 Global Big Cities. Environ. Sci. Technol. 46, 696–703 (2012). 72. Venter, Z. S., Chakraborty, T. & Lee, X. Crowdsourced air temperatures contrast satellite measures of the urban heat island and its mechanisms. Sci. Adv. 7, eabb9569 (2021). 73. Zhou, B., Rybski, D. & Kropp, J. P. On the statistics of urban heat island intensity. Geophys. Res. Lett. 40, 5486–5491 (2013). 74. Guerreiro, S. B., Kilsby, C. & Fowler, H. J. Assessing the threat of future megadrought in Iberia. Int. J. Climatol. 37, 5024–5034 (2017). 75. Schwingshackl, C., Daloz, A. S., Iles, C., Aunan, K. & Sillmann, J. High-resolution projections of ambient heat for major European cities using different heat metrics. Nat. Hazards Earth Syst. Sci. 24, 331–354 (2024). 76. Huszar, P. et al. Regional climate model assessment of the urban land-surface forcing over central Europe. Atmospheric Chem. Phys. 14, 12393–12413 (2014). 77. Karlický, J. et al. The “urban meteorology island”: a multi-model ensemble analysis. Atmospheric Chem. Phys. 20, 15061–15077 (2020). 78. Lo, J. C. F., Lau, A. K. H., Chen, F., Fung, J. C. H. & Leung, K. K. M. Urban Modification in a Mesoscale Model and the Effects on the Local Circulation in the Pearl River Delta Region. J. Appl. Meteorol. Climatol. 46, 457–476 (2007). 79. Zhang, P., Imhoff, M., Wolfe, R. & Bounoua, L. Potential Drivers of Urban Heat Island in Northeast USA Cities. AGU Fall Meet. Abstr. (2010). 80. Walt, S. van der et al. scikit-image: image processing in Python. PeerJ 2, e453 (2014). 81. Daniel, M. et al. Benefits of explicit urban parameterization in regional climate modeling to study climate and city interactions. Clim. Dyn. 52, 2745–2764 (2019). 82. Diez-Sierra, J. et al. Dataset of urban and surrounding reference rural regions for CORDEX-CORE. Zenodo https://doi.org/10.5281/zenodo.17257489 (2025). 83. Florczyk A.J. et al. GHSL Data Package 2019. https://human-settlement.emergency.copernicus.eu/documents/GHSL_Data_Package_2019.pdf (2019). 84. Melchiorri, M. et al. Stats in the City: The GHSL Urban Centre Database 2025 : Public Release GHS UCDB R2024. (Publications Office of the European Union, 2024). 85. Zhou, B. et al. Assessing Seasonality in the Surface Urban Heat Island of London. J. Appl. Meteorol. Climatol. 55, 493–505 (2016). 86. Siswanto, S. et al. Spatio-temporal characteristics of urban heat Island of Jakarta metropolitan. Remote Sens. Appl. Soc. Environ. 32, 101062 (2023). 87. Deilami, K., Kamruzzaman, Md. & Liu, Y. Urban heat island effect: A systematic review of spatio-temporal factors, data, methods, and mitigation measures. Int. J. Appl. Earth Obs. Geoinformation 67, 30–42 (2018). 88. Diez-Sierra, J. et al. URCLIMASK: A Python Package for Delineating Urban Areas and Their Surrounding Reference Rural Regions from Regional Climate Models (RCMs). Zenodo https://doi.org/10.5281/zenodo.17257445 (2025). 89. Maraun, D. & Widmann, M. Statistical Downscaling and Bias Correction for Climate Research. (Cambridge University Press, Cambridge, 2018). doi:10.1017/9781107588783. 90. Cornes, R. C., van der Schrier, G., van den Besselaar, E. J. M. & Jones, P. D. An Ensemble Version of the E-OBS Temperature and Precipitation Data Sets. J. Geophys. Res. Atmospheres 123, 9391–9409 (2018).