Granular mapping of UHI and heatwave effects: Implications for building performance and urban resilience

Granular mapping of UHI and heatwave effects: Implications for building performance and urban resilience Alireza Karimi * , David Moreno-Rangel , Antonio García-Martínez Instituto Universitario de Arquitectura y Ciencias de la Construcci´ on, Escuela T´ ecnica Superior de Arquitectura, Universidad de Sevilla, 41012 Sevilla, Spain ARTICLE INFO Keywords: Urban Heat Island (UHI) Climate adaptation strategies Green infrastructure effectiveness Multi-Criteria Decision Analysis (MCDA) Interactive climate data visualization ABSTRACT This study presents a comprehensive framework for addressing the growing challenges posed by the Urban Heat Island (UHI) effect on heatwave (HW) phases in urban environments. By generating customized, high-resolution weather datasets (EPW) that capture both spatial and temporal variations of UHI across different periods, this framework is applied to the Canillas neighbourhood in Madrid, providing valuable insights into the UHI-HW interactions specific to this urban area. Through the analysis of historical, mid-term, and future climate scenarios, the findings reveal a concerning trend: While the frequency of HW events may decrease, their duration and intensity are projected to increase significantly. Future climate scenarios suggest the possibility of up to 30 heatwaves occurring annually, with a combined total duration of 102 days. Moreover, UHI within an urban cell exhibits varying intensities and sizes depending on the region and measurement time. The use of Multi-Criteria Decision Analysis (MCDA) sheds light on how factors such as building density, road infrastructure, and building height amplify UHI effects, driving temperatures in densely developed areas above 37 ◦C during HW. Depending on the urban geometry, there is a difference between the amount of heat and the time it takes to release that heat. While urban vegetation has a cooling effect, its ability to mitigate extreme temperatures during intense HW is limited, emphasizing the need for green infrastructure as a foundational, though inadequate, solution when applied alone. Additionally, analyses using the Normalized Difference Vegetation Index (NDVI) reveal a strong inverse relationship between vegetation and temperature, particularly during historical periods. However, this effect is expected to diminish in the future, underscoring the need for innovative resilience strategies such as: such as cool roofs, reflective surfaces, or green rooftops, as traditional approaches may no longer suffice in the face of intensifying climate change. Therefore, this study offers crucial insights to bolster urban resilience, emphasizing the need for the strategic integration of green infrastructure, innovative cooling solutions, and balanced urban density to mitigate the combined effects of the UHI phenomenon and HW in cities experiencing rapid warming. 1. Introduction The Sixth Assessment Report (AR6) confirms that global surface temperatures are projected to rise throughout this century, with warming likely to exceed 1.5 ◦C or even 2 ◦C unless substantial reductions in greenhouse gas emissions are achieved [1]. Urban areas, home to over half the world’s population and responsible for more than 70 % of CO2 emissions [2], are at the forefront of this challenge. Among the most pressing urban climate phenomena is the Urban Heat Island (UHI) effect, where dense cities experience higher temperatures than their rural surroundings due to reduced vegetation, heat-retaining materials, and anthropogenic activities [3]. The convergence of UHI with heatwaves (HW) has intensified the risks cities face globally, threatening public health, infrastructure, and ecosystems [4]. HW, already a deadly consequence of climate change, are amplified by UHI, creating hotspots of extreme thermal stress. For example, the 2003 European HW caused 45,000 fatalities, with Mediterranean regions particularly vulnerable [5]. As climate change intensifies, the increasing frequency and severity of such events strain the energy systems of buildings {REF: Low-energy buildings in combination with grid decarbonization, life cycle assessment of passive house buildings in Northern Ireland} [6] and escalate cooling demands [7], while also exacerbating UHI effects through waste heat emissions and thermal storage in construction materials [8], ultimately reducing energy efficiency under extreme weather conditions [9]. Accurate, high-resolution weather datasets, generated via * Corresponding author. E-mail address: [email protected] (A. Karimi). Contents lists available at ScienceDirect Building and Environment journal homepage: www.elsevier.com/locate/buildenv https://doi.org/10.1016/j.buildenv.2025.112705 Received 25 November 2024; Received in revised form 21 January 2025; Accepted 10 February 2025 Building and Environment 273 (2025) 112705 Available online 13 February 2025 0360-1323/© 2025 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ ). dynamical or statistical downscaling, are critical for evaluating building performance and defining boundary conditions for energy and resilience simulations. However, uncertainties in forecasting extreme climate events hinder effective adaptation and planning [10]. Closing these gaps is essential to developing resilient buildings and cities capable of withstanding UHI and HW risks. Accurate weather data is essential for building models, with dynamical and statistical downscaling being common methods [11]. Dynamical downscaling uses a Regional Climate Model (RCM) to generate localized climate data, based on boundary conditions from either a Global Climate Model (GCM) or an observation-based dataset. By incorporating high-resolution topographical data, land-sea contrasts, and other Earth-system elements, RCMs simulate atmospheric and land surface processes, providing climate information at a finer resolution than GCMs [12]. Additionally, RCMs are typically nested within GCMs, and the quality of the output is contingent upon the accuracy of the underlying GCM. Efforts to quantify uncertainties in climate projections involve combining different GCM-RCM pairings and conducting simulations referred to as "ensembles." These ensemble simulations generate multiple climate projections by varying the input parameters, model configurations, or even the underlying GCMs. However, it demands substantial computational power and storage capacity [13,14]. On the other hand, Climate data derived from statistical downscaling methods, which estimate local climate variables from larger-scale climate data using time series adjustments (Morphing), deterministic, or stochastic methods. The process of morphing entails a mathematical transformation of current weather data to match climate change projections from models like the GCM and RCM [15]. This process ensures accuracy by aligning current records with anticipated future changes over time, aiming to maintain local climate characteristics. This approach operates on the assumption that existing weather patterns will persist into the future, distinguishing it from alternative weather generation methods that require more extensive computational resources and larger datasets to achieve comparable spatial and temporal resolutions [16]. Additionally, these alternative methods require post-processing to address bias in climate models and involve computing multi-year hourly climate data and downscaling it when the resolution is insufficient. These procedures assume that the interaction between broad-scale meteorology and geographical features influences the local weather and climate [17]. Meanwhile, stochastic weather generators help complete data gaps and create continuous synthetic weather sequences. This becomes achievable by replicating key characteristics found in actual meteorological records, such as average daily values, variations, correlations, frequencies of events, and extreme occurrences [18]. It offers the benefit of incorporating the distribution utilized for the climate change signal. It also considers potential shifts in weather patterns and climate variability. However, a potential drawback of this approach seems to be the requirement for a substantial quantity of data to train the model, as the distributions used to generate future data rely on the initial data provided to the model [19] (For a detailed comparison of the differences and characteristics of each climate modeling approach, Appendix A). In this regard, recent research efforts have aimed to address this issue. For example: Nik. [20], developed a method to synthesize weather data from RCMs, creating three sets Typical Downscaled Year (TDY), Extreme Cold Year (ECY), and Extreme Warm Year (EWY) for 30-year periods using multiple climate scenarios. The combined Triple Set ensured accurate energy simulations with reduced computational effort. Zhu et al. [21] developed the Typical Principal Component Year (TPCY) method as a more accurate and efficient alternative to the widely used Typical Meteorological Year (TMY), showing better alignment with 30-year average energy use across various climate zones in China. Furthermore, Machard et al. [22] proposed new method for making EPW files by bias-correcting RCM projections with observed data for each city, ensuring accurate representation of future climate changes in terms of extreme temperature frequency, duration, and magnitude. Similarly, Li et al. [23] proposed a Dual-Periodic Time Series Model to predict future temperatures more accurately than GCMs. Using Morphing, they generated future TMYs, showing fluctuating energy demand trends in Shanghai. The choice of a downscaling approach depends on the specific research objectives, available data, computational resources, and the desired level of detail required for the analysis. Each approach has its advantages and limitations, and researchers often opt for dynamical downscaling when seeking a more comprehensive representation of future climate conditions at a regional or local scale by consideration extreme condition. On the other hand, several researchers have explored the UHI variations during HW, highlighting the complex interactions between urban environments and extreme weather events. Jie He et al. [24] studied the combined effects of HW and UHI on thermal comfort in Shanghai. They found that less urbanized districts, like Jing’an District (JD), experienced the most UHI amplification, with a 1.3 ◦C increase during AppT-based HW. Air moisture and wind helped reduce heat impacts, but the relief varied by location and time of day. Chew et al. [25] studied the HW-UHI interaction in Singapore during the April 2016 HW. They found that, despite a 3 ◦C rise in daytime temperatures, UHI intensity remained steady at 2.5 ◦C, showing no synergy between HW-UHI. The study suggests that tropical cities may experience different HW-UHI dynamics compared to temperate regions and calls for more research in these areas, as tropical cities are expected to face greater heat stress. Founda et al. [26] investigated the interaction between HW and UHIs in Athens Nomenclature UHI Urban Heat Island HW Heatwaves MCDA Multi-Criteria Decision Analysis NDVI Normalized Difference Vegetation Index AR6 Sixth Assessment Report CO2 Carbon dioxide RCM Regional Climate Model GCM Global Climate Model TDY Typical Downscaled Year ECY Extreme Cold Year EWY Extreme Warm Year TPCY Typical Principal Component Year TMY Typical Meteorological Year EPW EnergyPlus Weather JD Jing’an District (specific urban district in Shanghai) UWG Urban Weather Generator GWR Geographically Weighted Regression LCZ Local Climate Zones CORDEX Coordinated Regional Downscaling Experiment CMIP5 Coupled Model Intercomparison Project 5th Phase RCP Representative Concentration Pathway QDM Quantile Delta Mapping MBCn Multivariate Bias Correction using N-dimensional Probability Density Function Transform NETCDF4 Network Common Data Form version 4 BDM Baidu Map OSM Open Street Map KDE Kernel Density Estimation MSI Multi-Spectral Instrument m meters Q-Q Quantile-Quantile A. Karimi et al. Building and Environment 273 (2025) 112705 2 during the summer of 2012. They found that during HW, UHI intensity increased by up to 3.5 ◦C compared to background summer conditions, highlighting a positive feedback loop between UHI and HW. Moreover, Li et al. [27] and Gao et al. [28] examined the impact of UHI effect in relation to urban morphology, utilizing a geographically weighted regression model for their calculations. Lastly, Dumas et al. [29] developed a methodology using a mesoscale weather model to design and evaluate neighbourhood layouts, demonstrating in a Chicago test case that new developments affect both their own and existing neighbourhood microclimates, influencing energy consumption. Previous studies on UHI-HW interaction mainly use remote sensing data to estimate surface temperatures and map thermal anomalies. While useful, these methods often miss the temporal dynamics and meteorological context of HW events and are limited to surface conditions, lacking the detail required for air temperature analysis during HW phases. To address this, we developed a methodology to create customized weather datasets that integrate UHI effects into air temperature data. Our approach captures both spatial variability and temporal evolution of UHI during HW phases. It also incorporates future climate projections, enabling a forward-looking analysis of UHI-HW interactions. Fig. 1 provides an illustration of the conceptual framework of this study (a more comprehensively detailed is provided in the Method section). The methodology consists of several key steps: Generating weather Fig. 1. Conceptual framework highlighting the primary steps of the study process. A. Karimi et al. Building and Environment 273 (2025) 112705 3 datasets tailored to HW detection is based on percentile-based temperature thresholds (Spic, Sdeb, Sint) and quantifying severity using a dimensionless metric derived from cumulative temperature exceedances. Adding UHI effects to the datasets is achieved through the Urban Weather Generator (UWG), which incorporates urban parameters such as vegetation cover, building height, building density, and albedo. The tool integrates these factors by modeling the energy exchanges and radiative impacts within urban environments, accounting for variations in surface materials, vegetation shading, and heat retention in buildings, to simulate the urban thermal environment more accurately. Air temperature data generated for HW phases, including the periods one week prior and after each event, is extracted to analyse temporal transitions and UHI influences across different HW phases. Accuracy is ensured by comparing model-simulated climatic variables with observed field data. The validation method supports the evaluation of urban resilience to HW. Lastly, the multi-criteria decision-making (MCDM) method is applied to assess the impact of urban settings on air temperature variations during HW phases. By systematically weighting and ranking relevant spatial criteria, the method highlights their relative significance in shaping temperature dynamics. This approach facilitates a detailed understanding of spatial heterogeneity, complementing the analysis of temporal transitions and UHI influences described earlier. 2. Methodology This study builds upon the highly influential methodology developed by [22], which created HW-specific weather files by utilizing macroclimate data from weather stations located outside urban areas. While their approach effectively corrects biases in climate data, due to limitations such as the lack of urban observations and the spatial heterogeneity of UHI across different Local Climate Zones (LCZs), they do not fully account for the UHI effect. In our research, we extend and enhance this foundational methodology by explicitly integrating UHI effects into HW phase analyses. This includes the creation of high-resolution urban weather files that incorporate UHI across various LCZs (1013 for each period), addressing the spatial variability of UHI within urban environments. Additionally, our study provides a dynamic analysis of UHI’s impact prior, during, and after HW events, offering a temporal perspective on its influence on urban resilience. To further improve upon previous methods, which rely on post-processed UHI corrections, we directly integrate UHI effects using the ENTROPY method, ensuring a more objective and bias-free quantification of resilience needs in future climate scenarios. By incorporating these key advancements, our study fills critical gaps in understanding the interactions between UHI and HW, offering more refined insights that can inform urban climate resilience planning. 2.1. Data extraction and conversion The methodology involves using historical and future climate simulations to prepare weather files, based on data from the Coordinated Regional Downscaling Experiment (CORDEX). The EURO – CORDEX initiative is an integral component of the global CORDEX, which can be accessed through the work of [30]. This initiative strives to establish and coordinate an international framework aimed at refining regional climate scenarios for land regions worldwide. This data source delivers regional climate projections for Europe with resolutions of 50 km and 12.5 km [31]. These projections involve the downscaling of global climate projections from CMIP5 and the latest RCP scenarios [32]. For comprehensive insights into grid configuration and divergences in parametrization schemes among participating models, detailed information can be found in [33–35]. This ensemble can be categorized into two groups: the first, with a historical set spanning the period from 1971 to 2005, comprises ten model simulations. The second type encompasses 21st-century projections driven by two Representative Concentration Pathway (RCP) scenarios, namely RCP4.5 and RCP8.5. represent the highest emission scenario, with its CO2 concentration projected to reach 936 ppm by 2100, while RCP4.5 is projected to reach 538 ppm by the same year. For both ensembles driven by RCP4.5 and RCP8.5 scenarios, eight simulations have been employed, providing daily data from 2006 to 2100 [36]. Arranging data in weather file format has been conducted for historical (2006 to 2020), mid-future (2040 to 2060), and future (2080 to 2100) periods. Given that RCP8.5 is recognized as the most extreme greenhouse gas emission scenario in the Coupled Model Intercomparison Project 5th Phase (CMIP5) and aligns with current global emission trends, it was selected to assess the worst-case scenario for resilience and adaptation planning. The selection of climate simulations from the CORDEX database was guided by McSweeney et al. [37] who identified three reliable GCMs with different climate sensitivities: NCC – NORESM, MPI-ESM-LR, and HadGEM-ES. After reviewing the CORDEX simulations for sub-daily temporal frequency, the MPI-ESM-LR (GCM) and REMO (RCM) combination was chosen for its medium climate sensitivity and closest alignment with median temperature projections. This specific simulation, referred to as "MPI-REMO," provided the necessary temporal frequency for the study. The selected GCM, MPI-ESM-LR, was downscaled using the RCM, REMO, developed by the Max Planck Institute for Meteorology and maintained by GERICS in Hamburg [38]. REMO is based on the former Europa Model of the German Weather Service and simulates various atmospheric parameters, such as wind, temperature, humidity, pressure, and cloud water with a spatial resolution of 12.5 km for Europe. The weather data, stored in NETCDF4 format, were processed using a Python script to extract and compile key variables for the selected city, including temperature, humidity, pressure, shortwave irradiance, wind speed, and cloud cover (for Europe) [22]. Additionally, after correcting the global solar irradiance, the Zhang method was applied to decompose it into direct and diffuse solar irradiance components, ensuring accurate input for subsequent building performance simulations. 2.2. Adjustment of climate model simulation biases Climate models often exhibit biases due to their coarse spatial resolutions, which can lead to misrepresentations of climate-related hazards if not properly addressed. To correct these biases, in this study, two primary methods for bias correction were employed according to the recent research [22]: Quantile Delta Mapping (QDM) and Multivariate Bias Correction using N-dimensional Probability Density Function Transform (MBCn). The QDM method focuses on correcting the systematic biases in climate variables while retaining the projected changes in their quantiles. This is achieved by de-trending model data and mapping it to observed values through quantile mapping, followed by adjustments based on future projections [39]. The MBCn method enhances the QDM approach by extending its application to multiple variables. Initially, each climate variable is corrected using the QDM technique [40]. Following this, an iterative reshuffling method is applied to refine the dependence relationships among the variables, ensuring that the corrected data maintains the necessary correlations. For our analysis, while all climate variables underwent correction using the MBCn method, the QDM method was specifically applied to global solar irradiance. This choice was made to preserve the diurnal patterns of solar irradiance, which could be distorted if corrected using the MBCn approach. Calibration of these bias correction methods was carried out on a monthly basis to maintain the natural variability in the corrected data. The methods are predicated on the assumption that biases present in historical data will persist into the future. Since the climate model simulations are sourced from CORDEX, the bias correction methods—QDM and MBCn—were applied to these data to account for biases inherent in the model’s output. The CORDEX dataset provides reliable regional climate projections, but adjustments were necessary to improve the accuracy of the model’s output, especially for critical variables like solar A. Karimi et al. Building and Environment 273 (2025) 112705 4 irradiance. 2.3. Calculating direct and diffuse solar irradiance To calculate direct and diffuse solar irradiance, Zhang’s method [41] Eq. (1) was employed, beginning with the estimation of global solar radiation on a horizontal surface (I h ). This estimation is achieved through an empirical model that integrates various meteorological variables, as defined by the following equation: Ih= I0sinh(C0+C1CC 10 +C2(CC 10)2 +C3(un−un−3) + C4f) k(1) Where, I h =Estimated global solar radiation (W/m²), I 0 =Solar constant (approximately 1361 W/m²), h =Sun’s altitude angle (degrees), C = Amount of cloud cover (tenths), u n and u n−3 =Dry-bulb temperature at hours n and n −3 ( ◦C), f =Relative humidity (%) and, C0, C1, C2, C3, C4, and k =Regression coefficients specific to our location. On the other hand, a method for dividing total global solar radiation into direct and diffuse components is employed. A common approach involves using a parameter such as direct beam transmittance K n [42], which is often be modelled using empirical relationships or functions, such as the Gompertz function. Direct Irradiance (I d ): This is typically calculated using a formula that incorporates total irradiance and the angle of incidence, expressed as I d =I h ⋅K d Diffuse Irradiance (I diff ): This represents the remaining portion of solar radiation after accounting for direct irradiance, calculated as I diff = I h −I dI Moreover, the methodology for creating TMYs follows the approach established by [22], which is based on the international standard EN ISO 15,927–4 for hygrothermal performance of buildings. This method enables the assessment of long-term energy loads while acknowledging that it may not effectively address extreme meteorological events. TMYs are constructed from twelve typical months selected from multiyear hourly datasets, using primary parameters such as dry-bulb air temperature, global solar irradiance, and relative humidity, along with secondary parameters like wind speed. The procedure involves calculating daily mean values for each primary parameter, sorting them, and determining their cumulative distribution functions. The Finkelstein–Schafer statistic (Fs) [43] is then employed to evaluate the goodness-of-fit for the primary climatic parameters across each month and year. This process was implemented using 20 years of bias-corrected RCM data to generate TMYs, which were subsequently converted into EPW using the Energy Plus weather converter tool. More detailed information can be found in [22]. 2.4. Implementation procedure on effect of HW in EPW file HW pose a serious threat to infrastructure and public safety, placing significant strain on power systems. Rising electricity demand from increased air conditioning use during these events can reduce the efficiency of thermal power plants or force shutdowns due to overheating or water shortages [44]. Transmission and distribution systems also face higher stress, risking failures in power lines and transformers, particularly in regions with aging infrastructure. Outages can have cascading effects across essential services such as water, communications, transport, and healthcare. With climate change driving more frequent and severe HW, strengthening grid resilience is crucial to minimizing disruptions [45]. Early detection plays a critical role in implementing preventive measures, optimizing energy use, and ensuring grid stability during extreme weather events. HW lack a universally accepted definition, but key aspects include duration, intensity, and severity. Duration refers to the number of days exceeding a threshold, intensity measures peak values during a HW, and severity reflects the cumulative heat load over time [46,47]. Among the various approaches, Ouzeau’s method [36,48] is particularly notable for integrating building resilience analysis. Initially developed for France using mortality data to define temperature thresholds, it was later adapted for predicting future HW in Paris and validated against historical data using CORDEX simulations. This methodology provides valuable insights for resilience studies, offering a structured approach to analyse HW impacts (Fig. 2). Additional methodological details and threshold calculations are elaborated in the Appendix B. 2.5. Modelling future urban weather emphasising local UHI The UHIs effect significantly influences EPW files by altering local temperature patterns due to urbanization [49]. However, the urban thermal layer is not uniform and varies spatially and temporally. The variability in the urban thermal layer, which can be measured using near-surface air cover, depends on land use/cover types and human activities. Consequently, the diversity and intensity of human activities in different parts of the urban environment create varying microclimates. This effect enhances heat retention, leading to higher ambient temperatures compared to surrounding rural areas. Consequently, this impact building energy simulations and HVAC system designs in urban environments [50]. Notably, the Energy Plus simulation engine does not inherently account for local wind speed and air temperature influenced by urban morphology, relying instead on meteorological data from EPW files [51]. To address this limitation, microclimate weather data is generated using the Dragonfly plugin for Grasshopper, employing the UWG method. The UWG estimates UHI effects by applying energy conservation principles to model urban canopy and boundary layer impacts on microclimate conditions [52]. Surface roughness effects on airflow are resolved through a vertical diffusion sub-model and a logarithmic wind speed profile [53]. Anthropogenic heat gains are factored in based on building type and area, while reduced evaporation due to decreased vegetation coverage is also considered using vegetation surface coverage ratios [54]. UWG divides the urban model into modules to simulate interactions between neighbourhoods, termed urban mapping, where each module is designated as upwind or downwind based on wind direction, calculating average boundary layer air temperatures for subsequent neighbourhoods [55]. In this context, this method requires a set of indicators such as: 2.5.1. Building footprint and height determination Two open data sources, Baidu Map (BDM) [56], and Open Street Map (OSM) [57] were employed to ascertain building footprints through a structured process involving three steps: 1) extracting footprints from BDM, 2) extracting footprints from OSM, and 3) integrating the building footprints derived from both BDM and OSM. In this process, the use of OSM is to supplement missing components from the BDM, which are mainly big shopping centers as they are shown in BDM as indoor scenes which are not downloadable for geometry determination [56]. However, buildings’ heights and altitudes are not available for the target coordination. Thus, building footprints are projected to the terrain’s mesh to locate them in the actual altitudes. Then, buildings’ heights are defined from Google Street View based on the number of floors [58]. 2.5.2. Building density determination KDE calculates the density of point features around each output raster cell. Conceptually, a smooth curved surface is fitted to each point, creating a continuous surface of density values. For linear features such as streets and alleys, KDE estimates the density within the neighborhood of each output raster cell by considering the proximity and distribution of these linear features. When dealing with polygon features like buildings, the process involves finding the center point (centroid) of each polygon first and then calculating the density based on these centroid points [59]. This approach allows for a detailed understanding of spatial patterns and concentrations in urban environments, aiding in A. Karimi et al. Building and Environment 273 (2025) 112705 5 effective urban planning, resource allocation, and spatial analysis. By visualizing these density estimates, planners can identify areas of high and low density, facilitating better decision-making for urban development and infrastructure improvements. In this regard, the predicted density at a new (x, y) location in KDE is determined by the kernel density estimator formula (Eq. (2)). Density =1 (radius)2∑ n i=1[3 π ×popi(1−(disti radius)2)2](2) Where i=1,……,nare the input points. Only include points in the sum if they are within the radius distance of the (x, y) location, popi is the population field value of point i, which is an optional parameter, and disti is the distance between point I and the (x, y) location. 2.5.3. Albedo assessment Our approach to evaluating albedo characteristics utilizes Sentinel-2 satellite data. We begin by acquiring Multi-Spectral Instrument (MSI) Level-2A Surface Reflectance data from Sentinel-2A and Sentinel-2B via reliable platforms like Google Earth Engine [60]. This method ensures thorough temporal coverage while minimizing the effects of cloud cover. We then enhance the quality of this dataset through preprocessing steps, including potential atmospheric correction. To estimate albedo, we extract key spectral bands from the Sentinel-2 data: Band 2 (Blue, 490 nm), Band 3 (Green, 560 nm), Band 4 (Red, 665 nm), and Band 8 (Near Infrared, 842 nm). These bands form the foundation for our analysis. We calculate shortwave broadband albedo using established narrow-to-broadband conversion coefficients. The formula used incorporates surface reflectance (pb), conversion coefficient (wb), and a constant (C), as outlined by Bonafoni [61] (Eq. (3)). a=∑ n B=1 (pb×wb) + C(3) We analyse spatial heterogeneity in albedo patterns by comparing Sentinel-2-derived data at various resolutions. This involves evaluating spatial distribution characteristics and discrepancies across three urban landscapes: business districts, high-density residential areas, and urbanrural mixed zones, along predefined sampling transects. Additionally, we investigate seasonal variations in albedo by comparing values across different seasons within these urban landscapes. 2.5.4. Urban greenery assessment Assessment of vegetation characteristics using Sentinel-2 satellite data involves MSI Level-2A Surface Reflectance data from Sentinel-2A and Sentinel-2B acquired with a focus on ensuring minimal cloud coverage (0–10 %) during 2015 and 2017 [62]. The data used has a spatial resolution of 10 m for bands 2, 3, and 4, and 20 m for other bands. Data retrieval and initial processing were executed using the Sen2r package in Python, with Google Earth Engine (GEE) and ArcGIS leveraged for enhanced accuracy and preprocessing capabilities in remote sensing and spatial analysis. For visualization, true-colour (RGB) composites were generated using bands 4 (Red), 3 (Green), and 2 (Blue), with all sources clearly cited from the European Space Agency (ESA). Additionally, false-color composites created by substituting the Red band with the Near Infrared (NIR) band (Band 8), facilitating clearer identification of vegetation features critical for calculating the Normalized Difference Vegetation Index (NDVI). Vegetation areas are categorized into three classes—non-vegetated areas, grass, and trees—using winter and summer indices within the ArcGIS environment [63]. This classification is achieved by applying the reclassify command based on NDVI values, which are calculated as (Eq. (4)). NDVI =NIR[Band8] − RED[Band4] NIR[Band8] + RED[Band4](4) Where NIR (Band 8), and RED (Band 4) represent the near-infrared and the red band of Sentinel-2A image product, respectively. Fig. 3 illustrates the extraction methods for these urban layers. Fig. 2. The conceptual framework of HW definition and analysis. A. Karimi et al. Building and Environment 273 (2025) 112705 6 2.5.5. Integrating entropy, as multi-criteria decision-making (MCDM) methods in UHI The ENTROPY method is a robust MCDM technique widely used to quantify spatial heterogeneity and assess urban complexity. It leverages Shannon’s entropy formula (Eq. (5)) to evaluate the distribution and significance of various criteria. The formula is expressed as: Entropy =∑ I i=1 p(xi) × log(1 p(xi))(5) Where the Iis the total pixel and i is the its pixel. p(xi)is the probability of the its outcome, and log(1 p(xi))is the measurement of information from xi. 2.5.5.1. Application in UHI analysis. In this study, the ENTROPY method is employed to analyse UHI dynamics during HW events. The method helps identify the most influential criteria affecting UHI by examining their fragmentation across three phases: One week prior to the HW, During the HW and, One week after the HW. 2.5.5.2. Decision matrix and indicator weighting. To apply the ENTROPY method, we construct a decision matrix X, which aggregates expert assessments of a defined set of indicators: Fig. 3. Urban layer extraction process: building and green infrastructure characteristics. A. Karimi et al. Building and Environment 273 (2025) 112705 7 X=(xij)m×n=⎡ ⎣ x11 x12 x13 …x1m x21 x22 x23 …x2m xn1xn2xn3…xnm ⎤ ⎦(6) In this matrix, x nm represents the n-th expert’s evaluation of the m-th indicator. •Normalization: Matrix values are normalized by summing each column and dividing each element by the respective column sum. •Entropy Calculation: The entropy value Ej for each indicator is computed as: Ej= − 1 In m ∑ m i=1 rij Inrij (7) Weight Assignment: Indicator weights Wj are derived as: Wj=1−Ej ∑n j=1(1−Ej)(8) 2.5.5.3. Clustering of indicators. To simplify decision-making and identify priority intervention areas, indicators are clustered based on their weights. The K-means algorithm is applied as follows: •Initialization: Random selection of K centroids. •Assignment: Each indicator is assigned to the nearest cluster using Euclidean distance: min ∑ F J=1∑ ˙ j=1k‖Wi−Mj‖2 (9) Where: Wi is the weight of the i th indicator, μ j is the centroid of cluster j, Cj is the set of indicators belonging to cluster j. •Iteration: Steps are repeated until the centroids stabilize. 2.5.5.4. Rationale for using entropy. Compared to other MCDM methods, such as AHP and ANP, the ENTROPY method provides an objective framework for classifying and ranking criteria based on their inherent variability. This is particularly advantageous in analysing UHI dynamics, as it captures the spatial and temporal fragmentation of UHI distribution [64]. By identifying the most influential factors during HW events, this method enhances our understanding of urban thermal environments and informs policy interventions aimed at mitigating UHI impacts [65]. 2.6. Study area The Canillas neighborhood was selected for testing the framework due to its distinctive features and significance. Located in the northeastern part of Madrid, Spain, Canillas covers an area of 2.52 km² and has a population of 39,708. It was incorporated into Madrid’s urban area in 1949 and has experienced substantial growth and transformation since then. The neighborhood was chosen not only for its vibrant community amenities—such as sports facilities, a cycling track, and green spaces—but also because our research group has worked in this area in previous projects [66]. Canillas presents a unique case, with its diverse topography and a K¨ oppen–Geiger climate classification of Bsk, which indicates a hot summer Mediterranean climate characterized by very hot, dry summers and mild, wet winters. These factors contribute to its susceptibility to UHI effects. This study’s framework is thus tested in an ideal context for examining the spatial and temporal dynamics of environmental variables. The findings are expected to offer valuable insights for similar conditions in southern European and Mediterranean cities with a Bsk climate classification. 2.7. Validation The methodological framework of this research is based on the study presented in [22], which uses a comparative analysis of observed and model-simulated climatic variables to assess the resilience of buildings to climate change and HW. For the city of Madrid, field data were collected during the periods of 10–21 July and 3–11 June 2021. During these periods, key climatic variables including ambient air temperature, relative humidity, solar radiation, and wind speed were measured under extreme summer conditions. These data provided valuable information on climatic features relevant to the analysis of heat waves. The observations were purposively selected based on their relevance to previous research [67–69], which investigated the impact of outdoor thermal comfort conditions on tourism activities in urban squares in Madrid. This selection ensures coherence and continuity in the study of the effects of extreme climatic conditions on urban environments. 2.7.1. Comparison of observational and simulated data To assess the accuracy of the bias correction methods applied to the RCM outputs, observational data were compared with raw and corrected simulations of the model. The main evaluation criteria were mean temperature, solar radiation, wind speed, and relative humidity. The mean values and standard deviations of these climate variables during the validation periods are presented in Appendix C. The findings show that the corrected RCM-bc simulations are more consistent with the observed data than the RCM-raw simulations, especially for variables such as temperature and relative humidity. For instance, during the period 10–21 July, the observed average temperature was 29.7 ◦C with a standard deviation of 2.3 ◦C. The corrected model produced an average temperature of 28.9 ◦C, closer to the observed value than the raw model’s 29.2 ◦C. Similarly, the corrected model improved the accuracy of solar irradiance estimates, reporting an average of 251.4 W/m² compared to the observed 250.3 W/m², while the raw model reported 253.1 W/m². These results demonstrate that the bias correction methods enhance the model’s ability to simulate key climate variables, which is critical for accurately modeling extreme climate events such as HW. However, higher standard deviations for some variables, such as relative humidity (11.8 % in July), indicate unresolved complexities in modeling dynamic climate conditions. To validate the findings and extend the time range of analysis, ERA5 reanalysis data were also incorporated. This approach bridges short-term observations with long-term trends, affirming the effectiveness of the bias correction methods and enhancing the reliability of climate projections for assessing urban resilience to extreme climate events. 3. Results and discussion 3.1. Heat waves detected by the studied model HW were identified using Ouzeau’s method, as detailed in Fig. 4 and Appendix D, across three intervals: historical (2006–2020), mid-future (2040–2060), and future (2080–2100). During the historical period, six HW were detected, lasting 4 to 14 days. In August 2014, a notable event spanned 14 days with an intensity of 31 ◦C. Projections indicate an increase in frequency and duration: the mid-future period is expected to experience 26 HW (3 to 51 days), while the future period projects 30 HW (4 to 102 days). HW severity, defined by Ouzeau’s method, ranged from 2.59 to 12.93 in the historical period, with the mid-future and future periods showing maximum severities of 31.32 and 171.92, respectively. Duration trends reveal a shift towards longer HW over time. Short events (3–4 days) comprised 53 % of occurrences historically but are projected to drop to 8.2 % by the century’s end. Conversely, longer events are A. Karimi et al. Building and Environment 273 (2025) 112705 8 expected to rise, impacting energy demand, indoor comfort, and public health. The most intense, severe, and longest HW occurred in 2014 during the historical period. In the mid-future, these events were recorded in 2054, 2058, and 2051, respectively. For the future period, the most intense and longest events occurred in 2097, while the most severe was observed in 2084. The analysis of HW reveals not only the increasing frequency and duration of these extreme events but also highlights the critical role of urban environments in amplifying their impacts. To understand the dynamics of these events, UHI patterns are analyzed in three key phases: before, during, and after each HW. For the historical period, the HW from August 1 to 14, 2014, is examined, while for mid-future and future scenarios, HW from July 22 to August 19, 2058, and from July 1 to September 27, 2084, are analyzed. The results from earlier steps, as shown in Appendix D, are used to define these periods. By linking HW characteristics with UHI patterns, we aim to uncover the intricate relationships between urbanization, temperature variations, and public health, ultimately providing insights into potential adaptive strategies for mitigating heat-related risks. •Quantile-Quantile (Q-Q) Analysis: This examines how UHI influences air temperature distributions during HW across historical, midfuture, and future periods, offering insights into the temporal changes in temperature anomalies related to UHI. •Clustering UHI criteria by using entropy method: In this kind of decision making make the data clustering into important and most effective criteria to low impact criteria: •Role of Vegetation AND Trees: This evaluation focuses on how vegetation impacts UHI Intensity during various HW phases, highlighting the role of green spaces in mitigating UHI effects. •Data Visualization: Temperature data are visualized using histograms with KDE plots to depict the distribution and density of temperatures, clarifying UHI effects. 3.2. Quantile-Quantile (Q-Q) analysis of UHI impacts on air temperature distributions during HW phases across historical, mid-future, and future periods To evaluate the impact of UHI on air temperature distributions during HW, we conducted a Q-Q analysis across three periods—historical, mid-future, and future—and examined three phases: preHW, during the HW, and post-HW. This analysis provides insights into how UHI influences temperature distributions at different HW stages and traces the evolving intensity of UHI effects over time. 3.2.1. Q-Q analysis of temperature extremes in historical urban HW events The Q-Q plots for the historical period detail the UHI effect’s influence on temperature distributions across three phases: prior, during, and after a HW. In the pre-HW phase, minor deviations from normality, with slight curvatures at both ends of the temperature distribution, indicate the initial influence of UHI, characterized by elevated nighttime temperatures and reduced cooling rates. This pre-existing skewness reflects heat retention due to infrastructure and anthropogenic emissions. During the HW, the Q-Q plot shows pronounced deviations, especially in the lower quantiles, with a significant concave curvature, signifying the intensified UHI effect, as maximum temperatures rise sharply and the distribution compresses at the lower end. This pattern shows that heat accumulation limits cooling ability. Post-HW, the Q-Q plot suggests a partial return to the pre-HW distribution but still displays notable lowertail deviations. This ongoing skewness indicates that UHI continues influencing urban temperatures, preventing a full return to pre-event baselines and signaling persistent shifts in urban temperature dynamics (Fig. 5). 3.2.2. Q-Q analysis of temperature extremes in MID -Future urban HW events For the mid-future period, Q-Q plots reveal key insights into UHI effects at different HW phases. Pre-HW plots display significant deviations from normality, especially at the tails, suggesting urban temperatures are skewed lower prior the HW. During the HW, the distribution appears closer to normal, with an S-shaped curve showing Fig. 4. All HW detected in three distinct periods in Madrid (BSk climate): historical (2001–2020), mid-term future (2041–2060), and long-term future (2081–2100). A. Karimi et al. Building and Environment 273 (2025) 112705 9 below the pre-HW peak, with a distribution showing reduced variability compared to the HW period. This cooling trend suggests gradual heat dissipation and possible contributions from increased vegetation or improved cooling strategies. These trends underscore the significant impact of HW on urban temperatures, stressing the need for green spaces, cooling materials, and resilient urban planning to mitigate the UHI effect and protect public health. Ongoing research and monitoring are vital for enhancing HW management and future climate adaptation (Fig. 15). 3.6. Limitations and future research This study used a 50 ×50 m² grid to analyze UHI across three phases (pre-HW, during HW, post-HW) in different time periods (Historical, Mid-future, Future). This high resolution enabled precise analysis of urban parameters and UHI dynamics but posed significant computational challenges, especially for large urban areas. Coarser grids, such as 250 ×250 m² in India [27] or 600 ×600 m² in Karlshamn [58], are more efficient but fail to capture localized phenomena like street-level shading or building-specific interactions critical for urban resilience during HW. Future research could explore the use of machine learning to optimize grid resolution, striking a balance between accuracy and computational efficiency. Additionally, integrating real-time urban climate monitoring systems could enable dynamic grid adjustments, producing weather files customized for specific building locations under extreme conditions, thereby greatly enhancing urban design and resilience. While this study focused on integration between UHI and HW, future research could apply similar methods to other extreme events, like heavy rainfall or cold spells, to refine adaptive modeling for diverse urban climates. Moroever, the use of Sentinel-2 data for vegetation mapping in this study has limitations, mainly due to its 10 to 60 m spatial resolution. While sufficient for large-scale urban analysis, it struggles to capture the fine-scale heterogeneity of vegetation in dense urban environments [70, 71]. Cloud cover and variable lighting conditions also compromise the consistency of NDVI assessments, impacting the reliability of vegetation dynamics over time [72,73]. NDVI, although widely used, lacks details on plant health, species diversity, and physiological states, which are crucial for modeling the thermal properties of vegetation and its effects on UHI. Future research could combine Sentinel-2 data with higher-resolution datasets, such as drone imagery or LIDAR, to improve vegetation mapping in complex urban areas. 4. Conclution This methodology offers a comprehensive approach to studying UHIHW interactions by generating customized weather datasets that capture both spatial and temporal variations of UHI during HW phases across different periods. The analysis of historical, mid-term, and future trends reveals concerning patterns: while the frequency of HW may decrease, their duration and intensity are projected to significantly increase, with future scenarios suggesting up to 30 HW lasting for 102 days in scale of macroclimate. The use of MCDA to assess UHI entropy provides further insights into the factors influencing HW dynamics in the scale of microclimate. In this regard, this framework was applied to the Canillas district in Madrid. Results indicate that factors such as building density, road density, and building height significantly affect HW intensity. In densely populated urban areas, temperatures can exceed 37 ◦C, Fig. 14. Mid-future of temperature distribution by HW phase. Fig. 15. Future of temperature distribution by HW phase. A. Karimi et al. Building and Environment 273 (2025) 112705 16 exacerbating the effects of HW. In contrast, areas with greater vegetation experience lower temperature increases and recover more quickly to normal conditions. However, the analysis revealed that urban greenery had a limited impact on mitigating extreme temperatures during HW events. This suggests that while vegetation offers some cooling benefits, its effectiveness in significantly reducing extreme heat during such events is constrained. Furthermore, the analysis of the NDVI shows a strong inverse relationship with temperature, indicating that increased vegetation can effectively help lower temperatures, particularly in historical contexts. However, the declining effectiveness of NDVI in future scenarios suggests that reliance on existing vegetation may not be enough, highlighting the need for exploring new urban approaches to enhance resilience against rising temperatures. CRediT authorship contribution statement Alireza Karimi: Writing – review & editing, Writing – original draft, Validation, Software, Resources, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization. David Moreno-Rangel: Writing – review & editing, Conceptualization. Antonio García-Martínez: Writing – review & editing, Conceptualization. Declaration of competing interest The authors declare no conflict of interest related to this study. The research was conducted independently, and no funding, affiliations, or personal relationships influenced the findings reported in this manuscript. Supplementary materials Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.buildenv.2025.112705. Data availability Data will be made available on request. References [1] S. Speizer, C. Raymond, C. Ivanovich, Concentrated and intensifying humid heat extremes in the IPCC AR6 regions, Geophys. Res. Lett. 49 (2022), https://doi.org/ 10.1029/2021GL097261. [2] F. Dokhanian, M. Mohajerani, H. Estaji, M. Nikravan, Shading design optimization in a semi-arid region: considering energy consumption, greenhouse gas emissions, and cost, J. Clean. Prod. (2023) 139293. [3] A. Karimi, P. Mohammad, A. García-Martínez, D. Moreno-Rangel, D. Gachkar, S. Gachkar, New developments and future challenges in reducing and controlling heat island effect in urban areas, Environ. Dev. Sustain. 25 (2023) 10485–10531, https://doi.org/10.1007/s10668-022-02530-0. [4] I. Ahmed, M. van Esch, F. van der Hoeven, Heatwave vulnerability across different spatial scales: insights from the Dutch built environment, Urban. Clim. 51 (2023) 101614. [5] T. Ahmed, P. Kumar, L. Mottet, Natural ventilation in warm climates: the challenges of thermal comfort, heatwave resilience and indoor air quality, Renewable and Sustainable Energy Reviews 138 (2021) 110669. [6] Y. Chen, W. Moufouma-Okia, V. Masson-Delmotte, P. Zhai, A. Pirani, Recent progress and emerging topics on weather and climate extremes since the fifth assessment report of the intergovernmental panel on climate change, Annu Rev. Environ. Resour. 43 (2018) 35–59. [7] A. Karimi, Y.J. Kim, N.M. Zadeh, A. García-Martínez, S. Delfani, R.D. Brown, D. Moreno-Rangel, P. Mohammad, Assessment of outdoor design conditions on the energy performance of cooling systems in future climate scenarios—a case study over three cities of Texas, Unites States, Sustainability 14 (2022) 14848. [8] H. Liu, Z. Du, T. Xue, T. Jiang, Enhancing smart building performance with waste heat recovery: supply-side management, demand reduction, and peak shaving via advanced control systems, Energy Build. 327 (2025) 115070. [9] M. Panteli, P. Mancarella, Influence of extreme weather and climate change on the resilience of power systems: impacts and possible mitigation strategies, Electric Power Syst. Res. 127 (2015) 259–270. [10] A.T.D. Perera, F. Khayatian, S. Eggimann, K. Orehounig, S. Halgamuge, Quantifying the climate and human-system-driven uncertainties in energy planning by using GANs, Appl. Energy 328 (2022) 120169. [11] V.M. Nik, Application of typical and extreme weather data sets in the hygrothermal simulation of building components for future climate–A case study for a wooden frame wall, Energy Build. 154 (2017) 30–45. [12] V. Singh, S.K. Jain, P.K. Singh, Inter-comparisons and applicability of CMIP5 GCMs, RCMs and statistically downscaled NEX-GDDP based precipitation in India, Sci. Total Environ. 697 (2019) 134163. [13] L.P. Seaby, J.C. Refsgaard, T.O. Sonnenborg, S. Stisen, J.H. Christensen, K. H. Jensen, Assessment of robustness and significance of climate change signals for an ensemble of distribution-based scaled climate projections, J. Hydrol. 486 (2013) 479–493. [14] A. Karimi, A. Bayat, N. Mohammadzadeh, M. Mohajerani, M. Yeganeh, Microclimatic analysis of outdoor thermal comfort of high-rise buildings with different configurations in Tehran: insights from field surveys and thermal comfort indices, Build. Environ. (2023) 110445. [15] L. Troup, D. Fannon, Morphing climate data to simulate building energy consumption, Proc. SimBuild 6 (2016). [16] S. Farah, D. Whaley, W. Saman, J. Boland, Integrating climate change into meteorological weather data for building energy simulation, Energy Build. 183 (2019) 749–760. [17] M.P. Tootkaboni, I. Ballarini, M. Zinzi, V. Corrado, A comparative analysis of different future weather data for building energy performance simulation, Climate 9 (2021) 37. [18] M.A. Semenov, E.M. Barrow, Use of a stochastic weather generator in the development of climate change scenarios, Clim. Change 35 (1997) 397–414. [19] P. Ailliot, M. Boutigny, E. Koutroulis, A. Malisovas, V. Monbet, Stochastic weather generator for the design and reliability evaluation of desalination systems with renewable energy sources, Renew. Energy 158 (2020) 541–553. [20] V.M. Nik, Making energy simulation easier for future climate–Synthesizing typical and extreme weather data sets out of regional climate models (RCMs), Appl. Energy 177 (2016) 204–226. [21] L. Yang, K.K.W. Wan, D.H.W. Li, J.C. Lam, A new method to develop typical weather years in different climates for building energy use studies, Energy 36 (2011) 6121–6129. [22] A. Machard, A. Salvati, M.P. Tootkaboni, A. Gaur, J. Zou, L.L. Wang, F. Baba, H. Ge, F. Bre, E. Bozonnet, Typical and extreme weather datasets for studying the resilience of buildings to climate change and heatwaves, Sci. Data 11 (2024) 531. [23] M. Zhu, Y. Pan, Z. Huang, P. Xu, An alternative method to predict future weather data for building energy demand simulation under global climate change, Energy Build. 113 (2016) 74–86. [24] B.-J. He, J. Wang, H. Liu, G. Ulpiani, Localized synergies between heat waves and urban heat islands: implications on human thermal comfort and urban heat management, Environ. Res. 193 (2021) 110584. [25] L.W. Chew, X. Liu, X.-X. Li, L.K. Norford, Interaction between heat wave and urban heat island: a case study in a tropical coastal city, Singapore, Atmos. Res. 247 (2021) 105134. [26] D. Founda, M. Santamouris, Synergies between urban heat island and heat waves in Athens (Greece), during an extremely hot summer (2012), Sci. Rep. 7 (2017) 10973. [27] Y. Li, B. Zhou, M. Glockmann, J.P. Kropp, D. Rybski, Context sensitivity of surface urban heat island at the local and regional scales, Sustain. Cities. Soc. 74 (2021) 103146. [28] Y. Gao, J. Zhao, L. Han, Exploring the spatial heterogeneity of urban heat island effect and its relationship to block morphology with the geographically weighted regression model, Sustain. Cities. Soc. 76 (2022) 103431. [29] M.R. Allen-Dumas, A.N. Rose, J.R. New, O.A. Omitaomu, J. Yuan, M.L. Branstetter, L.M. Sylvester, M.B. Seals, T.M. Carvalhaes, M.B. Adams, Impacts of the morphology of new neighborhoods on microclimate and building energy, Renew. Sustain. Energy Rev. 133 (2020) 110030. [30] F. Giorgi, C. Jones, G.R. Asrar, Addressing climate information needs at the regional level: the CORDEX framework, World Meteorological Organization (WMO), Bulletin 58 (2009) 175. [31] K.E. Taylor, R.J. Stouffer, G.A. Meehl, An overview of CMIP5 and the experiment design, Bull. Am. Meteorol. Soc. 93 (2012) 485–498. [32] R.H. Moss, J.A. Edmonds, K.A. Hibbard, M.R. Manning, S.K. Rose, D.P. Van Vuuren, T.R. Carter, S. Emori, M. Kainuma, T. Kram, The next generation of scenarios for climate change research and assessment, Nature 463 (2010) 747–756. [33] D. Jacob, J. Petersen, B. Eggert, A. Alias, O.B. Christensen, L.M. Bouwer, A. Braun, A. Colette, M. D´ equ´ e, G. Georgievski, EURO-CORDEX: new high-resolution climate change projections for European impact research, Reg. Environ. Change 14 (2014) 563–578. [34] R. Vautard, A. Gobiet, D. Jacob, M. Belda, A. Colette, M. D´ equ´ e, J. Fern´ andez, M. García-Díez, K. Goergen, I. Güttler, The simulation of European heat waves from an ensemble of regional climate models within the EURO-CORDEX project, Clim. Dyn. 41 (2013) 2555–2575. [35] S. Kotlarski, K. Keuler, O.B. Christensen, A. Colette, M. D´ equ´ e, A. Gobiet, K. Goergen, D. Jacob, D. Lüthi, E. Van Meijgaard, Regional climate modeling on European scales: a joint standard evaluation of the EURO-CORDEX RCM ensemble, Geosci. Model. Dev. 7 (2014) 1297–1333. [36] G. Ouzeau, J.-M. Soubeyroux, M. Schneider, R. Vautard, S. Planton, Heat waves analysis over France in present and future climate: application of a new method on the EURO-CORDEX ensemble, Clim. Serv. 4 (2016) 1–12. [37] C.F. McSweeney, R.G. Jones, R.W. Lee, D.P. Rowell, Selecting CMIP5 GCMs for downscaling over multiple regions, Clim. Dyn. 44 (2015) 3237–3260. A. Karimi et al. Building and Environment 273 (2025) 112705 17 [38] S. Pfeifer, Modeling cold cloud processes with the regional climate model REMO, (2006). [39] A.J. Cannon, S.R. Sobie, T.Q. Murdock, Bias correction of GCM precipitation by quantile mapping: how well do methods preserve changes in quantiles and extremes? J. Clim. 28 (2015) 6938–6959. [40] T. Su, J. Chen, A.J. Cannon, P. Xie, Q. Guo, Multi-site bias correction of climate model outputs for hydro-meteorological impact studies: an application over a watershed in China, Hydrol. Process. 34 (2020) 2575–2598. [41] Q. Zhang, Development of the typical meteorological database for Chinese locations, Energy Build. 38 (2006) 1320–1326. [42] K.N. Musselman, S.A. Margulis, N.P. Molotch, Estimation of solar direct beam transmittance of conifer canopies from airborne LiDAR, Remote Sens. Environ. 136 (2013) 402–415. [43] J.M. Finkelstein, R.E. Schafer, Improved goodness-of-fit tests, Biometrika 58 (1971) 641–645. [44] A. Sengupta, D. Al Assaad, M. Steeman, H. Breesch, Thermal resilience to overheating assessment in a Belgian educational building with passive cooling strategies during heatwaves and power outages, in: E3S Web of Conferences, EDP Sciences, 2023, p. 1018. [45] J. Nairn, R. Fawcett, Defining heatwaves: heatwave defined as a heat-impact event servicing all, Europe 220 (2011) 224. [46] T.J. Parker, G.J. Berry, M.J. Reeder, The structure and evolution of heat waves in southeastern Australia, J. Clim. 27 (2014) 5768–5785. [47] S.E. Perkins, L.V. Alexander, On the measurement of heat waves, J. Clim. 26 (2013) 4500–4517. [48] A. Machard, C. Inard, J.M. Alessandrini, C. Pel´ e, J. Rib´ eron, A methodology for assembling future weather files including heatwaves for building thermal simulations from regional climate models multi-years datasets, Energies. (Basel) (2020) 1–34. [49] L.G.R. Santos, A. Afshari, L.K. Norford, J. Mao, Evaluating approaches for districtwide energy model calibration considering the Urban Heat Island effect, Appl. Energy 215 (2018) 31–40. [50] J. Huang, R. Reitberger, F. Banihashemi, W. Lang, A novel risk-based design framework for urban heat island: a case study of Kempten, Germany, Build. Environ. 228 (2023) 109671. [51] P.G. Ellis, P.A. Torcellini, Simulating tall buildings using EnergyPlus, Natl. Renew. Energy Lab., Golden, CO (US) (2005). [52] B. Bueno, M. Roth, L. Norford, R. Li, Computationally efficient prediction of canopy level urban air temperature at the neighbourhood scale, Urban. Clim. 9 (2014) 35–53. [53] B. Bueno, G. Pigeon, L.K. Norford, K. Zibouche, C. Marchadier, Development and evaluation of a building energy model integrated in the TEB scheme, Geosci. Model. Dev. 5 (2012) 433–448. [54] B. Bueno, J. Hidalgo, G. Pigeon, L. Norford, V. Masson, Calculation of air temperatures above the urban canopy layer from measurements at a rural operational weather station, J. Appl. Meteorol. Climatol. 52 (2013) 472–483. [55] B. Bueno, M. Roth, L. Norford, R. Li, Computationally efficient prediction of canopy level urban air temperature at the neighbourhood scale, Urban. Clim. 9 (2014) 35–53. [56] C. Wang, Y. Li, X. Shi, Information mining for urban building energy models (UBEMs) from two data sources: openStreetMap and baidu map. Building Simulation 2019, IBPSA, 2019, pp. 3369–3376. [57] P. Mooney, M. Minghini, A review of OpenStreetMap data, Mapp. Citizen Sensor (2017) 37–59. [58] M. Hosseini, K. Javanroodi, V.M. Nik, High-resolution impact assessment of climate change on building energy performance considering extreme weather events and microclimate–Investigating variations in indoor thermal comfort and degree-days, Sustain. Cities. Soc. 78 (2022) 103634. [59] J.-Q. Chen, Y.-L. He, Y.-C. Cheng, P. Fournier-Viger, J.Z. Huang, A multiple kernelbased kernel density estimator for multimodal probability density functions, Eng. Appl. Artif. Intell. 132 (2024) 107979. [60] I. Caballero, R.P. Stumpf, Retrieval of nearshore bathymetry from Sentinel-2A and 2B satellites in South Florida coastal waters, Estuar. Coast. Shelf. Sci. 226 (2019) 106277. [61] S. Bonafoni, A. Sekertekin, Albedo retrieval from Sentinel-2 by new narrow-tobroadband conversion coefficients, IEEE Geosci. Remote Sens. Lett. 17 (2020) 1618–1622. [62] H.K. Zhang, D.P. Roy, L. Yan, Z. Li, H. Huang, E. Vermote, S. Skakun, J.-C. Roger, Characterization of Sentinel-2A and Landsat-8 top of atmosphere, surface, and nadir BRDF adjusted reflectance and NDVI differences, Remote Sens. Environ. 215 (2018) 482–494. [63] S.L. Ustin, E.M. Middleton, Current and near-term earth-observing environmental satellites, their missions, characteristics, instruments, and applications, Sensors 24 (2024) 3488. [64] E. Wang, N. Alp, J. Shi, C. Wang, X. Zhang, H. Chen, Multi-criteria building energy performance benchmarking through variable clustering based compromise TOPSIS with objective entropy weighting, Energy 125 (2017) 197–210. [65] Q. Song, Z. Wang, T. Wu, Risk analysis and assessment of water resource carrying capacity based on weighted gray model with improved entropy weighting method in the central plains region of China, Ecol. Indic. 160 (2024) 111907. [66] K.A. García-Pardo, D. Moreno-Rangel, S. Domínguez-Amarillo, J.R. García-Ch´ avez, Characterisation of morphological patterns for land surface temperature distribution in urban environments: an approach to identify priority areas, Climate 12 (2023) 4. [67] A. Karimi, P. Mohammad, Effect of outdoor thermal comfort condition on visit of tourists in historical urban plazas of Sevilla and Madrid, Environ. Sci. Pollut. Res. (2022) 1–21. [68] A. Benarroch, G.A. Siles, J.M. Riera, Relationship between cloud cover and rainfall rate in Madrid: first results, in: 2023 17th European Conference on Antennas and Propagation (EuCAP), IEEE, 2023, pp. 1–4. [69] D. Iskandaryan, F. Ramos, S. Trilles, Bidirectional convolutional LSTM for the prediction of nitrogen dioxide in the city of Madrid, PLoS. One 17 (2022) e0269295. [70] N. Badaro-Saliba, J. Adjizian-Gerard, R. Zaarour, G. Najjar, LCZ scheme for assessing Urban Heat Island intensity in a complex urban area (Beirut, Lebanon), Urban. Clim. 37 (2021) 100846. [71] K. Iizuka, Y. Akiyama, M. Takase, T. Fukuba, O. Yachida, Microscale temperaturehumidity index (THI) distribution estimated at the City scale: a case study in Maebashi City, Gunma Prefecture, Japan, Remote Sens. (Basel) 16 (2024) 3164. [72] S. Mhanna, L.J.S. Halloran, F. Zwahlen, A.H. Asaad, P. Brunner, Using machine learning and remote sensing to track land use/land cover changes due to armed conflict, Sci. Total Environ. 898 (2023) 165600. [73] P. Tysiąc, R. Ossowski, Ł. Janowski, D. Moskalewicz, Utilizing UAV and orthophoto data with bathymetric LiDAR in google earth engine for coastal cliff degradation assessment, Sci. Rep. 15 (2025) 704. A. Karimi et al. Building and Environment 273 (2025) 112705 18