Exploring the Urban Heat Island Effect: A Bibliometric and Topic Modeling Analysis

Abstract

The urban heat island (UHI) effect, intensified by urbanisation and climate change, leads to increased urban temperatures and poses a serious environmental challenge. Understanding its causes, impacts, and mitigation strategies is essential for sustainable urban planning. The aim of this study is to systematically analyse how the Urban Heat Island (UHI) effect has been addressed in the scientific literature, to identify key research themes and their temporal evolution, and to critically highlight knowledge gaps in order to provide guidance for future research and urban planning policies. Using BERTopic, an advanced natural language processing (NLP) tool, the study extracts dominant themes from a large corpus of academic literature and tracks their evolution over time. A total of 9061 research articles from the Web of Science database were collected, pre-processed, and analysed. BERTopic clustered semantically related topics and revealed their temporal dynamics, offering insights into emerging and declining research areas. The results show that pavement materials and urban vegetation are among the most studied themes, highlighting the importance of surface materials and green infrastructure in mitigating UHI. In line with this aim, the study identifies a rising interest in urban cooling strategies, particularly reflective surfaces and ventilation corridors. Consistent with its aim, the study provides a comprehensive overview of UHI literature, critically identifies existing gaps, and proposes clear directions for future research. It provides supports for urban planners, policymakers, and researchers in developing data-driven strategies to mitigate UHI impacts and strengthen enhance urban climate resilience.

1. Introduction

The Urban Heat Island Effect (UHI) is an important problem with environmental, social and economic dimensions that has arisen as a result of modern urbanisation and increasing population density [1,2]. The phenomenon of urban heat islands, characterised by the elevated temperatures of city centres relative to their surroundings, has been demonstrated to precipitate a number of deleterious consequences. These include an escalation in energy consumption, an acceleration in environmental degradation, and an exacerbation in public health issues [3,4]. In the contemporary era, characterised by the exacerbation of the repercussions of climate change, the consequences of UHI have become increasingly evident. In conjunction with extreme heat waves, this phenomenon exerts substantial pressure on public health systems [5]. This situation has prompted a diverse range of actors, including urban planners, scientists, politicians and public health experts, to explore potential solutions [6]. The primary factors contributing to the phenomenon of Urban Heat Island (UHI) include the utilisation of artificial materials such as concrete and asphalt, which replace natural surfaces, the diminution of green areas, anthropogenic heat emissions resulting from fossil fuel consumption, and elevated energy consumption [7,8]. The synergistic effect of these factors contributes to the sustained elevated temperatures in urban areas, both during daylight hours and throughout the nighttime period. This phenomenon has been observed to result in an escalation in mortality rates during periods of extreme heat. Moreover, the UHI effect has been demonstrated to exert an influence on both individual comfort levels and energy demand, thereby exacerbating economic pressures and engendering heat-related health complications [5].

Research aiming to comprehend and alleviate the consequences of UHI is often oriented towards an examination of urban physical characteristics, climatic factors, and population density [9,10]. The utilisation of remote sensing technologies and Geographic Information Systems (GIS) has been a pervasive approach in the analysis of the spatial distribution of the Urban Heat Island (UHI) phenomenon. Furthermore, the integration of climate modelling and thermal imaging techniques has facilitated a more profound comprehension of the temporal variability in UHI impact [11,12]. However, the extant literature is predominantly constrained to a specific regional or city scale and does not adequately address the different contexts of UHI on a global scale. For instance, in developed countries, green infrastructure solutions for UHI mitigation are a priority, while in developing countries the UHI impact is more pronounced due to uneven urbanisation and inadequate infrastructure.

The utilisation of big data and artificial intelligence (AI) holds considerable promise in terms of facilitating novel advancements in the field of analysing the impact of UHI [13]. In this context, the employment of natural language processing tools such as BERTopic (version 0.16.4) facilitates the identification of knowledge gaps in the literature by extracting meaningful topics and trends from large-scale text data, thus enabling the exploration of new research areas. BERTopic is a sophisticated instrument that facilitates the systematic analysis of the subjects that have gained importance over time in the academic literature on UHI, as well as the subjects that have been less extensively studied. The objective of this study is to systematically reveal trends, knowledge gaps and future research areas in the existing literature on the UHI effect using BERTopic. The aim of this study is to systematically analyse how the Urban Heat Island (UHI) effect has been addressed in the scientific literature, to identify key research themes and their temporal evolution, and to critically highlight knowledge gaps in order to provide guidance for future research and urban planning policies. To operationalize this aim, the study also addresses three guiding research questions, which structure the analysis presented in the following sections.

RQ1: Which thematic categories and specific aspects of Urban Heat Island (UHI) research can be systematically identified in the scientific literature, and how can they be distinguished based on disciplinary scope, methodological approach, and conceptual focus?

RQ2: How have research themes in UHI studies evolved across the periods 2004–2024 and what patterns mark the rise or decline of specific topics?

RQ3: What knowledge gaps remain in UHI research, especially regarding geographic underrepresentation, limited integration of socio-economic dimensions, and methodological shortcomings?

It is an established fact that strategies developed with the intention of mitigating the negative impacts of UHI frequently include methods such as the following: increasing green infrastructure, improving energy efficiency, increasing water surfaces and the use of reflective materials. Nevertheless, the efficacy of these strategies is contingent on various climatic conditions and regional dynamics. To illustrate this point, it is notable that cooling wind corridors and shading measures are more effective in tropical cities, while water management strategies may play a more critical role in mitigating the impact of UHI in arid regions. The analyses conducted with BERTopic in this study will facilitate a more profound comprehension of the variability of the UHI impact across diverse geographical and climatic contexts, thereby contributing to the development of more efficacious strategies at both regional and global levels. The present study has two further aims. Firstly, it seeks to establish a framework for addressing the UHI effect which incorporates the disciplines of environmental sciences, meteorology, urban planning, energy engineering and public health [14]. The visualisation tools offered by BERTopic will provide an important resource for understanding current trends and identifying future research and implementation strategies.

In conclusion, the aim of this study is to systematically analyse how the UHI effect has been addressed in the scientific literature, to identify key research themes and their temporal evolution, and to critically highlight knowledge gaps in order to provide guidance for future research and urban planning policies. Based on this critical review, the study also provides a roadmap for future research and policy recommendations aimed at mitigating the UHI effect. The findings will contribute not only to academic studies but also to the formulation of more sustainable urban policies by providing practical recommendations for urban planners, policy makers and local governments. In this direction, the fundamental principles of the research, the literature review, the methods employed and the analyses conducted are discussed in detail in the following sections.

2. Research Background and Related Work

The UHI effect is widely acknowledged within academic discourse as a major consequence of urbanisation on ecological systems. In the closing decades of the twentieth century, the prevailing scientific consensus was that urban areas experienced higher temperatures in comparison with rural regions. Consequently, scientific studies commenced in order to ascertain the reasons for this discrepancy [15,16]. The scientific foundations of the UHI effect have become increasingly quantitative and model-oriented, particularly following Myrup’s seminal 1969 study, entitled ‘A Numerical Model of the Urban Heat Island’. In this period, studies concentrated on ascertaining the correlation between UHI and factors such as urbanisation, surface materials and energy consumption. However, it was not until the 2010s that the impact of big data analytics and artificial intelligence-based modelling methods in this field became widely apparent. Recent advancements in geographic information systems (GIS) and remote sensing techniques, in conjunction with text mining and topic modelling methodologies, have facilitated comprehensive data analysis within the domain of UHI research [13]. These developments have enabled UHI studies to be approached in a more systematic and interdisciplinary manner [17]. Bibliometric analyses of the extant literature demonstrate that studies on UHI are primarily categorised into four distinct classifications: root causes, environmental and social impacts, mitigation strategies, and measurement and modelling techniques.

The replacement of natural surfaces with materials such as concrete and asphalt, which has been a consequence of urbanisation, increased energy consumption and anthropogenic heat emissions from fossil fuels, is among the most significant causes of the UHI effect [18]. Moreover, the restriction of airflow by high-rise buildings and the reduction in green areas also contribute to the increase in temperature. The UHI effect has been demonstrated to engender a multitude of deleterious consequences, including but not limited to: air pollution, heat-related diseases, increased energy consumption and loss of thermal comfort [19]. In the context of extreme heat waves, elevated temperatures in urban areas have been demonstrated to exert a deleterious effect on public health, as evidenced by increased mortality rates and the imposition of a considerable strain on public health systems [20]. A plethora of methodologies have been employed to mitigate the impact of UHI, including the implementation of green infrastructure practices, the utilisation of reflective materials, the integration of cool roofs and cool asphalt technologies [21,22,23]. Nevertheless, the efficacy of these strategies is contingent upon the climatic and geographical characteristics of the respective cities. In the field of UHI research, a wide range of modelling techniques have been employed, drawing upon diverse scientific disciplines. These include remote sensing, Geographic Information Systems (GIS), numerical simulations and machine learning. In recent years, topic modelling methods have begun to assume a significant role in these analyses.

The UHI effect is known to cause both an increase in physical temperature and a change in the perception of urban environments. In their 2023 study, Hu and Chen conducted a comprehensive analysis of the identity, structure and meaning of coastal cities. The researchers demonstrated the intricate interplay between environmental factors and urban tourism, as well as urban planning [24]. Furthermore, the repercussions of UHI on tourist cities necessitate additional research in relation to city image and sustainability policies.

Topic modelling is a powerful data analysis method that allows the analysis of large text datasets to identify main themes and examine temporal changes in these themes. In the context of UHI research, the application of text mining techniques to the analysis of scientific publications has proven to be a fruitful endeavour. In this context, classical topic modelling methods such as Latent Dirichlet Allocation (LDA) have been utilised for a considerable duration to thematically map studies on the UHI effect [25]. Nevertheless, the fact that LDA is based on static analyses and has limited flexibility means that it is not suitable for addressing the dynamic aspects of UHI.

In recent years, there has been an increasing use of more advanced topic modelling methods, such as BERTopic, in the field of UHI studies [25,26,27,28,29,30]. BERTopic facilitates the provision of more flexible and contextually meaningful clustering of topics by employing the HDBSCAN clustering algorithm and word embedding models. This methodological approach facilitates a dynamic analysis of the temporal changes in themes in the literature in UHI research.

The issue of UHI has implications for a wide range of disciplines, including, but not limited to, urban planning, energy policy and public health. The BERTopic initiative promotes interdisciplinary collaboration by facilitating enhanced visualisation of the relationships between these fields [31,32,33,34,35,36]. This is particularly noteworthy in light of the extensive corpus of UHI studies. BERTopic conducts a thorough analysis of the available data and identifies research gaps [22,37,38,39,40,41]. The analysis of temporal trends is facilitated by BERTopic, which demonstrates the periods during which specific themes attain significance within the domain of UHI research. Furthermore, it identifies subjects that merit further investigation in the future.

The utilisation of topic modelling techniques is not confined to the analysis of academic articles; these techniques are also employed extensively to comprehend the information and discourse disseminated on social media platforms [42,43,44,45]. In their seminal 2019 study, Smith and Graham analysed anti-vaccination movements on Facebook, examining the impact of social media discourses on public health policies [46]. In a similar manner, techniques such as BERTopic can be utilised to analyse the manner in which UHI is addressed on social media and online news platforms.

Notwithstanding the extant research, studies on UHI exhibit certain lacunae and there is room for improvement in future research. There exists a paucity of research into UHI in developing countries. The principal reasons for the dearth of UHI research in regions such as Africa and South Asia are the absence of data, insufficient financial resources and the Western-centred nature of academic research [13]. This complicates the interpretation of regional variations. The paucity of research into the interaction of climate change and UHI is evidenced by the use of static analyses to address the UHI effect. It is recommended that future studies incorporate a greater number of time series analyses and machine learning-based modelling techniques with the objective of determining the long-term impacts of climate change on UHI [47]. It is imperative that further investigation is conducted into the impact of socioeconomic inequalities on UHI. In this context, BERTopic-based analyses should be utilised to ascertain how the UHI effect varies in low-income neighbourhoods [2,48].

3. Materials and Methods

The purpose of this study is to identify themes in the literature regarding the UHI effect and to analyse temporal trends. In this direction, a large-scale academic article dataset was analysed by topic modelling method, thereby systematically revealing the trends in existing research. (Figure 1). To achieve this aim, the BERTopic model, a Natural Language Processing (NLP) technique, was employed to cluster the UHI literature and analyse temporal variations in research themes. BERTopic processes text-based data to automatically extract meaningful topics and analyses their changes over time. In this section, the data collection, pre-processing, modelling and result analysis stages are discussed in detail.

3.1. Data Collection and Preprocessing

The data collection process was executed through the utilisation of the Web of Science (WoS) database. The selection of WoS was predicated on the fact that it contains peer-reviewed academic publications and provides access to high quality interdisciplinary research. In the data collection phase, a broad literature review on UHI was conducted. In this context, key terms such as ‘urban heat island’, ‘urban warming’, ‘urban microclimate’, ‘city heat island’, ‘urban thermal environment’, ‘surface urban heat island (SUHI)’, ‘urban albedo’, ‘heat island mitigation’ and ‘sky view factor’ were identified and scanned in title, abstract and keyword sections. The present study exclusively encompasses articles published between 2004 and 2024 in the English language. These limitations are designed to facilitate a comprehensive understanding of prevailing trends in UHI research, whilst ensuring that the study is founded exclusively upon international academic literature. In accordance with the predetermined criteria, a total of 9061 articles were collated and prepared for the analysis process.

In the course of the data preprocessing procedure, a range of natural language processing (NLP) techniques were employed to render the raw texts suitable for modelling. Initially, the article texts were subjected to a process of purification, whereby superfluous numerals, punctuation marks, non-essential symbols, and stopwords were eliminated. In this study, the term “non-essential symbols” specifically refers to digits (0–9), punctuation marks (.,!?;:), special characters (#, %, @, &, *), and HTML/XML tags. These elements were removed using regular expression–based filtering, and multiple consecutive spaces were collapsed into a single space. The rationale behind this step was to eliminate noise and improve semantic clarity of the text.

Subsequently, stemming and lemmatisation techniques were employed to preserve the standard forms of the words and ensure semantic unity. In the pre-processing stage, the texts were transformed into a document-term matrix (DTM) in order to facilitate the effective operation of the BERTopic model. This transformation facilitated the determination of the frequencies of the words in the texts and the extraction of themes. Furthermore, the TF-IDF (Term Frequency—Inverse Document Frequency) method was employed to ascertain the importance of keywords and optimise word distribution [49,50,51]. Subsequent to the pre-processing stage, the BERTopic model was implemented in order to ascertain the themes associated with UHI. The model employed deep learning-based natural language processing methods to interpret the extensive dataset and elucidate the interrelationships between subjects. The results obtained in this process have provided a critical basis for identifying scientific trends, knowledge gaps and solutions for temperature increase in cities. The modelling process and analysis method are visualised in Figure 1.

3.2. Topic Modelling with BERTopic Model

In the present study, the BERTopic model was utilised to identify themes in the extant literature and to analyse their temporal changes. BERTopic is a topic modelling method that combines NLP techniques and machine learning algorithms to extract meaningful themes from text data [30,52,53]. The model calculated word weights with TF-IDF (Term Frequency-Inverse Document Frequency) and performed dimension reduction with the UMAP (Uniform Manifold Approximation and Projection) algorithm. Subsequently, thematic clustering was performed utilising the HDBSCAN (Hierarchical Density-Based Spatial Clustering of Applications with Noise) algorithm. The BERTopic study not only identified themes in the literature but also analysed the temporal changes in these themes. The temporal patterns of the identified themes were characterised by their intensity and trend, and these were visualised with regard to the publication years of the articles. This process clearly highlights the dominant topics in the existing literature, while also emphasising areas that have been studied to a limited extent.

BERTopic is a model that identifies topics by converting texts into semantic vectors and applying density-based clustering. Initially, texts are embedded into a high-dimensional vector space using a pre-trained language model.

$$V_i=f\left(T_i\right)$$

(1)

Here, $T_i$ represents the i-th text, and $f$ is a transformation function such as BERT or SBERT. Consequently, $V_i$ is a vector in ${\mathbb{R}}^d$. Next, a density-based clustering algorithm, HDBSCAN, clusters the obtained vectors (1).

$$C=HDBSCAN(V,\epsilon , {min}_{samples})$$

(2)

Here, $C$ denotes the cluster labels, while $\epsilon$ and ${min}_{samples}$ represent clustering parameters. After clustering, the Class-Based TF-IDF (c-TF-IDF) method is applied to select the most distinctive words for each cluster (2).

$${c-TF-IDF}_{w,k}={\displaystyle \frac{f_{w,k}}{{\sum }_{w\prime ϵk}{f}_{w^{\prime },k}}}\times \mathit{log}\left({\displaystyle \frac{N}{n_w}}\right)$$

(3)

Here, $f_{w,k}$ represents the frequency of word $w$ in cluster k, N is the total number of documents, and $n_w$ is the number of documents containing word $w$. For cluster representation, high-dimensional vectors can be reduced using UMAP (3).

$$V^{\prime }=UMAP (V, d_{target})$$

(4)

Finally, words with the highest c-TF-IDF values in each cluster are assigned as topic labels. This way, BERTopic identifies meaningful topics while clustering texts (4).

3.3. Model Configuration and Parameterization

To ensure reproducibility and methodological transparency, the full configuration of preprocessing, dimensionality reduction, clustering, and topic modelling components used in the BERTopic pipeline is described (

Table 1. Preprocessing Steps and Model Parameters of the BERTopic Workflow.

Step	Parameter/Procedure	Value/Description
Preprocessing	Lowercasing	All text converted to lowercase
	Stopwords	NLTK English stopwords + custom list
	Lemmatization	WordNetLemmatizer (nltk)
	Normalization	Dictionary mapping
	Cleaning rules	Remove digits, punctuation, special characters; collapse multiple spaces
Vectorization	Model	TF-IDF (max_features = 5000)
	Analyzer	Word-level
Dimensionality Reduction (UMAP)	n_neighbors	Best param (selected via grid search: 5, 10, 15, 20, 30, 40, 50)
	min_dist	Grid search (0.0–0.99)
	n_components	3
	random_state	42
Clustering (HDBSCAN)	min_cluster_size	Optimised from [10,15,20,25]
	min_samples	Optimised from [5,10,15,20]
	prediction_data	True
Topic Modelling (BERTopic)	Embedding model	TF-IDF vectorizer (no transformer embedding used)
	top_n_words	15
	min_topic_size	10
	low_memory	False
	Vectorizer model	TF-IDF (above)

). Titles, abstracts, and author keywords were merged into a single text field and converted to lowercase. Standard English stopwords from NLTK were extended with a custom list such as writing, article, abstract, available, research, study, method. Lemmatization was performed with the NLTK WordNetLemmatizer, and normalisation rules were applied to unify variants. Digits, punctuation, and special characters were removed, and multiple spaces were collapsed into one. The texts were then transformed into numerical vectors using TF-IDF with a maximum of 5000 features at the unigram level, which provided a balance between vocabulary coverage and computational efficiency.

Dimensionality reduction was carried out with UMAP, reducing vectors to three dimensions. A grid search tested different neighbour values (5, 10, 15, 20, 30, 40, 50) and distance thresholds (0.0, 0.1, 0.2, 0.3, 0.5, 0.8, 0.99). The configuration that maximised silhouette scores and minimised Davies–Bouldin indices was selected, resulting in 15 neighbours and a minimum distance of 0.10, with three components and random state fixed at 42. Clustering was performed with HDBSCAN, where min_cluster_size values of 10, 15, 20, and 25 and min_samples values of 5, 10, 15, and 20 were tested. The best performance was obtained with min_cluster_size set to 20 and min_samples set to 10, with prediction data enabled to estimate topic probabilities.

Finally, BERTopic was trained with the optimised UMAP and HDBSCAN settings and the TF-IDF vectorizer as the embedding model. The parameters were top_n_words = 15, min_topic_size = 10, and low_memory = False. This configuration yielded coherent and interpretable topics while avoiding fragmentation into overly small clusters. The explicit description of preprocessing steps, parameter ranges, and final settings ensures full replicability of the modelling procedure and directly addresses reproducibility concerns in bibliometric and topic modelling research.

4. Results

The results of this study on the UHI effect are presented under four main headings: bibliometric analysis, topic modelling results, temporal trends and classification of themes. The purpose of this structure is to reveal the main themes, trends and knowledge gaps in the literature.

4.1. Bibliometric Analysis (RQ1)

A comprehensive data set consisting of 9061 articles was analysed for the purpose of conducting a bibliometric analysis of the literature. The distribution of the articles was as follows: 36.99% were in environmental sciences, 24.21% in meteorological and atmospheric sciences, and 18.23% in construction technology (Figure 2).

This distribution illustrates the interdisciplinary nature of the UHI effect and its significance across various scientific disciplines.

Furthermore, an analysis of the journals with the highest number of publications on UHI revealed that the most prolific were ‘Urban Climate’ (6.42%), ‘Sustainable Cities and Society’ (5.69%) and ‘Building and Environment’ (5.41%), respectively. These are followed by ‘Sustainability’ (4.2%), ‘Remote Sensing’ (3.23%) and ‘Energy and Buildings’ (2.95%). Interdisciplinary environmental science journals such as ‘Science of The Total Environment’ (2.65%) and ‘Atmosphere’ (2.09%) are also among the important publication platforms (Figure 3).

A thorough investigation into keyword usage revealed that the terms ‘urban heat island’ (7.3%), ‘land surface temperature’ (2.35%) and ‘climate change’ (1.12%) were the most prevalent (Figure 4). These findings indicate that the extant literature predominantly focuses on the consequences of rising temperatures and climate change in urban areas.

4.2. Topic Modelling Results (RQ1)

Pursuant to the analysis undertaken with the BERTopic model, the primary themes (

Table A1. Discovered topics, keywords and percentages.

Topic Name	BERTopic Keywords	%
Pavement Materials	pavement, asphalt, concrete, material, property, pervious, mixture, coating, performance, permeable, strength, aggregate, temperature, heat, cool	16.39
Urban Vegetation	tree, specie, leaf, canopy, cooling, planting, street, crown, transpiration, urban, effect, shade, growth, temperature, thermal	9.41
Air Pollution	concentration, pollution, pm, air, pollutant, ozone, aerosol, quality, atmospheric, urban, model, mu, emission, layer, island	7.03
Rainfall Patterns	rainfall, precipitation, urbanization, event, lightning, extreme, convective, thunderstorm, storm, urban, area, region, beijing, convergence, cloud	6.76
Habitat Diversity	specie, habitat, tolerance, bee, ant, urbanization, insect, size, response, trait, abundance, population, body, urban, community	6.16
LST Resolution	lst, resolution, algorithm, image, downscaling, retrieval, data, surface, land, infrared, spatial, landsat, tir, modis, lsts	4.24
Urban Cooling	park, cooling, pci, effect, urban, intensity, pce, surrounding, factor, island, landscape, area, efficiency, planning, size	3.88
Brazilian Climate	brazil, paulo, city, sao, island, urban, brazilian, area, tropical, heat, surface, degree, climate, temperature, region	2.79
Reflective Materials	material, solar, reflectance, retroreflective, facade, building, envelope, rr, film, energy, property, radiation, incident, reflective, optical	2.21
Sky View Factor	svf, view, sky, factor, fisheye, image, skyview, solar, diffuse, street, model, camera, photograph, panorama, hemispherical	2.12
Ventilation Corridor	ventilation, corridor, wind, urban, precinct, circuit, environment, potential, planning, area, city, air, assessment, path, based	2.07
Ecosystem Infrastructure	ecosystem, infrastructure, service, gi, green, benefit, planning, space, management, naturebased, cemetery, stormwater, resilience, urban, quality	2.07
Soil Respiration	soil, carbon, forest, tree, ecosystem, service, respiration, litter, management, biomass, urban, decomposition, sequestration, planting, community	1.93
Climate Adaptation	climate, adaptation, change, policy, mitigation, strategy, planning, urban, knowledge, challenge, city, action, measure, plan, implementation	1.85
Groundwater	groundwater, subsurface, geothermal, shallow, aquifer, temperature, heat, ground, flow, anthropogenic, energy, potential, underground, heating, source	1.85
Monitoring Sensor	monitoring, sensor, mobile, wearable, microclimate, environmental, network, data, measurement, intraurban, fixed, air, temperature, station, smartphone	1.71
Satellite Remote	satellite, remote, sensing, image, data, gnss, surface, temperature, sensor, island, uhi, suhi, urban, heat, measurement	1.66
Radiative Cooling	radiative, cooling, material, passive, coating, rc, subambient, film, daytime, drc, paint, solar, polymer, photonic, potential	1.54
Snow Irradiance	terrain, snow, radiation, topography, sky, topographic, view, solar, glacier, forest, slope, parameterization, irradiance, snowmelt, flux	1.54
Plant Diversity	specie, plant, alien, native, trait, fern, habitat, diversity, forest, de, gradient, fallopia, woody, richness, community	1.35
Solar Energy	pv, photovoltaic, panel, solar, energy, system, deployment, photovoltaics, pvsps, building, bipv, rooftop, impact, rpvps, thermal	1.3
Color Pigments	pigment, reflectance, coating, nir, tile, solar, reflective, ceramic, cool, property, glaze, color, nearinfrared, copper, inkjet	1.25
Water Infrastructure	water, stormwater, infrastructure, wastewater, rainwater, service, green, br, gsi, ecosystem, management, solution, nb, runoff, naturebased	1.25
Drone Mapping	uav, aerial, unmanned, vehicle, thermal, image, uavs, camera, infrared, environment, surface, temperature, tir, accuracy, resolution	1.23
Lidar Mapping	visualization, lidar, relief, archaeological, dem, landslide, digital, feature, technique, skyview, elevation, openness, terrain, archaeology, visualisation	1.23
Local Climate Zone	lcz, zone, local, climate, classification, suhi, mapping, study, urban, scheme, lst, wudapt, surface, class, land	1.13
Mosquito Disease	mosquito, disease, dengue, transmission, vector, abundance, malaria, ae, outbreak, mosquitoborne, aedes, aegypti, albopictus, vivax, culex	1.08
Radiant Temperature	tmrt, road, radiant, radiation, rst, longwave, forecast, view, mean, flux, thermal, temperature, svf, data, factor	1.04
Phenology	phenology, so, vegetation, urbanization, spring, response, warming, eos, phenological, growing, change, plant, date, start, gud	0.99
Circulation Velocity	circulation, velocity, convective, boundary, layer, flow, heat, wind, turbulent, mesoscale, flux, stratified, island, uhic, convection	0.99
Flowering Phenology	flowering, phenology, phenological, pollen, date, plant, ffd, specie, onset, change, phenophases, day, response, temperature, cherry	0.94
LCZ Classification	lcz, classification, sentinel, mapping, zone, convolutional, accuracy, feature, local, network, classifier, cnn, neural, deep, learning	0.94
Pandemic Pollution	lockdown, covid, pandemic, activity, restriction, air, quality, anthropogenic, human, reduced, period, prelockdown, aod, suhi, pollution	0.94
Cooling Corridor	corridor, network, connectivity, node, source, ecological, island, patch, cold, theory, circuit, pattern, landscape, resistance, cooling	0.94
Fog Impact	fog, hole, visibility, dense, pdo, occurrence, aerosol, frequency, california, pacific, angeles, los, pollution, decrease, particulate	0.89
Solar Radiation	solar, radiation, canyon, street, sun, model, geometry, view, urban, availability, building, incident, morphology, simplified, duration	0.87
Neural Network	neural, network, artificial, model, indoor, ann, prediction, building, london, predictive, hourly, temperature, air, island, predict	0.82
Japan Climate	japan, trend, precipitation, air, temperature, warming, station, season, kumagaya, rural, udi, seasonal, area, nlni, site	0.77
Solar Irradiance	solar, irradiance, facade, sky, radiation, shadow, diffuse, potential, rooftop, cadaster, building, view, model, dsm, pv	0.72
Australian Climate	melbourne, adelaide, sydney, australia, cbd, heat, australian, uhi, island, heatwaves, urban, day, station, anthropogenic, city	0.72
Wind Flow	flow, canyon, vortex, wind, street, buoyancy, turbulence, turbulent, heating, removal, budget, statistic, direction, particle, buoyant	0.72
Sustainable Farming	food, agriculture, garden, sustainability, farming, farm, urban, greenhouse, security, production, horticulture, environmental, ua, sustainable, challenge	0.67

) that emerged in UHI research were identified (Figure 5). The study identified the following topics as the most significant: ‘Pavement Materials’ (16.39%), which focuses on the thermal properties of asphalt, concrete and permeable surface materials, insulation performance and cooling coatings. Other significant issues pertaining to building materials include ‘Urban Vegetation’ (9.41%) and ‘Air Pollution’ (7.03%). The role of urban afforestation and vegetation in reducing UHI is addressed, along with its effects on ecological diversity.

Air pollution is a significant topic in the context of the effects of ozone and particulate matter emissions on UHI. The study ‘Rainfall Patterns’ (6.76%), which examines the relationship between rainfall patterns and atmospheric processes related to the UHI, explores the impact of urbanisation on rainfall regimes. The impacts of urbanisation on ecosystems and biodiversity were evaluated under the heading ‘Habitat Diversity’ (6.16%).

The LST Resolution (4.24%) programme is dedicated to the measurement and mapping of surface temperatures through the utilisation of remote sensing methodologies. Research on urban cooling strategies is gathered in ‘Urban Cooling’ (3.88%), ‘Reflective Materials’ (2.21%) and ‘Radiative Cooling’ (1.54%). It is evident that a significant proportion of the methods employed in order to reduce the heat load of urban areas include the use of reflective surfaces and passive radiative cooling techniques. In the context of urban planning approaches aimed at mitigating the impact of UHI, the topics of ‘Ventilation Corridor’ (2.07%), ‘Cooling Corridor’ (0.94%), and ‘Wind Flow’ (0.72%) encompass studies focused on optimising urban airflow and preventing the accumulation of hot air. The following subjects are of particular interest in terms of regional analyses: ‘Brazilian Climate’ (2.79%), ‘Japanese Climate’ (0.77%) and ‘Australian Climate’ (0.72%). The impact of UHI in different climate zones has been evaluated under these headings, and the manner in which urban planning approaches are shaped according to regional differences has been analysed. The concept of ‘Climate Adaptation’ (1.85%) is centred on the policies and practices that have been developed to combat UHI in the context of climate change.

Remote sensing and modelling techniques have an important place in the study of urban heat islands (UHI). The focus on satellite remote sensing (1.66%) and drone mapping (1.23%) is on mapping using satellite imagery and unmanned aerial vehicles (UAVs), whilst lidar mapping (1.23%) includes methods used for three-dimensional urban analysis. The term ‘Neural Network’ (0.82%) is a broad category encompassing research studies that employ machine learning algorithms and artificial neural networks to model Urban Heat Island (UHI) effects. Notable public health concerns encompass ‘Mosquito Disease’ (1.08%), ‘Pandemic Pollution’ (0.94%), and ‘Fog Impact’ (0.89%). In the course of examining the relationship between climate change and rising global temperatures on the one hand, and mosquito-borne diseases on the other, the links between atmospheric events such as changes in air pollution and fog formation during the period of the pandemic of the Coronavirus (SARS-CoV-2) were also investigated in the studies examined.

4.3. Temporal Subject Trends (RQ2)

This study analyses temporal trends in research on the Urban Heat Island (UHI) effect, drawing upon data from the most recent five-year period (2020–2024) (Figure 6). The primary rationale for this phenomenon is the confluence of three major factors: the rapid advancements in urbanism, climate change, and artificial intelligence-based analysis methodologies, which have exerted a direct influence on the realm of UHI research. Moreover, the emergence of novel concepts related to UHI in the academic literature, coupled with shifts in the perceived significance of existing topics, necessitated the identification of contemporary trends. The increasing salience of the impact of global temperature increases on cities, and the emergence of sustainable urban planning approaches, are further factors that support the observed temporal pattern. A close analysis of the most recent five-year period reveals that specific subjects within the domain of UHI research have garnered increased academic interest. This shift in focus is exemplified by the subjects of ‘Pavement Materials’, ‘Urban Cooling’ (urban cooling solutions), ‘Drone Mapping’ and ‘Neural Networks’. Accordingly, the analysis employing the colour coding in Appendix A,

Table A2. Annual percentages and trends of topics under UHI.

Topic Name	2004	2005	2006	2007	2008	2009	2010	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020	2021	2022	2023	2024	Trend
Air Pollution	10.00%	14.29%	4.55%	4.55%	21.74%	6.67%	14.29%	4.35%	8.93%	14.55%	7.94%	9.09%	6.06%	8.04%	6.82%	7.55%	9.63%	3.66%	7.03%	5.12%	5.56%	0.031
Australian Climate	0.00%	0.00%	0.00%	4.55%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	6.06%	3.57%	0.00%	1.26%	0.53%	0.41%	0.00%	0.39%	0.00%	0.000
Brazilian Climate	0.00%	0.00%	0.00%	4.55%	0.00%	0.00%	7.14%	4.35%	5.36%	5.45%	1.59%	1.30%	3.03%	3.57%	4.55%	3.14%	2.67%	2.85%	3.13%	1.18%	2.63%	−0.313
Circulation Velocity	0.00%	7.14%	0.00%	4.55%	4.35%	3.33%	3.57%	0.00%	0.00%	1.82%	3.17%	1.30%	5.05%	0.00%	0.76%	1.26%	0.53%	0.41%	0.00%	0.39%	0.58%	−1.000
Climate Adaptation	0.00%	0.00%	0.00%	4.55%	0.00%	3.33%	0.00%	0.00%	0.00%	1.82%	1.59%	1.30%	2.02%	2.68%	2.27%	3.77%	3.21%	3.25%	0.39%	0.79%	1.46%	−0.828
Color Pigments	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	1.82%	3.17%	3.90%	1.01%	1.79%	0.76%	1.26%	0.00%	0.00%	1.17%	2.36%	2.05%	0.547
Cooling Corridor	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.41%	1.56%	2.36%	2.92%	0.000
Drone Mapping	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	1.59%	0.00%	0.00%	0.89%	0.00%	0.63%	0.53%	0.81%	1.95%	1.97%	2.63%	0.000
Ecosystem Infrastructure	0.00%	0.00%	4.55%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	1.82%	0.00%	0.00%	1.01%	2.68%	1.52%	1.89%	2.67%	3.25%	2.34%	1.57%	3.51%	0.547
Flowering Phenology	0.00%	0.00%	9.09%	4.55%	0.00%	6.67%	0.00%	8.70%	1.79%	3.64%	0.00%	2.60%	1.01%	0.00%	0.00%	1.89%	0.00%	0.00%	0.78%	0.39%	0.58%	0.000
Fog Impact	0.00%	7.14%	0.00%	4.55%	13.04%	0.00%	0.00%	0.00%	3.57%	0.00%	0.00%	2.60%	0.00%	0.00%	4.55%	0.63%	1.07%	0.00%	0.00%	0.00%	0.58%	−1.000
Groundwater	20.00%	0.00%	0.00%	0.00%	0.00%	6.67%	3.57%	0.00%	1.79%	3.64%	1.59%	2.60%	0.00%	5.36%	0.76%	1.89%	3.21%	2.85%	1.56%	0.79%	0.29%	1.063
Habitat Diversity	0.00%	14.29%	0.00%	0.00%	4.35%	6.67%	3.57%	0.00%	3.57%	5.45%	6.35%	6.49%	3.03%	7.14%	11.36%	11.95%	6.42%	6.50%	3.52%	5.91%	5.85%	−0.691
Japan Climate	0.00%	0.00%	9.09%	4.55%	8.70%	6.67%	0.00%	8.70%	3.57%	1.82%	1.59%	0.00%	0.00%	0.89%	0.76%	0.00%	0.00%	0.00%	0.39%	0.00%	0.29%	−0.484
LCZ Classification	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.76%	1.89%	1.60%	0.81%	2.34%	1.97%	0.29%	2.094
Lidar Mapping	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	8.70%	3.57%	1.82%	0.00%	0.00%	3.03%	4.46%	0.00%	0.63%	1.60%	1.63%	0.78%	1.18%	0.29%	0.000
Local Climate Zone	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	1.59%	0.00%	0.00%	0.00%	1.52%	3.14%	1.07%	1.63%	1.56%	0.00%	2.05%	0.031
LST Resolution	0.00%	0.00%	4.55%	0.00%	0.00%	6.67%	0.00%	4.35%	7.14%	5.45%	4.76%	6.49%	5.05%	3.57%	2.27%	5.03%	3.21%	4.07%	5.47%	3.54%	4.68%	1.406
Monitoring Sensor	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	7.14%	0.00%	0.00%	3.64%	1.59%	2.60%	0.00%	0.00%	3.03%	1.26%	2.67%	2.85%	1.95%	1.57%	1.17%	−0.355
Mosquito Disease	0.00%	0.00%	0.00%	0.00%	4.35%	0.00%	0.00%	0.00%	0.00%	0.00%	3.17%	1.30%	4.04%	0.89%	0.76%	0.63%	1.60%	1.22%	1.17%	1.18%	0.29%	0.547
Neural Network	10.00%	0.00%	0.00%	4.55%	0.00%	3.33%	3.57%	8.70%	1.79%	0.00%	1.59%	1.30%	1.01%	0.00%	0.76%	0.00%	0.53%	0.41%	1.17%	0.39%	0.29%	0.547
Pandemic Pollution	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	3.25%	1.95%	1.97%	0.88%	0.000
Pavement Materials	10.00%	0.00%	4.55%	13.64%	8.70%	6.67%	3.57%	13.04%	16.07%	10.91%	14.29%	19.48%	15.15%	11.61%	15.91%	19.50%	16.04%	19.51%	21.48%	19.69%	14.33%	0.350
Phenology	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	3.03%	0.00%	0.76%	0.00%	1.60%	2.03%	0.78%	1.57%	1.17%	0.031
Plant Diversity	0.00%	7.14%	0.00%	4.55%	0.00%	6.67%	0.00%	0.00%	0.00%	1.82%	4.76%	1.30%	1.01%	0.89%	1.52%	1.89%	1.07%	0.81%	1.17%	1.18%	1.17%	−0.227
Radiant Temperature	10.00%	7.14%	4.55%	0.00%	4.35%	0.00%	3.57%	0.00%	0.00%	0.00%	3.17%	0.00%	2.02%	0.00%	0.76%	1.26%	0.00%	0.81%	1.95%	1.18%	0.29%	1.578
Radiative Cooling	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	1.79%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	1.26%	2.67%	0.41%	0.39%	4.33%	3.22%	0.000
Rainfall Patterns	10.00%	7.14%	13.64%	18.18%	8.70%	16.67%	10.71%	4.35%	1.79%	10.91%	4.76%	7.79%	2.02%	8.93%	6.06%	6.92%	5.35%	5.69%	6.25%	7.48%	7.02%	0.031
Reflective Materials	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	1.79%	0.00%	3.17%	3.90%	4.04%	6.25%	3.03%	1.26%	1.07%	2.85%	1.95%	1.97%	2.05%	−0.355
Satellite Remote	0.00%	0.00%	9.09%	9.09%	0.00%	10.00%	3.57%	4.35%	8.93%	1.82%	3.17%	2.60%	2.02%	1.79%	0.00%	1.26%	0.00%	1.22%	1.95%	0.39%	0.88%	0.000
Sky View Factor	10.00%	0.00%	4.55%	4.55%	0.00%	0.00%	7.14%	4.35%	1.79%	1.82%	6.35%	3.90%	2.02%	4.46%	4.55%	1.26%	2.14%	2.03%	0.78%	0.79%	1.17%	−0.828
Snow Irradiance	10.00%	7.14%	13.64%	0.00%	8.70%	0.00%	0.00%	4.35%	12.50%	3.64%	1.59%	0.00%	2.02%	2.68%	2.27%	0.00%	1.07%	0.41%	0.39%	0.79%	0.58%	−0.828
Soil Respiration	0.00%	14.29%	0.00%	0.00%	4.35%	3.33%	0.00%	0.00%	0.00%	0.00%	4.76%	0.00%	4.04%	2.68%	0.76%	2.52%	1.07%	2.03%	1.56%	0.79%	2.92%	1.063
Solar Energy	0.00%	0.00%	9.09%	4.55%	0.00%	0.00%	0.00%	0.00%	0.00%	3.64%	1.59%	1.30%	1.01%	1.79%	0.00%	0.00%	2.14%	0.00%	0.78%	2.76%	1.75%	0.000
Solar Irradiance	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	1.82%	0.00%	0.00%	0.00%	0.00%	2.27%	0.63%	1.60%	0.81%	0.39%	0.79%	0.88%	−0.828
Solar Radiation	0.00%	0.00%	4.55%	0.00%	0.00%	0.00%	0.00%	4.35%	0.00%	0.00%	0.00%	2.60%	2.02%	0.00%	0.00%	1.26%	2.14%	0.81%	0.78%	0.00%	0.88%	0.000
Sustainable Farming	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	1.79%	0.00%	1.59%	0.00%	2.02%	0.89%	0.76%	0.00%	0.00%	0.41%	0.78%	0.39%	1.46%	0.031
Urban Cooling	0.00%	0.00%	0.00%	4.55%	0.00%	0.00%	7.14%	0.00%	0.00%	1.82%	3.17%	3.90%	4.04%	1.79%	3.03%	1.26%	3.21%	4.07%	4.69%	4.72%	7.02%	0.547
Urban Vegetation	10.00%	7.14%	0.00%	0.00%	4.35%	0.00%	14.29%	8.70%	8.93%	3.64%	3.17%	7.79%	11.11%	8.93%	11.36%	7.55%	11.23%	11.38%	7.42%	11.81%	10.53%	−0.347
Ventilation Corridor	0.00%	7.14%	4.55%	0.00%	4.35%	6.67%	3.57%	4.35%	3.57%	0.00%	0.00%	1.30%	1.01%	0.89%	0.76%	0.00%	2.14%	1.63%	3.52%	3.54%	2.05%	3.641
Water Infrastructure	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	4.35%	0.00%	3.64%	1.59%	0.00%	0.00%	0.89%	0.76%	1.89%	2.14%	1.63%	1.95%	0.79%	1.17%	1.578
Wind Flow	0.00%	0.00%	0.00%	0.00%	0.00%	0.00%	3.57%	0.00%	0.00%	1.82%	1.59%	1.30%	0.00%	0.00%	2.27%	0.63%	0.53%	1.22%	0.78%	0.00%	0.58%	−0.656

facilitates comprehension of the evolution of specific subjects in research on UHI over time. The utilisation of white tones is indicative of subjects with minimal academic interest, subjects with escalating academic interest as they transition towards orange, and grey tones signify subjects of interest during a specific period. This distribution facilitates a more comprehensive understanding of the themes that have garnered greater attention in the academic literature and their evolution over time.

It is evident that a decrease in academic interest in specific subjects has been observed in recent years. Among these subjects, those that have seen a decline include ‘Air Pollution’, ‘Fog Impact’, ‘Neural Network’, ‘Sky View Factor’, ‘Snow Irradiance’, ‘Water Infrastructure’, ‘Solar Irradiance’, ‘Climate Adaptation’, ‘Monitoring Sensor’ and ‘Solar Radiation’.

A decline in interest has been observed, particularly in the areas of solar radiation (−0.59%), monitoring sensors (−0.56%) and climate adaptation (−0.54%). Despite the fact that studies on solar radiation and irradiation have been the focus of considerable research in previous years, particularly in the context of the relationship between energy systems and urbanisation, there has been an increased interest in more application-oriented topics such as urban cooling methods and remote sensing solutions in recent years.

Figure 6 shows both increases (e.g., urban cooling) and relative decreases (e.g., climate adaptation) between 2020 and 2024. However, this does not imply that recent studies have neglected climate adaptation research or that all studies have been conducted under the concept of urban cooling. A review of the literature shows that in recent studies, many “cooling” measures, such as tree cover, shading, and green/reflective roofs, have been framed under the heading of urban cooling as components of urban heat island adaptation [2,4,5,17,54]. Therefore, it should be noted that part of this observed change may be an example of terminological evolution rather than a genuine conceptual break.

4.4. Classification of Themes (RQ3)

The BERTopic analysis shows that seven key areas are prominent in UHI research (Figure 7). The study’s most significant finding, as reported in ‘Plants & Ecosystem’ (22.85%), concerns the impact of urban green spaces and ecosystem services on UHI. The subject of ‘Urban Vegetation’ (9.41%) is the examination of planting strategies, while ‘Habitat Diversity’ (6.16%) and ‘Ecosystem Infrastructure’ (2.07%) focus on the role of biodiversity in urban environments.

‘Urban Climate’ (20.5%), which deals with urban temperature dynamics, focuses on remote sensing and urban cooling solutions. ‘LST Resolution’ (4.24%) discusses the methods used to determine surface temperatures, while ‘Urban Cooling’ (3.88%) includes cooling urban planning strategies. The subject of air flow and shading is addressed in the sections titled ‘Sky View Factor’ (2.12%) and ‘Ventilation Corridor’ (2.07%). The subject of ‘Materials & Surfaces’ (19.85%) is the examination of how building materials shape the heat island effect, while ‘Pavement Materials’ (16.39%) focuses on the thermal properties of asphalt and concrete surfaces. The categories identified as significant contributors to this phenomenon include ‘Reflective Materials’ (2.21%) and ‘Colour Pigments’ (1.25%), with research focusing on the development of surface treatments that enhance the reflectance of sunlight, thereby leading to a reduction in surface temperature.

The environmental impacts and health dimension of UHI are assessed under the remit of ‘Climate, Environment & Health’, which accounts for 11.57% of the total.The subject of ‘Air Pollution’ (7.03%) is the relationship between rising global temperatures and the quality of the atmosphere, while ‘Climate Adaptation’ (1.85%) examines how urban areas can adapt in the context of climate change. The article entitled ‘Mosquito Disease’ (1.08%) discusses the risks to public health and analyses the impact of temperature changes on vector-borne diseases. A significant proportion of studies pertaining to the interplay between water and energy (14.29%) are collated in a collection of literature entitled ‘Water & Energy’. ‘Rainfall Patterns’ (6.76%) addresses the impact of urbanisation on rainfall regimes, while topics such as “Solar Energy” (1.3%) and “Groundwater” (1.85%) focus on energy production and water management processes.

This analysis, which evaluates the topics focused on in the studies, shows that research on the urban heat island (UHI) phenomenon is not limited to physical materials, but rather encompasses a comprehensive scientific field that includes ecological, health, and energy dimensions. However, the integration of socioeconomic dimensions, including inequality, social vulnerability, and urban justice issues, appears limited in understanding the differentiated impacts of UHI [2,48]. The evaluation of mapping and monitoring techniques is undertaken under the category of ‘Monitoring & Mapping’, accounting for 6.84% of the total. ‘Satellite Remote’ (1.66%) and ‘Drone Mapping’ (1.23%) are concerned with the examination of temperature distributions through the implementation of remote sensing techniques. The field of ‘Lidar Mapping’ (1.23%) focuses on three-dimensional urban analyses, while artificial intelligence-based modelling studies are included under the title of ‘Neural Network’ (0.82%). Methodological diversity and the use of new technologies appear to have a limited place in the studies. In particular, methodological deficiencies are evident in the adoption of new approaches such as big data analytics, advanced sensor networks, and interdisciplinary modelling. Current research also supports this conclusion [47]. The study ‘Global Climate’ (4.28%), which addresses regional variations, explores the variability of the UHI effect across diverse climate zones. The relationship between regional climate dynamics and UHI is examined by ‘Brazilian Climate’ (2.79%), ‘Japanese Climate’ (0.77%) and ‘Australian Climate’ (0.72%).

This situation show that regional diversity and interregional comparisons are underrepresented. While cities in the Global North are represented in the research, the absence of Africa or the Middle East highlights the geographical imbalance in the scope of the research [13]. These gaps clearly indicate a need for more geographically inclusive, socially integrated, and methodologically innovative research in the field of UHI studies and in cities to mitigate the effects of UHI.

5. Discussion

The UHI impact is widely recognised in the literature as a complex environmental problem that requires an interdisciplinary approach. The findings of this study provide important contributions in terms of identifying temporal changes, knowledge gaps and future research directions by revealing how the UHI effect is addressed in different disciplines. The discussion herein details the interdisciplinary nature of the UHI effect, key themes and temporal trends, with implications for future research. The findings of this study demonstrate that the UHI impact is not solely a technical issue, but rather a multifaceted issue that necessitates consideration of its environmental, social and economic dimensions. The analysis of extant literature indicates that the majority of studies on UHI impact are concentrated in three fields: environmental sciences (36.99%), meteorology and atmospheric sciences (24.21%) and construction technology (18.23%).

The BERTopic model has been employed in this study to identify the main themes, which have been found to correspond with the most critical issues highlighted in the literature on UHI. It has been determined that the themes of “Green Infrastructure Applications,” “Energy Efficiency” and “Thermal Comfort” are of significant importance in the existing literature [54,55].

Building on these insights, the main discussion points of this study can be synthesised as follows:

• Green Infrastructure Applications, Energy Efficiency, and Thermal Comfort are the most prominent themes, underscoring their centrality in UHI mitigation strategies. Practices such as green roofs and urban afforestation not only lower urban temperatures but also provide co-benefits including biodiversity conservation and improved air quality.
• The UHI effect disproportionately affects low-income neighbourhoods due to the absence of cooling infrastructures. Addressing social vulnerability and urban justice issues is essential for equitable adaptation.
• Research remains dominated by the Global North, while regions such as Africa and South Asia are underrepresented due to data scarcity and financial constraints. Greater investment and geographically inclusive studies are needed to capture regional variations.
• BERTopic provides clear advantages over static models (e.g., LDA, NMF) by capturing temporal dynamics. This allowed the study to link thematic shifts with milestones such as the Paris Agreement, COP26, and extreme climate events.
• Early research focused on descriptive climatology and documenting temperature differences, whereas recent work reflects a transition towards interdisciplinary, solution-oriented frameworks directly informing urban governance.
• Emerging use of IoT sensor networks, remote sensing, and AI-powered models offers real-time monitoring and predictive capabilities, transforming urban planning into a data-driven, adaptive process.
• Beyond environmental sciences, this study contributes methodologically to Management Information Systems by demonstrating AI-driven text mining on large bibliographic datasets.
• Evidence-based strategies such as green infrastructure, reflective materials, and ventilation corridors are gaining traction, providing actionable guidance for planners and policymakers aiming to build climate-resilient cities.

These focal points confirm the dominant themes in the literature while emphasising their broader implications for sustainability, equity, and urban resilience.

The study’s most significant findings relate to the temperature-reducing effects of green infrastructure applications and their contributions to sustainability [56,57]. As demonstrated by [58], the implementation of green roof systems and urban afforestation has been shown to be an effective measure in reducing the impact of UHI. Such practices have been demonstrated to reduce temperatures, whilst concomitantly offering secondary benefits, including the support of biodiversity and the enhancement of air quality [58].

The relationship between UHI and socioeconomic inequalities was examined by [48] who stated that the UHI effect was more pronounced in low-income regions. This study demonstrated analogous findings, indicating that the absence of cool infrastructure practices in economically disadvantaged neighbourhoods exacerbates the UHI effect.

In addition, the issue of the scarcity of UHI studies in developing countries was addressed by [13] and it was stated that the impact of UHI in regions such as Africa and South Asia has not been sufficiently studied. The findings of this study demonstrate that in order to further explore the implications of UHI in the context of developing countries, it is essential to allocate increased financial resources and data.

The merits of the BERTopic model become more evident when juxtaposed with other topic modelling approaches documented in the literature (LDA, NMF) [59]. Ref. [47] posited that the static structure of the LDA model is inadequate for the analysis of temporal trends. The present study lends support to the findings by means of an analysis of the evolution of the main themes in the literature over time, thanks to the dynamic structure of BERTopic.

The bibliometric analyses conducted by previous studies [60,61] already highlighted the growing scholarly interest in green infrastructure and energy efficiency after 2010. Our findings corroborate these shifts and, thanks to the dynamic capabilities of the BERTopic model, allow for a clearer monitoring of their evolution over time. Importantly, these temporal patterns can also be interpreted in relation to major historical milestones. For instance, the increased attention to reflective materials and cooling strategies after 2015 coincides with the Paris Agreement, which emphasised sustainable urban planning and climate mitigation. Similarly, the surge in research on thermal comfort and health impacts reflects societal responses to extreme climate events such as the 2003 European heatwave and the 2021 Canadian heat dome. The post-2020 rise in ecosystem-based and resilience-oriented studies also parallels global discussions during COP26 in Glasgow and the broader post-pandemic focus on sustainability. By situating thematic shifts within these historical and policy contexts, the analysis gains a stronger interpretive framework that connects literature dynamics with real-world climate and urban governance milestones.

The evolution of UHI studies demonstrates a clear paradigm shift from descriptive climatology towards interdisciplinary, solution-oriented frameworks. Early research predominantly focused on measuring urban–rural temperature differences and documenting the physical manifestations of the UHI effect. However, as global climate policies and urbanisation pressures intensified, the field expanded into new domains such as remote sensing, urban planning, public health, and sustainable architecture. This shift reflects a broader transition in climate science from purely observational research to integrative approaches that directly inform policy and urban governance. Notably, the increasing prominence of themes such as thermal comfort, ecosystem-based strategies, and reflective materials indicates the movement from understanding “what UHI is” towards addressing “how UHI can be mitigated and managed.”

Finally, recent research has highlighted the increasing importance of IoT and AI-enabled solutions for UHI monitoring and management [62]. The findings of this study support these perspectives and show how IoT sensor networks and satellite data can be critical for the spatial analysis of UHI, thereby informing more effective urban planning strategies.

Nevertheless, analyses conducted with BERTopic demonstrate that the subjects of green infrastructure and energy efficiency have become increasingly prevalent in research since 2010. This phenomenon may be associated with the emergence of sustainable urbanism policies and the escalating measures undertaken to combat the repercussions of climate change.

The BERTopic model implemented in the present study offers a substantial enhancement over conventional topic modelling techniques in the context of analysing the temporal variations in themes in UHI studies and identifying the relationships between topics.

The primary benefits provided by BERTopic, in contrast to conventional topic modelling techniques such as LDA, are its capacity to monitor temporal changes, ascertain the geographical distribution of themes associated with UHI, facilitate comparative analysis between cities, and examine interactions between multidisciplinary topics.

Beyond its interdisciplinary contributions to environmental sciences and urban climatology, this study also provides added value for the field of Management Information Systems (MIS) by demonstrating how advanced text mining and natural language processing techniques, such as BERTopic, can be systematically applied to large-scale bibliographic datasets. This methodological perspective enriches MIS research through showcasing the integration of artificial intelligence tools into decision support systems and research evaluation frameworks.

From the perspective of urban and regional planning, the findings offer actionable insights by revealing which strategies (e.g., green infrastructure, reflective materials, ventilation corridors) have gained prominence in the literature and how these themes have evolved over time. Such evidence-based mapping can support planners and policymakers in designing more climate-resilient cities, aligning scientific knowledge with practical planning strategies.

6. Conclusions

This study analyses in detail how the UHI effect is addressed in the scientific literature, which topics come to the fore and how these topics change over time. The main objective of the study is to identify trends in the scientific literature on the UHI effect, identify knowledge gaps and provide a roadmap for future research. In this context, BERTopic, one of the topic modelling techniques, was used to systematically examine a large academic dataset. BERTopic has made a significant contribution to understanding trends in UHI research by performing clustering and temporal analysis of themes in the literature. Demonstrating that the UHI effect should be approached in an interdisciplinary framework is one of the key findings of this study. Urban green spaces and ecosystem services, building materials and thermal properties, air pollution and climate change, urban cooling strategies, remote sensing and artificial intelligence-based analysis are among the key themes identified in the study. In particular, the application of green infrastructure and urban cooling strategies are among the topics that have received increasing attention in the scientific literature over the last few years. In addition, airflow and shading solutions, water and energy management strategies are also important areas of research aimed at reducing the impact of the UHI.

This study has several limitations. First, the dataset was limited to articles obtained only from the Web of Science (WoS) database, which may result in the exclusion of relevant studies available in other academic sources. This reliance on a single, English-language database introduces a linguistic bias and geographical restrictions, potentially underrepresenting findings from developing regions such as Africa, South Asia, or Latin America. Second, the performance of the BERTopic model is directly related to dataset quality and parameter configurations, and the absence of semantic validation by domain experts may limit the interpretability of some clusters. Although the algorithm ensures internal coherence, expert-driven cross-validation could further improve reliability. To address these limitations, future studies should expand to multilingual sources, integrate broader databases (e.g., Scopus, Google Scholar), and consider hybrid approaches supported by expert interpretation and complementary topic modelling methods.

In addition, while BERTopic ensures internal consistency through c-TF-IDF scoring, no external semantic validation procedures (e.g., expert review or coherence metrics) were applied in this study. This omission represents a methodological limitation, as the interpretability of some clusters may be affected. Future studies should complement algorithmic clustering with expert-driven validation and quantitative coherence metrics to ensure semantic reliability.

Despite this evolution, several persistent gaps remain in UHI research. A first structural limitation is geographical bias, with the majority of studies focusing on cities in the Global North, while rapidly urbanising regions in Africa, South Asia, and Latin America are underrepresented. Second, methodological constraints remain evident, particularly the limited application of semantic validation in topic modelling and the underutilization of mixed-method designs that combine quantitative and qualitative insights. Third, socio-economic blind spots persist: while UHI’s physical mechanisms are well documented, the intersections with inequality, vulnerability, and governance are insufficiently addressed. These gaps are not merely incidental but stem from systemic research practices, including the dominance of English-language publications, reliance on a small number of bibliographic databases, and limited cross-disciplinary collaboration. Addressing these root causes will be essential for future research to achieve a more inclusive and policy-relevant understanding of the UHI phenomenon.

Future studies should address the following areas in greater detail:

• The socio-economic dimensions of UHI, particularly its impact in low-income neighbourhoods, its links to health problems, and its broader influence on social inequalities.
• The regional dynamics of UHI, with a focus on tropical and arid climates, where field studies can strengthen localised strategies and improve the effectiveness of climate adaptation measures.
• The use of advanced technologies such as IoT sensor networks, artificial intelligence, and remote sensing for more accurate measurement, real-time monitoring, and predictive analysis of UHI impacts.
• The development of comparative methodological approaches by combining different topic modelling techniques (LDA, NMF, BERTopic) to generate more comprehensive insights into research trends.

Future research should further strengthen topic modelling results by integrating semantic validation procedures. This may include expert-based validation workshops, where domain specialists assess cluster coherence, or the application of topic coherence metrics such as UMass, CV, or NPMI. Such hybrid approaches will enhance the interpretability and validity of BERTopic analyses, thereby reinforcing their utility for policy and research design.

In this context, the study aims to provide a strong reference for future academic research and urban planning policy by systematising the existing knowledge on the UHI effect. One of the most important future goals of sustainable urbanism will be to develop regional and global strategies to reduce the effects of the UHI.