UTHECA_USE: A Multi-Source Dataset on Human Thermal Perception and Urban Environmental Factors in Seville

Abstract

This paper introduces UTHECA_USE, a dataset of 989 observations collected in Seville, Spain (2023–2025), integrating microclimatic, personal, and urban morphological data. It comprises 55 variables, including in situ measurements of air and globe temperatures, humidity, wind speed, derived indices such as the Universal Thermal Climate Index (UTCI), demographic and physiological participant data, subjective thermal perception, and detailed urban form characteristics. The surface temperature data of urban materials are included in a subset. The dataset is openly accessible under a permissive license, and this data descriptor documents the collection methods, calibration, survey design, and data processing to ensure reproducibility and transparency. The UTHECA project aims to develop a more accurate and adaptive outdoor thermal comfort (OTC) assessment model to guide effective, inclusive urban strategies to improve human thermal perception and climate resilience. UTHECA_USE facilitates research on outdoor thermal comfort and urban microclimates, supporting diverse analyses linking human perception, environmental conditions, and urban morphology.

1. Introduction

1.1. Background and Motivation

Rising urban temperatures and the growing frequency of heatwaves have positioned outdoor thermal comfort as a crucial area of research [1,2]. To assess thermal environments, indices like the Universal Thermal Climate Index (UTCI) are often employed, integrating microclimatic variables such as the air temperature, humidity, wind speed, and radiant temperature [3,4]. While these indices offer an objective assessment, they fail to incorporate additional influential elements shaping individual thermal perception, such as personal attributes (e.g., age, gender) and environmental factors like vegetation density or ambient noise [5,6]. This restriction hinders their capacity to completely encompass human thermal sensations within intricate urban environments. Moreover, recent advances in machine learning have created promising opportunities to model and predict subjective thermal comfort more precisely by integrating these diverse influences [7,8]. While outdoor thermal comfort is recognized as a multifaceted phenomenon [9], there remains a scarcity of datasets that simultaneously combine microclimatic data, morphological factors, and perception-based information from outdoor settings. Specifically, data collections are still lacking adequate representation from Southern European cities, which are particularly susceptible to intense heat stress [10,11]. Recent datasets, notably those of Tiago et al. [12] and Silva et al. [13], have offered significant insights through itinerant and roving methodologies, delivering additional urban thermal comfort data across various seasons and spatial dimensions.

The urban heat island (UHI) phenomenon, resulting from increased surface temperatures in urban areas, significantly impacts pedestrian thermal comfort, underscoring the importance of the thermal properties of urban surfaces [14,15]. Recent studies, such as that by Middel et al. [16], have employed panoramic infrared thermography to evaluate human thermal exposure, demonstrating that high-resolution 3D surface temperature data are critical in accurately capturing the complex longwave radiation from various materials and enhancing the precision of comfort indices like the UTCI. Likewise, Lamberti et al. [17] integrated infrared thermography with 3D urban modeling to measure directional radiation on specific body parts, advancing the understanding of pedestrian heat exposure and highlighting the importance of detailed surface temperature measurements in urban microclimate studies.

These contributions emphasize that combining surface temperature information from urban materials with microclimatic and morphological variables considerably enhances evaluations of thermal comfort and aids in developing mitigation strategies. Urban planning initiatives like the use of reflective pavements, the extension of green infrastructure, and the selection of adaptive materials have demonstrated quantifiable potential to lessen heat stress [18,19].

To obtain reliable datasets, it is crucial to adhere to well-defined measurement protocols involving calibrated sensors and standardized field techniques, along with quota sampling. This method is vital in accurately capturing subjective thermal sensations, ensuring that environmental and human factors are consistently incorporated into thermal comfort research [20,21,22].

1.2. Related Work

The examination of outdoor thermal comfort and its interaction with urban microclimates has gained more attention due to the acceleration of urbanization and the challenges posed by climate change. Significant research underscores the importance of blending objective measurements of microclimates with personal thermal perceptions to effectively assess and enhance urban thermal comfort [20,21]. These comprehensive methods emphasize that outdoor thermal comfort is shaped by intricate interactions among the climate, urban form, and human physiological and psychological adjustments [9,23]. Recently, expansive datasets that incorporate environmental factors with subjective feedback and urban structural data have improved the understanding of the multifaceted elements influencing outdoor comfort and aided in planning for resilient urban climates [10,12,13,24]. Even with these advancements, there persists a significant demand for freely available and thoroughly annotated datasets that facilitate reproducibility and allow for robust cross-comparisons across various urban settings and climatic conditions. The dataset introduced in this study expands upon the groundwork set by previous initiatives such as the RUROS project [20] and modern machine learning research [25,26] by integrating detailed microclimate data, demographic details, subjective thermal comfort responses, and urban morphological characteristics for an entire city. Such expansive datasets are essential in creating sophisticated thermal comfort models that account for both macro- and microclimate factors, as well as human diversity [7,23,27]. Furthermore, the application of deep learning and convolutional neural networks has recently improved the spatial mapping and predictive precision of thermal comfort metrics, providing urban planners and designers with significant data-driven insights [8,28].

1.3. Study Objectives and Dataset Contribution

This paper introduces an innovative open database gathered from diverse public urban areas in Seville, Spain, over several warm seasons between 2023 and 2025. It integrates on-site microclimatic data collection and in-depth characterizations of the urban morphology, such as street layouts, vegetation, and building densities, along with comprehensive surveys of personal thermal comfort that provide subjective experiences and demographic information. The dataset reflects spatial and temporal variations in residential, tourist, and mixed-use areas, offering a comprehensive foundation for a multifactorial analysis of the factors influencing outdoor thermal comfort. Aimed at promoting the development and validation of sophisticated thermal comfort evaluation tools, this dataset acknowledges the diversity in human physiological and psychological responses, the intricacy of urban structures, and environmental fluctuations. It seeks to support evidence-informed urban planning, climate adaptation plans, and public health strategies by enabling comprehensive multivariate and machine learning evaluations of outdoor thermal comfort adaptability and responses [25,26,29].

2. Study Area and Experimental Context

2.1. Case Study: Seville, Spain

Located in the southern region of Spain, Seville exemplifies a Mediterranean urban area facing heightened thermal stress driven by both climate change and urbanization. With its hot-summer Mediterranean climate (codified as Csa in Köppen’s classification [30]), the city endures extended warm periods and recurrent heatwaves that profoundly impact outdoor thermal comfort. The varied urban layout of Seville, which spans from densely packed historical districts to expansive green areas, renders it an ideal subject for the examination of microclimatic differences and human thermal experiences in intricate urban environments.

2.2. Urban Environment and Climate Overview

The research area encompasses diverse urban layouts with varying characteristics, including the building density, street orientation, vegetation presence, and surface materials. These elements significantly influence local microclimates and subsequently impact the thermal comfort of outdoor environments. The Mediterranean climate intensifies thermal stress during the warmer months from April to October, highlighting the need for an in-depth analysis of these factors to guide urban planning and climate adaptation strategies in Seville.

2.3. Site Selection and Morphological Diversity

Observation sites were strategically selected to include a broad range of morphological types, such as residential zones, commercial areas, and public squares. Figure 1 shows the map of Seville with the different measurement locations included in this study. This was intended to examine how the characteristics of the built environment—like the street canyon geometry, the presence of vegetation, and the thermal properties of materials—affect outdoor thermal comfort [6,31]. Studies show that the height-to-width (H/W) ratio of street canyons and the presence of trees and water features can greatly influence thermal sensations and comfort in urban outdoor environments [23,32]. By sampling these diverse areas, we compiled a comprehensive dataset that reflects the varied thermal environments experienced by Seville’s residents, facilitating advanced modeling and machine learning efforts to accurately predict human thermal perception [10,25].

Table 1. Urban morphological characteristics of the measurement sites.

Longitude	Latitude	Height	Width	HW_Ratio	Pavement	Façades	Nature	Shadow	Water
−6.000	37.399	10.000	50.000	0.200	Asphalt–black pavement	White paint	20–29	20–29	Abundant
−5.998	37.401	10.000	50.000	0.200	Ochre pavement	Ochre brick	20–29	70 or more	Abundant
−5.998	37.401	14.000	20.000	0.700	Ochre pavement	Ochre brick	20–29	40–49	Abundant
−5.997	37.402	10.000	50.000	0.200	Ochre pavement	Ochre brick	20–29	70 or more	Abundant
−5.995	37.404	15.000	14.000	1.071	Ochre pavement	White paint	10–19	Less than 10	Abundant
−5.993	37.404	9.750	31.125	0.313	Ochre pavement	White paint	40–49	50–59	None
−5.992	37.403	8.000	15.000	0.533	Ochre pavement	White paint	Less than 10	50–59	None
−5.990	37.403	16.000	37.500	0.427	Ochre pavement	Ochre brick	Less than 10	40–49	None
−5.989	37.403	16.000	38.333	0.417	Ochre pavement	Ochre brick	Less than 10	40–49	None
−5.989	37.403	16.000	40.000	0.400	Ochre pavement	Ochre brick	20–29	40–49	None
−5.988	37.403	16.000	40.000	0.400	Ochre pavement	White paint	Less than 10	20–29	None
−5.987	37.401	12.000	7.000	1.714	Ochre pavement	Ochre brick	Less than 10	70 or more	None
−5.988	37.401	12.000	27.000	0.444	Ochre pavement	Ochre brick	10–19	20–29	None
−5.999	37.385	12.000	24.000	0.500	Ochre pavement	Concrete–red brick	Less than 10	Less than 10	Abundant
−5.996	37.383	11.000	10.000	1.100	Concrete–red pavement	Concrete–red brick	Less than 10	30–39	Abundant
−5.996	37.385	3.000	12.000	0.250	Asphalt–black pavement	White paint	Less than 10	10–19	Occasional
−5.997	37.386	10.000	50.000	0.200	Concrete–red pavement	White paint	70 or more	70 or more	None
−5.999	37.388	20.000	68.000	0.294	Asphalt–black pavement	White paint	Less than 10	50–59	None
−5.997	37.388	12.000	30.000	0.400	Asphalt–black pavement	Concrete–red brick	Less than 10	40–49	None
−5.995	37.387	9.000	5.250	1.714	Concrete–red pavement	White paint	30–39	40–49	None
−5.994	37.385	8.000	15.000	0.533	Concrete–red pavement	White paint	10–19	10–19	None
−5.992	37.386	16.000	40.000	0.400	Concrete–red pavement	White paint	10–19	10–19	Abundant
−5.993	37.387	9.000	6.000	1.500	Ochre pavement	White paint	30–39	50–59	None
−5.985	37.380	12.000	30.000	0.400	Asphalt–black pavement	Vegetation	30–39	Less than 10	None
−5.985	37.382	15.000	35.000	0.429	Concrete–red pavement	Concrete–red brick	Less than 10	20–29	None
−5.984	37.383	17.000	3.000	5.667	Asphalt–black pavement	Concrete–red brick	30–39	10–19	Occasional
−5.986	37.383	20.000	26.000	0.769	Asphalt–black pavement	White paint	10–19	20–29	None
−5.987	37.383	14.000	15.000	0.933	Asphalt–black pavement	White paint	10–19	10–19	Occasional

shows the different combinations of morphological characteristics present at the measurement sites where data were collected.

3. Materials and Methods

3.1. Data Collection Protocols

Data on microclimatic conditions were gathered by architecture graduates utilizing precise instruments like the ones shown in Figure 2. These measurements adhered to the protocols recommended in urban outdoor comfort studies. Following ASHRAE Standard 55 [33], all probes were placed at a height of approximately 1.1 m above the ground to represent the pedestrian level. The thermohygrometer and the anemometer were kept in an unobstructed orientation to ensure accurate readings. During each measurement, the instruments were held stable at the specified height in accordance with standard thermal comfort analysis practices. Researchers collected morphological data directly from the site, as illustrated in Figure 3, adhering to methods described in studies pertaining to the urban form and thermal comfort. The methodology for personal surveys utilized a quota sampling technique segmented by gender and age, conducted via intercept surveys in urban public areas. The complete questionnaire used for the surveys can be found in Appendix A. These surveys predominantly utilized questionnaires modified from the works of [20] This approach, which involved randomly encountering participants, is a standard practice in surveys examining thermal comfort. Of the individuals approached, roughly 20% opted to participate, with most completing the surveys completely while on location. In total, 989 surveys were completed. This participation rate is consistent with those found in other studies of outdoor comfort. The basic descriptive statistics of the respondents showed a balanced gender distribution (53.4% female, 46.0% male, and 0.6% preferred not to say). The survey aimed to cover all age groups, with the largest representation in the 30–39 age group. In addition, the respondents represented 32 different nationalities.

3.2. Fixed Outdoor Monitoring

In Seville, Spain, data collection occurred during the warmer months, from April to October, over the years 2023 to 2025. As outlined in Section 2.2, the observation locations were carefully selected across public urban areas to represent various urban morphologies, including residential zones, tourist areas, plazas, street canyons, and parks. The criteria for selecting these sites were designed to ensure diversity in vegetation, building density, and surface materials, providing comprehensive coverage of Seville’s diverse urban environment, in line with methodologies for the study of urban microclimates and thermal comfort. The surface temperature measurements were taken on up to three different materials located near the observation point. These included a wide variety of urban surfaces, such as land, metal, playgrounds, and multiple types of pavement (e.g., black asphalt, cobblestone, grass, grey concrete, light stone, ochre, red concrete, white, dark stone, green concrete), as well as façades (e.g., ochre brick, red brick–concrete, vegetation, white paint). Surveys were administered at each site multiple times, creating groups of responses within identical microclimatic and morphological settings. Architecture graduates carried out all data collection tasks, a practice recommended in biometeorological studies to maintain high data quality and accurate contextual understanding. Moreover, they documented the specific morphological features of each location, capturing variations in urban form that are essential in evaluating thermal comfort.

Figure 3. Data gathering: personal surveys using a standardized questionnaire and obtainment of microclimatic and morphological data.

Figure 3. Data gathering: personal surveys using a standardized questionnaire and obtainment of microclimatic and morphological data.

3.3. Mobile Transects

Mobile survey campaigns, such as those conducted on foot or by bike, follow established protocols to ensure the simultaneous gathering of both subjective and objective data, as outlined in recent machine learning studies on urban thermal comfort. Routes were chosen to ensure the broad spatial representation of various urban environments, while portable microclimatic devices and survey tools allowed for on-the-go data collection. The alignment of microclimatic measurements with thermal sensation surveys from participants adheres to the recommended practices for the accurate capture of dynamic outdoor comfort levels.

3.4. Measured Parameters and Equipment

Fixed Stations and Manual Measurements

Environmental data were collected on-site using portable, calibrated instruments, selected for their accuracy and reliability in outdoor urban conditions. The key measurement devices included the following:

• Thermohygrometer: Employed to measure the air temperature, globe temperature, and relative humidity. The black globe measuring head has a diameter of approximately 5 cm, while international standards recommend 15 cm globes. The smaller size alters the balance between radiation and convection, increases the sensitivity to changes in solar radiation and wind, and reduces thermal inertia. Under direct sunlight, it may overestimate shortwave radiation, while, at higher wind speeds, it tends to underestimate the radiant effect. These limitations prevent direct comparison with measurements taken using standardized globes, especially in comfort or heat stress studies. The instrument offers accuracy of ±0.6 °C for air temperature measurements, ±0.6 °C for globe temperature measurements in the 20–50 °C range and ±1 °C outside this range, and ±5% for relative humidity. The air temperature can be measured from 0 to 50 °C, the globe temperature from 0 to 80 °C, and relative humidity from 0 to 99.9%, ensuring precision in capturing microclimatic conditions relevant to thermal comfort assessment.
• Anemometer: Utilized to capture wind speed data, with accuracy of ±0.2 m/s, allowing the detailed characterization of airflow variability in urban canyons and open spaces.
• Surface Thermometer: Used to measure the surface temperature in a range of −50 to 350 °C, with accuracy of ±(1 °C + 1% of the measured value). Its probe allows measurements to be taken on various surfaces.

Morphological parameters, critical in contextualizing the microclimatic observations, were documented through systematic manual surveys conducted by trained researchers at each measurement site. These included the average building height, street width, and estimated vegetation cover (naturalization, measured on a 0–100% scale in 10% increments, with each increment corresponding to the estimated green cover at the site).

4. Dataset Description

4.1. Dataset Organization and File Formats

The dataset is provided in both .xlsx and .csv formats to ensure accessibility and usability for a wide range of users. The file UTHECA_USE.xlsx contains two spreadsheets. The first, named data, includes the complete main dataset. The second, named supplementary, contains additional variables for a subset of records. These correspond to the .csv files UTHECA_USE.csv and UTHECA_USE_additional.csv, respectively.

4.2. Variable Definitions and Metadata

The main dataset contains 989 observations and 55 variables. Each row represents an individual surveyed at a specific time and place, with multiple surveys possibly sharing the same microclimatic and morphological context when conducted at the same locations. The first column (id) uniquely identifies each record, followed by variables detailing temporal information (Date, Hour), geographic coordinates (Longitudinal.coord,Latitudinal.coord), and specific location descriptors (Location). The remaining variables fall into three categories:

• Microclimatic: Objective measurements including Air.temperature, Globe.temperature, Relative.humidity, Wind.speed, Radiant.temperature, UTCI, and UTCI.classification.
• Personal survey: Demographic and physical characteristics (Height.cm, Weight.kg, BMI), clothing insulation (Clothing.cover.clo), and subjective thermal sensation and well-being metrics (Subjective.thermal.sensation, Subjective.well.being.sensation), as well as perception-related variables such as Vegetation.likeable.
• Morphological: Urban environmental descriptors such as Average.height, Street.width, Naturalization, and Noise.level, among others.

The supplementary dataset contains 820 observations and 3 variables, providing surface temperature data for various urban materials (pavements, façades, urban furniture). Entries link to the main dataset via the id variable. Material types are classified under Surface.temperature.material, with corresponding temperature measurements in Surface.temperature.

In

Table 2. Identification and temporal variables.

Name	Type	Description
id	Integer	Identification number (1–989)
Hour	Numeric	Time of interview/measurement
Date	Date	Data collection date (yyyy-mm-dd)
Longitudinal.coord	Numeric	Site longitude
Latitudinal.coord	Numeric	Site latitude
Location	Categorical	Qualitative location name (e.g., Avenida De La Constitución, Calle Torneo, etc.)

,

Table 3. Microclimatic variables.

Name	Type	Description
Air.temperature	Numeric	Air temperature (°C), Trotec TC100 (±0.6 °C)
Globe.temperature	Numeric	Globe temperature (°C), Trotec TC100 (±0.6 °C)
Relative.humidity	Numeric	Relative humidity (%), Trotec TC100 (±5%)
Wind.speed	Numeric	Wind speed (m/s), Trotec TA300 (±0.2 m/s)
Radiant.temperature	Numeric	Mean radiant temperature (°C), UNE-EN ISO 7726:2002 [34]
UTCI	Numeric	Universal Thermal Climate Index value
UTCI.classification	Categorical	UTCI thermal stress class (e.g., no thermal stress, moderate heat stress, etc.)

,

Table 4. Personal survey variables.

Name	Type	Description
Height.cm	Numeric	Height in centimeters
Weight.kg	Numeric	Weight in kilograms
BMI	Numeric	Body mass index: weight/(height/100)2
Clothing.cover.clo	Numeric	Clothing insulation (clo units, 0.2–1.2)
Activity.met	Numeric	Metabolic activity (met units, 0.8–8)
Visit.frequency	Numeric	Visit frequency (1 = rarely to 5 = always)
Mood	Numeric	Mood (1–5)
Gender	Categorical	Female, male, prefer not to say
Age.range	Categorical	Age (e.g., 20–29, 30–39, etc.)
Nationality	Categorical	Respondent’s nationality
Living.in.Seville	Categorical	Duration living in Seville
Accompanied.by	Categorical	Accompanied during interview (Yes/No)
You.answered	Categorical	Posture (sitting, standing, walking)
Shaded.during.survey	Categorical	Sun/shade exposure during survey
Time.here	Categorical	Time at location before survey
Previously.in	Categorical	Previous environment (e.g., air-conditioned interior)
Reason.visit	Categorical	Reason for presence (family, tourism, work, etc.)
Would.change.Temperature	Logical	Wants to change temperature
Would.change.Humidity	Logical	Wants to change humidity
Would.change.Wind	Logical	Wants to change wind
Would.change.Radiation	Logical	Wants to change radiation
Would.change.Nothing	Logical	Would not change anything
Vegetation.likeable	Numeric	Vegetation: unpleasant (−1), neutral (0), pleasant (1)
Materiality.likeable	Numeric	Materiality: unpleasant (−1), neutral (0), pleasant (1)
Spatiality.likeable	Numeric	Spatiality: unpleasant (−1), neutral (0), pleasant (1)
Noise.likeable	Numeric	Noise: unpleasant (−1), neutral (0), pleasant (1)
Number.of.people.likeable	Numeric	Number of people: unpleasant (−1), neutral (0), pleasant (1)
Shade.likeable	Numeric	Shade: unpleasant (−1), neutral (0), pleasant (1)
Subjective.thermal.sensation	Categorical	Perceived thermal sensation (e.g., cold, neutral, hot)
Subjective.well.being.sensation	Categorical	Perceived comfort (very uncomfortable–very comfortable)

and

Table 5. Morphological variables.

Name	Type	Description
Average.height	Numeric	Average building height (m)
Street.width	Numeric	Street width (m)
Naturalization	Categorical	Naturalization degree (e.g., less than 10, 10–19, etc.)
Shadow	Categorical	Shade percentage at location
Water	Categorical	Water feature presence (none, occasional, abundant)
Pedestrians	Categorical	Pedestrian traffic level
Traffic	Categorical	Vehicular traffic level
Noise.level	Categorical	Ambient noise (almost none–very intense)
Orientation	Categorical	Street orientation (N-S, E-W, etc.)
Pavement	Categorical	Pavement type (e.g., asphalt, grass)
Façades	Categorical	Façade type (e.g., red brick, white paint)
Urban.furniture	Categorical	Urban furniture (none, occasional, abundant)

, we show all the main variables included in the UTHECA_USE dataset.

4.3. Data Cleaning, Quality Control, and Imputation

Meticulous procedures for data quality assurance were enacted throughout the research to guarantee the dataset’s thoroughness and dependability. A mixture of methods was specifically adapted to the types and patterns of the missing data for imputation. For missing age group data, mode imputation was applied, affecting about 0.8% of cases, thus maintaining the variable’s categorical essence. In the case of microclimatic variables, with 3.6% data absence, K-nearest neighbors (KNN) [35] imputation was utilized, incorporating temporal, spatial, and morphological covariates to enhance the plausibility of imputed values by considering environmental factors [36,37]. For a single participant (0.1%) with missing anthropometric data (height and weight), mean imputation was applied, which was appropriate given the minimal missing data. This process, along with other data cleaning and imputation tasks, was performed in the R programming language using the Tidyverse suite of packages. Tidyverse allows for efficient data manipulation, ensures script reproducibility, and increases transparency in data preprocessing workflows [38,39]. Its coherent data manipulation grammar and integration with specialized packages enable the effective management of complex datasets, such as those that incorporate microclimatic, morphological, and survey data. This methodological precision supports the robustness of subsequent analyses and predictive modeling endeavors.

Finally, making use of these tools, we validated the data by evaluating the consistency between objective measurements and survey responses to ensure that the results were coherent. For the sake of transparency, details about the sensor models, measurement heights, and data logging systems utilized are provided.

4.4. Supplementary Datasets and Derived Variables

4.4.1. Derived Thermal Comfort Metrics

To ensure the dataset’s integrity and comprehensiveness, data quality assurance measures were carefully executed. All preprocessing activities, from cleaning to the computation of the derived variables, were undertaken in R, leveraging the Tidyverse suite extensively [38,39]. The Tidyverse collection presents a unified and reproducible system for data manipulation, further improved by integration with packages like ’themis’ to deal with imbalanced data [40] and ’ROSE’ to apply oversampling methods [41]. This approach enables robust and transparent workflows when managing complex datasets, ensuring the integrity and dependability of later statistical analyses and the creation of machine learning models.

4.4.2. Supplementary Surface Temperature Dataset

A complementary dataset encompassing 819 observations was collected using a hand-held surface thermometer to measure the surface temperatures of various urban materials, including pavements, façades, and benches. Each measurement was classified by material type and linked to corresponding microclimatic and morphological data. Information about the supplementary data can be found in

Table 6. Supplementary variables.

Name	Type	Description
id	Integer	Identification number ranging from 1 to 989
Surface.temperature	Numeric	Surface temperature (°C) measured using a Testo 905-T2 surface thermometer (Testo Austria, Vienna, Austria) (accuracy: ±1 °C + 1% of the measured value)
Surface.temperature.material	Categorical	Material where the surface temperature was measured. Categories: land, metal, pavement—black asphalt, pavement—cobblestone, pavement—grass, pavement—grey concrete, pavement—light stone, pavement—ochre, pavement—red concrete, pavement—white, playground, façade—ochre brick, façade—red brick–concrete, façade—vegetation, façade—white paint, pavement—dark stone, pavement—green concrete

.

4.5. Data Documentation and Accessibility

To facilitate data reuse and clarity, the dataset is accompanied by a comprehensive documentation file, CODEBOOK_UTHECA_USE.pdf. This codebook summarizes all variables from both datasets, specifying the variable names, data types (numeric, categorical, logical), definitions, units or value ranges, and details of the measurement instruments when applicable. Additionally, footnotes describe the data imputation methods (mean, mode, and KNN-based techniques) and clarify their scope within the dataset.

5. Results

5.1. Exploratory Data Analysis

Understanding how environmental measurements relate to subjective thermal perceptions is crucial in assessing the credibility and usefulness of comfort indices. In this section, we examine the dataset’s framework and patterns using descriptive statistics and visual analyses, with a focus on the correlations (or their absence) between objective indicators—like the Universal Thermal Climate Index (UTCI)—and participants’ personal thermal perceptions. This analysis aims to emphasize underlying trends, inconsistencies, and potential biases that could impact thermal comfort evaluations and the formulation of predictive models.

The comparison between the biometeorological index UTCI and participants’ self-reported thermal sensations reveals a notable discrepancy between objective classification and subjective experience. Ideally, a reasonable correspondence between UTCI categories and perceived thermal sensations would be expected; however, as we can see in Figure 4, our findings suggest otherwise.

In the ’no thermal stress’ category, 62% of respondents reported feeling neutral, which aligns with the interpretation of the index. Nevertheless, a substantial portion (approximately 20%) reported feeling warm, including sensations such as ’heat’ and even ’very hot’. Similarly, under ’moderate heat stress’, more than 50 individuals perceived the conditions as neutral or even slightly cold, indicating the potential underestimation or overestimation of actual thermal comfort by the index.

This bias highlights that biometeorological indices based solely on physical variables—such as the air temperature, humidity, wind speed, and radiation—are not sufficient to accurately represent the complexity of human thermal perception. Sensation is influenced by various personal factors (age, gender, activity level, clothing), psychological components (mood, thermal expectations), and contextual conditions (shade, vegetation, urban form).

Therefore, it is essential to complement objective thermal indices with perceptual and contextual indicators, particularly when assessing thermal comfort in real urban environments. Such an approach allows for a more comprehensive understanding of the outdoor thermal experience and provides a stronger foundation for the design of thermally inclusive urban spaces.

5.1.1. Selected Variables

In the previous analysis, we used an ANOVA test to examine whether categorical variables had a significant effect on subjective thermal sensation, which is appropriate for ordered response data. For continuous variables, we applied linear regression to assess how changes in these variables influenced subjective thermal sensation. The reported p-values, shown in

Table 7. Statistical tests for variables affecting subjective thermal sensation. Values up to the 5th decimal.

Variable	p-Value	Type
Location	0.00000	Categorical
UTCI.classification	0.00000	Categorical
Nationality	0.00000	Categorical
Shaded.during.survey	0.00000	Categorical
Previously.in	0.00000	Categorical
Subjective.well.being.sensation	0.00000	Categorical
Shadow	0.00000	Categorical
Water	0.00000	Categorical
Noise.level	0.00000	Categorical
Pavement	0.00000	Categorical
Façades	0.00000	Categorical
Time	0.00000	Categorical
Traffic	0.00001	Categorical
Urban.furniture	0.00004	Categorical
Age.range	0.00006	Categorical
Reason.visit	0.00027	Categorical
Naturalization	0.00109	Categorical
Pedestrians	0.00251	Categorical
Gender	0.00429	Categorical
Living.in.Seville	0.06102	Categorical
Time.here	0.08524	Categorical
You.answered	0.08978	Categorical
Orientation	0.13214	Categorical
Accompanied.by	0.15197	Categorical

, indicate the strength of evidence for each variable’s association with the subjective thermal sensation.

The analysis of the relationship between gender and subjective thermal sensation, restricted to the time window between 12:00 p.m. and 6:00 p.m., revealed statistically significant differences in response patterns. As shown in Figure 5, although ’neutral’ remained the most frequent category across genders, women exhibited a higher proportion of responses at the thermal extremes (’heat’ and ’very hot’) compared to men. Specifically, 72.5% of women reported ’heat’ or ’very hot’, whereas this percentage was 53% among men. Conversely, men reported neutral sensations more frequently (38.5%) than women (22.5%).

These differences were confirmed through a chi-squared test (${\chi }^2$ = 8.47; df = 3; p = 0.037), indicating a statistically significant association between gender and subjective thermal sensation during the period of peak solar exposure.

Similarly, the age group analysis revealed an even stronger and statistically significant association (${\chi }^2$ = 50.47; df = 18; p < 0.0001). As detailed in Figure 6, groups under 40 years old exhibited a wider distribution of responses, including ’very hot’ sensations, whereas older groups tended to concentrate their responses in more moderate thermal categories.

These results reinforce the evidence that both gender and age significantly influence subjective thermal perception. Previous studies have consistently shown that physiological factors (such as thermoregulation and metabolic rate) and behavioral aspects (such as clothing and activity level) vary with sex and age, directly affecting thermal experiences. These findings highlight the importance of incorporating personal variables into thermal comfort models to achieve more accurate, inclusive, and user-centered assessments [21].

Figure 7 shows the values of the mean radiant temperature (MRT) recorded at 1:00 p.m. and 2:00 p.m. on April 2023, classified by the predominant façade type in the immediate surroundings. The data reveal clear differences in the thermal performance of urban spaces according to their construction materials and surface treatments.

Vegetation stands out as the façade type associated with the lowest MRT values, with a median of 30.45 °C, compared to the significantly higher medians found for red or ochre brick façades (above 46 °C) and white-painted façades (37.92 °C). This difference of approximately 7 degrees Celsius between vegetation and the hotter façades is particularly relevant from a thermal comfort perspective.

The median, as a measure of central tendency, is especially useful here because it is less influenced by extreme values than the mean. It better represents a typical condition in skewed distributions, which are common in thermal variables in urban environments.

These findings emphasize the role of vegetation as a mitigating element in urban thermal exposure and support climate-sensitive design strategies in public space planning.

5.1.2. Correlations

In Figure 8, we present a visualization of the correlation matrix, which depicts the pairwise Pearson correlation coefficients among the 25 numerical variables related to environmental conditions, personal attributes, and subjective likes or feelings in an observational study. Strong positive correlations are observed between Air.temperature, Globe.temperature, Radiant.temperature, and UTCI, with coefficients exceeding 0.9, indicating that these measures largely track the same thermal conditions. Relative.humidity is strongly negatively correlated with Air.temperature ($-0.82$) and Globe.temperature ($-0.75$), reflecting the typical inverse relationship between temperature and humidity. Personal physical metrics such as Height.cm, Weight.kg, and BMI show moderate positive correlations within each other, especially Weight.kg and BMI (0.81). Subjective variables like Mood and Visit.frequency have moderate positive correlations with each other and with Clothing.cover (clo) (e.g., Mood correlates with Clothing.cover at $0.15$ and Visit.frequency at $0.19$), suggesting some link between how often people visit, how they dress, and their moods. Environmental preference variables (Vegetation.likeable, Materiality.likeable, Spatiality.likeable, Noise.likeable, Number.of.people.likeable, and Shade.likeable) generally show weak correlations among themselves and with other variables, indicating that subjective likes are relatively independent of the measured environmental conditions and physical traits in this dataset. Overall, the matrix highlights the expected physical relationships between environmental variables and some modest connections between personal attributes and subjective experiences.

Figure 7. Mean radiant temperature by façade type.

Figure 7. Mean radiant temperature by façade type.

6. Discussion

6.1. Dataset Value and Key Insights

This dataset represents the robust integration of microclimatic data, urban morphological characteristics, and personal survey insights, enabling a comprehensive analysis of the multifaceted factors influencing outdoor thermal perception. By incorporating both environmental and personal elements, it assists researchers and urban planners in pinpointing key determinants of extreme thermal sensations and discomfort, crucial for effective climate adaptation strategies [21,42]. Additionally, this dataset provides a solid basis for the creation and validation of predictive models, including advanced machine learning tools such as random forest, XGBoost, and neural networks [43,44,45], to forecast subjective thermal responses using objective environmental and personal data, increasing the accuracy over traditional indices [27,46]. Incorporating detailed urban morphology variables, such as the widths of streets, average heights of buildings, and types of surface materials, enhances the ability to assess their impacts on thermal comfort and aids in creating urban areas resilient to heat [4,32]. Moreover, additional data that correlate the surface temperatures of different urban elements—like pavements, building façades, and street furniture—with local microclimatic conditions offer valuable insights into the thermal properties of these materials in real-world outdoor settings, supporting evidence-based choices in material selection to alleviate urban heat [47,48]. This extensive dataset is crucial in advancing scientific knowledge and practical solutions in urban thermal comfort and sustainable urban planning.

6.2. Comparison with Previous Studies

The UTHECA_USE dataset focuses on the city of Seville, Spain, during the warm months—a period of particular interest due to the pronounced urban heat island effect characterizing the city in summer. This temporal and geographical scope was deliberately chosen to accurately capture the critical microclimatic conditions affecting urban thermal comfort, rather than aiming for year-round representativeness. With 989 records, UTHECA_USE offers comprehensive sampling for a single urban environment. By comparison, the RUROS dataset [20]—collected between 2001 and 2002—comprises approximately 10,000 observations across five countries and fourteen sites, only two of which exceeded the number of interviews collected in the present study. UTHECA_USE’s questionnaire design was informed by RUROS, incorporating core variables such as age group, sex, shading availability, and reasons for visiting, while optimizing the survey length and relevance. Notably, UTHECA_USE enhances the thermal perception detail by using ten levels for thermal sensation and five levels for comfort, compared to RUROS’s simpler scales, thereby capturing wider variability and minimizing neutral response bias. Additional personal variables, such as height, weight, and mood, which the recent literature indicates may influence thermal perception, were also incorporated. The microclimatic data collected included all parameters required to compute the UTCI, complemented by detailed morphological variables for each measurement point—such as the dominant pavement and façade materials, vegetation, water presence, average building height, and street width. A unique feature of UTHECA_USE is the inclusion of surface temperature readings for various urban materials, allowing integrated studies of materials’ responses to environmental conditions.

In comparison, the Lisbon thermal comfort dataset from Tiago et al. [12] presents data obtained using itinerant meteorological measurements combined with surveys administered during walking routes across multiple urban morphological settings throughout all seasons. Their meteorological data are highly detailed, including the air temperature, relative humidity, wind speed/direction, dew point, black globe temperature, and both shortwave and longwave radiation components, allowing the calculation of the UTCI and mean radiant temperature at a high temporal resolution (every 5 to 10 s). Participants provided pre-walk sociodemographic and clothing insulation data, alongside thermal sensations, with more comprehensive thermal pleasantness and preference surveys conducted during the walks. This mobile and multi-seasonal approach contrasts with UTHECA_USE’s fixed stations focused on warm seasons, but both datasets offer critical complementary insights into pedestrian thermal comfort and the microclimate across different urban contexts.

The study by Silva et al. (2024) [13] similarly used mobile roving missions conducted throughout all seasons and during the day and night, explicitly incorporating local climate zone (LCZ) spatial scales to characterize pedestrian thermophysiological comfort. Their approach enabled the identification of typologies of urban areas with varying comfort levels, finding significant thermal stress differences temporally and spatially—such as a 44% increased UTCI at night and specific discomfort in compact, low-rise areas. The higher spatial and temporal resolutions of their dataset complement the UTHECA_USE dataset’s fixed monitoring and detailed morphological data, together enriching the understanding of urban thermal environments.

Both the Lisbon and Seville studies shared a common objective: to improve the understanding of outdoor thermal comfort in urban environments in order to support more sustainable planning and enhance citizens’ quality of life. In both cases, the Universal Thermal Climate Index (UTCI) was employed as a fundamental objective metric, and microclimatic variables at the pedestrian level were measured using portable equipment. This shared framework ensured comparability while allowing each study to address different aspects of the problem. The methodological approaches, however, diverge. The Lisbon study was structured around mobile transects across the city, using meteorological stations mounted on carts to capture high-resolution spatial variability in thermal conditions. This approach facilitated the identification of patterns at the scale of local climate zones (LCZs) and neighborhoods, producing a climate-oriented diagnosis of the city throughout the year. By contrast, the Seville study focused on strategically selected fixed points that represented diverse urban morphologies, where both environmental measurements and on-site surveys of passersby were conducted. This design allowed the direct coupling of physical conditions with individual perceptions, generating insights into the interplay between objective and subjective dimensions of thermal comfort. The datasets also differ in scope and depth. While both include core microclimatic variables such as the air temperature, humidity, and wind, the Lisbon dataset emphasizes spatial coverage and classification according to LCZs, whereas the Seville dataset incorporates a broader set of descriptors, including detailed morphological features for each location and individual-level data such as demographics, clothing, activity, and thermal sensations. As a result, the Lisbon study offers a comprehensive climatological mapping of urban comfort, while the Seville study provides a more integrated view that combines environmental, morphological, and personal dimensions. Together, these complementary approaches contribute valuable perspectives to advance the study of outdoor thermal comfort in urban contexts.

The integrated detail in UTHECA_USE, consisting of morphological and surface temperature data combined with refined subjective thermal sensation scales, provides a valuable contribution, complementing these earlier datasets and enabling advanced urban thermal comfort modeling focused on hot Mediterranean climates.

6.3. Limitations and Future Work

For variables such as height, weight, and age range, when the surveyed person had to leave prematurely, before completing the questionnaire, values were estimated by the interviewer when there was a reasonable degree of visual certainty; otherwise, the response was left blank. This methodological choice, while necessary to preserve anonymity, resulted in some missing values that were subsequently imputed.

It was also necessary to impute certain microclimatic variables due to occasional interruptions in measurements, caused by technical failures in the sensors or unstable environmental conditions, which prevented the collection of consistent data in some observations.

7. Conclusions

The UTHECA_USE dataset comprises 989 instances collected in Seville, Spain, during the years 2023 to 2025, specifically in the warmer months of April through October. It includes 55 variables organized into three primary categories: microclimatic information (like air and globe temperatures, relative humidity, wind speed, mean radiant temperature, and UTCI) and personal survey data (covering demographic details, clothing insulation, metabolic rates, and subjective thermal comfort assessments), as well as urban morphological characteristics (including building heights, street widths, naturalization, pavement and façade types, and noise levels). Some cases also documented the surface temperatures of urban materials.

Our study exposed significant differences between the objective Universal Thermal Climate Index (UTCI) categorizations and the subjective thermal perceptions of participants. Notably, in the UTCI’s ’no thermal stress’ classification, 62% of respondents indicated a neutral feeling, while about 20% felt warm, and a few described sensations such as ’heat’ or ’very hot’. In the ’moderate heat stress’ category, some participants felt neutral or even cold, indicating that traditional biometeorological indices might not always accurately reflect true human thermal comfort in specific situations (Figure 4).

Moreover, individual factors like gender and age played a substantial role in shaping thermal perception. Women reported extreme heat sensations more frequently (72.5% noting ’heat’ or ’very hot’) compared to men (53.0%), with a statistically significant link (chi-squared test: $p=0.037$; Figure 5). Age showed even more distinct variations; younger individuals (under 40 years) experienced a wider array of warmer sensations, while older participants predominantly reported moderate sensations (chi-squared: $p<0.0001$; Figure 6).

Morphologically, measurements of the mean radiant temperature (MRT) revealed evident differences based on the surrounding façade material. Areas with vegetation-covered surfaces showed median MRTs of around 30.5 °C, which was notably lower than for façades composed of red or ochre brick, which had temperatures exceeding 46 °C, demonstrating vegetation’s role in reducing thermal exposure (see Figure 7).

Correlation analyses indicated strong positive relationships between meteorological variables like the air temperature, globe temperature, radiant temperature, and UTCI, with correlation coefficients over 0.9, whereas relative humidity showed strong negative correlations. Personal characteristics (such as height, weight, and BMI) displayed moderate correlations, and subjective attributes like mood and frequency of visit were linked to clothing insulation, revealing the intricate interactions that affect thermal perception (refer to Figure 8).

Overall, these results confirm that the UTHECA_USE dataset effectively captures both the objective and subjective dimensions of urban thermal comfort, enabling complex modeling that considers a range of human and environmental factors.

Our comprehensive and diverse methodology, which merges standardized weather data, detailed morphological descriptions, and thorough personal survey information, offers a solid basis for upcoming investigations. The extensive scope and size of the dataset enable comparative research and sophisticated analyses, such as statistical modeling and machine learning, to more effectively guide urban planning and public health strategies targeting heat stress reduction.

In essence, UTHECA_USE showcases how cohesive datasets encompassing both empirical environmental data and personal human responses can reveal the subtleties of outdoor thermal comfort, particularly in Mediterranean urban areas confronting escalating heat dangers as a result of climate change.