Daytime Surface Urban Heat Island Variation in Response to Future Urban Expansion: An Assessment of Different Climate Regimes

Abstract

This study focuses on assessing the physical growth of cities and the land-cover changes resulting from it, which play a crucial role in understanding the environmental impacts and managing phenomena such as the Daytime Urban Surface Heat Island Intensity (DSUHII). Predicting the trends of these changes for the future provides valuable insights for urban planning and mitigating thermal effects in arid environments. This research aims to evaluate the spatial and temporal changes in the intensity of urban surface heat islands in cities under different climatic conditions, resulting from land-cover changes in the past, and to predict future trends. For this purpose, Landsat satellite data products, including Surface Reflectance with a 30-m resolution and Land Surface Temperature (LST) originally at a 100 (120)-meter resolution for Landsat 8 (Landsat 5) (resampled to 30 m for compatibility), along with a database of underlying criteria affecting urban growth, were used to analyze land-cover and LST changes. The land-cover classification was carried out using the Support Vector Machine (SVM) algorithm, and its accuracy was assessed. Spatial and temporal changes in LST and land-cover classes were quantified using cross-tabulation models and subtraction operators. Subsequently, the impact of land-cover changes on LST in different climates was analyzed, and the trends of land-cover and DUSHII changes were simulated for the future using the CA–Markov model. The results showed that in the humid climate (Babol and Rasht), built-up areas increased by over 100% from 1990 to 2023 and are projected to grow further by 2055, while green spaces significantly decreased. In the cold–dry climate (Mashhad), urban development increased dramatically, and green spaces nearly halved. In the hot–dry climate (Yazd and Kerman), built-up areas tripled, and the reduction of green spaces will continue. Additionally, in cities with hot and dry climates, a significant area of barren land was converted into built-up areas, and this trend is predicted to continue in the future. DSUHII in Babol increased from 2.5 °C in 1990 to 5.4 °C in 2023 and is projected to rise to 7.8 °C by 2055. In Rasht, this value increased from 2.9 °C to 5.5 °C, and is expected to reach 7.6 °C. In Mashhad, the DSUHII was negative, decreasing from −1.1 °C in 1990 to −1.5 °C in 2023, and is projected to decline to −1.9 °C by 2055. In Yazd, DSUHII also remained negative, decreasing from −2.5 °C in 1990 to −3.3 °C in 2023, with an expected drop to −6.4 °C by 2055. Similarly, in Kerman, the intensity of DSUHII decreased from −2.8 °C to −5.1 °C, and it is expected to reach −7.1 °C by 2055. Overall, the conclusions highlight that in humid climates, DSUHII has significantly increased, while green spaces have decreased. In moderate, cold, and dry climates, a gradual reduction in DSUHII is observed. In the hot–dry climate, the most substantial decrease in DSUHII is evident, indicating the varying impacts of land-cover changes on DSUHII across these regions.

1. Introduction

In recent years, global warming and its negative impacts on the quality of life in human societies have become a major international concern [1,2]. This phenomenon is primarily driven by the increase in CO₂ and other greenhouse gas emissions, which are closely associated with the rapid growth of industrial activities and climate-unfriendly anthropogenic behaviors. While urban growth and population expansion themselves are not direct causes of global warming, they contribute to higher emissions when accompanied by unsustainable lifestyles, such as increased private vehicle use, air travel, and high energy consumption [3,4]. Furthermore, uncontrolled urban development leads to the degradation of the surrounding natural ecosystems, compounding environmental challenges. Urban physical growth and land-use diversity, especially in industrial areas, play a crucial role in altering microclimatic patterns [5,6,7]. These effects are intensified due to differences in the material composition and thermal properties of urban surfaces, resulting in significant temperature variations between urban areas and their surrounding regions [8].

The Surface Urban Heat Island (SUHI) is one of the most significant environmental phenomena associated with global warming, primarily driven by human activities and urban development [9,10]. This phenomenon reflects the fact that certain urban areas, particularly city centers, are several degrees warmer than their surrounding regions due to changes in surface structures and land-use patterns [11,12]. A key factor contributing to SUHI formation is the replacement of natural covers with impermeable surfaces such as asphalt, pavements, buildings, and other urban structures [13,14]. This substitution reduces natural cooling capacities and significantly decreases evapotranspiration, a crucial process for temperature regulation [11,15]. Impervious surfaces absorb more solar energy and release a greater amount of sensible heat into the environment, whereas vegetative covers and permeable surfaces moderate air temperatures through natural evapotranspiration [16,17]. Additionally, it is important to note that in densely built-up areas, a considerable portion of the measured thermal signal, especially in 100-m resolution thermal data, originates from rooftops rather than ground-level surfaces. This three-dimensional effect can strongly influence the spatial pattern of SUHI, as observed in many European cities and cities like Cairo, where up to 60% of the thermal information may come from roof surfaces. Considering the urban structure of the Iran cities—particularly Tehran and Mashhad, which have areas with relatively dense mid-rise and high-rise developments—such effects may partially influence the observed LST patterns. These changes cause urban air temperatures—especially during the daytime—to consistently exceed those of rural and suburban areas [18,19]. These changes cause urban land surface temperatures (LST)—especially during the daytime—to consistently exceed those of rural and suburban areas [18,19]. The consequences of SUHI extend beyond temperature increases and have far-reaching impacts on the environment, public health, and urban economies [20]. These include a higher energy consumption for cooling, intensified air pollution, reduced quality of life, and increased heat-related illnesses [21,22]. Such negative effects have drawn the attention of urban managers, planners, and policymakers to the implications of urban growth and the need to enhance urban livability. Therefore, a comprehensive understanding of SUHI, its spatial and temporal patterns, and future projections is crucial for urban monitoring and management [23,24,25,26,27]. This knowledge can guide the development of effective strategies to mitigate negative impacts, improve urban livability, and enhance environmental and social resilience.

Multi-temporal remote sensing data, by providing continuous and repeated information on LST, enables the extraction of SUHI intensity (SUHII) and the monitoring of its changes over different periods [11,28]. The use of multi-temporal data allows for a precise evaluation of the spatial and temporal variations of the SUHII over past years and facilitates the analysis of its behavioral patterns. Beyond retrospective assessments, spatio-temporal modeling based on satellite data has become a crucial tool for predicting future SUHII [25,26,27,29,30,31]. This approach, by analyzing historical patterns and applying them to predict future scenarios, assists urban managers and policymakers in designing more effective strategies to mitigate heat island effects and make evidence-based decisions. These analyzes are essential for sustainable urban planning and for reducing the vulnerability of urban populations to the effects of global warming. Given the rising trends of global warming and heat-generating human activities, predicting SUHII using spatio-temporal models and multi-temporal remote sensing data has become a critical research area. This approach facilitates the analysis of future changes and supports the development of effective strategies for managing and mitigating the impacts of SUHI [23,25,26,27,29,30,31,32,33].

Ahmed et al. [34] predicted that in Dhaka, Bangladesh, areas with high temperatures would increase by 2029, with 87% of the region experiencing temperatures exceeding 30 °C. Mushore et al. [29], using the CA–Markov model, forecasted that urban growth in Harare, Zimbabwe would lead to increased LST and a reduction in green spaces and agricultural lands. Rahman et al. [33] demonstrated that the expansion of built-up areas in Dammam, Saudi Arabia, would increase surface temperatures while decreasing the extent of cooler regions. In the study by Deilami and Kamruzzaman [32], it was predicted that the intensity of the urban heat island in Brisbane, Australia, would increase in the future under a business as usual scenario. Firozjaei et al. [23] projected that the expansion of built-up areas in Babol, Iran, would lead to rising surface temperatures and significant changes in temperature classes. Wang et al. [35] predicted that the increase in built-up areas in Nanjing, China would increase surface temperatures, while vegetation and agricultural lands would help cool the city. Nadizadeh Shorabeh et al. [36], forecasted that the expansion of built-up areas in Tehran, Iran and human activities would intensify the urban heat island effect and increase the prevalence of higher temperature classes. Nurwanda and Honjo [30], using Landsat imagery combined with a multilayer neural network and the Markov chain, predicted that urban expansion in Bogor would increase by approximately 3760 hectares by 2027, leading to a rise in LST in the city center. Kiavarz et al. [24] predicted that the expansion of built-up areas in Tehran would lead to increased surface temperatures and a more intense urban heat island effect, with an average temperature rise of 1.6 kelvin as green areas are converted into built-up zones. Ramzan et al. [25] analyzed the temporal changes in LST and land cover in Lahore, Pakistan, and examined their impacts on predicting future LST and land-use/land-cover (LULC) patterns. Their findings indicated that between 1992 and 2020, built-up areas increased by 41.8%, while vegetated areas, barren land, and water bodies decreased by 24%, 17.4%, and 0.4%, respectively. Additionally, urban LST increased by 4.3 kelvin, and it was projected that LST would rise by another 1.3 kelvin by 2030. Li and Zheng [31] developed an LST prediction model for urban environments using generative adversarial networks (GANs) and LiDAR data. This model can rapidly and accurately generate LST maps from urban maps. They tested the model using data from New York and Landsat images, constructing and validating it with paired training images. The study demonstrated that accurate and rapid predictions could assist urban designers and planners in mitigating high surface temperatures and developing interactive design and planning tools. Li et al. [27], in this study, predicted LULC changes and LST changes in Harbin, China. The results showed that from 2005 to 2030, urban areas increased by 27.81%, while forests and grasslands decreased by 61.07%. High-temperature areas increased by 40.86% in winter and 60.90% in summer.

Numerous studies have been conducted on the impact of land-cover changes and the conversion of natural land to built-up areas on LST and the intensity of the SUHI. These studies have explored various aspects of these changes, including the effects of land-use changes on surface temperature, urban growth, and SUHI intensity in different areas. However, assessing how the results of these changes manifest in different climates is of high importance. In diverse climates, the type of land cover and how it influences LST and SUHI can vary significantly. These differences may stem from the specific climatic conditions of each region and the environmental features that influence the formation and intensity of SUHI. Additionally, predicting future changes in this area can provide valuable information for urban planning and natural resource management. Predictions related to land-cover and LST changes in different climates enable us to mitigate the adverse effects of SUHI and propose optimal solutions for reducing their intensity. Therefore, examining the effects of land-cover changes and related predictions in various climates can lead to a better and more accurate understanding of how SUHII form and change at the urban level. The primary objective of this study is to compare the spatial and temporal changes of the Daytime SUHI Intensity (DSUHII) in cities located in different climatic conditions in the past and make future predictions. The main question of this study is the following: do the spatial and temporal changes of DSUHII in different cities depend on the climatic regime of the region? The innovation of this study lies in evaluating and predicting changes in DSUHII driven by land-use changes associated with urban physical growth across different climatic conditions.

2. Study Area

In this study, to compare the spatial and temporal changes of the DSUHII in cities located in different climatic conditions in the past and to predict future trends, five cities were selected as follows: Mashhad, Rasht, Yazd, Kerman, and Babol as shown in Figure 1. The population of these cities has significantly increased over the past decades. According to the Iranian Statistical Center, the population of the metropolitan areas of Mashhad, Rasht, Yazd, Kerman, and Babol has increased from 2,020,000, 616,000, 370,000, 530,000, and 1,145,000 people in 1990 to 3,450,000, 980,000, 695,000, 1,050,000, and 1,810,000 people in 2024, respectively. These case studies were selected based on geographical location diversity and climatic conditions (Figure 1 and Figure 2).

-

Mashhad: Located in the northeast of Iran at a latitude of 36°17′ and longitude of 59°35′, Mashhad has a moderate, semi-cold, and dry climate. The city’s average annual temperature is approximately 17.1 °C, and the annual rainfall is around 300 mm. The city is situated at an elevation of 970 m above sea level.

-

Rasht: Situated in northern Iran at a latitude of 37°16′ and longitude of 49°36′, Rasht has a moderate, humid, Caspian climate. The average annual temperature is around 16 °C, with an annual rainfall of approximately 1500 mm. Rasht lies at an elevation of about 10 m above sea level.

-

Yazd: Positioned at a latitude of 31°25′ and longitude of 54°24′, Yazd experiences a hot, dry, and desert climate. The average annual temperature is about 19.2 °C, with an annual rainfall around 120 mm. Yazd is located at an elevation of 1240 m above sea level.

-

Kerman: Located in southeastern Iran at a latitude of 30°18′ and longitude of 56°56′, Kerman has a hot, dry climate. The average annual temperature is around 18.8 °C, with an annual rainfall of about 200 mm. Kerman is situated at an elevation of 1750 m above sea level.

-

Babol: Situated in northern Iran at a latitude of 36°33′ and longitude of 52°43′, Babol has a humid, Caspian climate. The average annual temperature is approximately 15.9 °C, and the annual rainfall is around 1500 mm. Babol lies at an elevation of about 30 m above sea level.

3. Data and Methods

3.1. Data

In the present study, surface reflectance and surface temperature Level-2 products were used to create land-cover and LST maps for five cities—Mashhad, Yazd, Rasht, Babol, and Kerman—during the years 1990 and 2024. These products are among the most important data sources for environmental studies and land-cover changes. For each city, images from the spring and summer months were selected, ensuring the inclusion of periods with the highest LST values. A total of eight images were chosen for each city, and the average of these images was calculated. The data from these products were analyzed annually, and the annual averages for the years under investigation were derived. Surface land-cover, LST, and DSUHII mapping were employed, providing a comprehensive approach to assess the changes and impacts across different cities and climatic conditions. These products are available through NASA’s EarthExplorer website is developed by United States Geological Survey (https://earthexplorer.usgs.gov, accessed on 29 October 2024). The data were analyzed on an annual basis, and the annual mean for the years studied was calculated. The Land Surface Temperature (LST) product originally has a spatial resolution of 100 (120) meters for Landsat 8 (Landsat 5), which has been resampled to 30 m using a cubic convolution method by the data provider to ensure compatibility with the Surface Reflectance products. The Surface Reflectance product includes seven spectral bands, including blue, green, red, near-infrared, SWIR 1, and SWIR 2. These bands provide essential information about built-up areas, vegetation cover, soil cover, and other surface characteristics. The spatial resolution of these products is 30 m, enabling a more accurate analysis of spatial and temporal changes. Specifically, the use of these data helps analyze the impact of land-cover changes on LST variations and urban heat islands at different temporal and spatial scales.

To prepare training and testing data for classification purposes, data collected during field visits were used alongside Landsat true-color composite images, and high-resolution images available in Google Earth. These data are particularly useful in more accurately identifying various land-cover classes (such as urban areas, agricultural lands, vegetation cover, etc.). In this study, approximately 500 samples from each class were selected for each city, with 70% of them used as the training dataset and the remaining 30% for the testing dataset. This training approach, especially when combined with automatic satellite image classification, yields a high accuracy in simulating land-cover changes.

Additionally, to create maps of spatial criteria affecting land-cover change predictions, this study used a database that includes the slope files of parks, rivers, water resources, and road networks for various cities, which were downloaded from OpenStreetMap. Moreover, a Digital Elevation Model (DEM), derived from the AW3D product, was used to generate elevation and slope maps with a spatial resolution of 30 m for each of the cities under study. These data is developed by Japan Aerospace Exploration Agency (JAXA), Remote Sensing Technology Center of Japan (RESTEC), and NTT DATA Corporation (NTTD). These data are accessible via the website https://www.aw3d.jp/en/ accessed on 29 October 2024.

3.2. Methods

To achieve the research objective, five main stages were implemented as illustrated in Figure 3. In the first step, land-cover maps for the selected cities in different years were prepared using the SVM algorithm, and accuracy was assessed. In the second step, LST maps for the various cities in different years were normalized. Then, spatial and temporal variations of the normalized land cover and LST were quantified using cross-tabulation models and the subtraction operator, respectively. In the third step, the impact of class transformations between land-cover types on normalized LST variations in cities with different climatic conditions was calculated and compared. In the fourth step, land-cover and LST prediction maps for future years for various cities were prepared based on the CA–Markov cellular model. In the fifth step, DSUHII for these cities were modeled for both past and future years, and the trends in spatial and temporal changes of this phenomenon were examined for both past and future time intervals. The trends of this phenomenon in cities with different climatic conditions were then compared.

3.2.1. Land-Cover Mapping

The land-cover classes in the studied cities include built-up areas, barren land, green/agricultural space, and water. The city of Babol lacks a barren land class, and the cities of Kerman and Yazd lack a water class. In this study, the SVM algorithm was used for land-cover classification in the different cities. SVM is a non-parametric supervised statistical method. In this algorithm, using all the spectral bands (reflective and thermal) and an optimization algorithm, the samples that form the class boundaries are obtained, and a linear decision boundary is calculated to separate the classes. These samples are known as support vectors. The support vector machine method is a binary classification method that separates different classes by determining an optimal separating hyperplane in the feature space of training data, maximizing the margin between the classes. The hyperplane that provides the maximum margin between the two classes is called the optimal hyperplane. One of the capabilities of the support vector machine is its ability to overcome the problem of the non-linear distribution of training data. In this case, using kernel functions, the data is mapped to a space with a higher dimension where a better separability is achieved.

The accuracy of the land-cover classification maps for the various cities was evaluated using a test dataset. Statistical metrics such as the overall accuracy and the Kappa coefficient were used to assess the accuracy. Finally, after evaluating the classification accuracy, the area corresponding to the land uses in all the years was extracted. The area changes and the spatial variations of the land-cover classes during the study period for the different cities were analyzed using the Cross-tab model.

3.2.2. Land-Cover Prediction

Among the various methods used to predict LULC changes, algorithms such as Artificial Neural Networks [37], Markov Chains [38], Logistic Regression [39], Decision Trees [40], Machine Learning Models [41], and Fuzzy Logic [42] have played a significant role. However, Cellular Automata [43] has been recognized as one of the most effective methods in modeling LULC dynamics due to its high capability in simulating complex and nonlinear processes [44]. As a result, various Cellular Automata-based approaches have been developed, including ANN-CA [45], CLUE-S models [46], LTM [47], the SLEUTH model [48], FLUS [49], and PLUS [50]. Despite these advancements, most Cellular Automata-based models can only simulate changes in one land-use class, whereas land-use changes typically occur simultaneously and interactively [51]. Markov Chains, due to their simplicity and ability to predict future trends based on the current state, are considered an appropriate method for modeling LULCC [52]. However, despite the ability to predict multidirectional changes [53], it does not specify the exact locations of the changes. The combination of Cellular Automata and Markov Chains has been proposed as an effective solution to overcome these limitations [44].

The CA–Markov model, due to its extensive capabilities in simulating dynamic processes, the bottom-up approach, a high data processing efficiency, and an ease of calibration, has been widely used in land-use and land-cover change studies [51]. This model has the ability to simulate complex and simultaneous land-use change patterns, making it one of the key tools in spatial dynamics analysis.

This is done using a transition probability matrix and a transition area matrix [54]. The transition probability matrix indicates the probability that each pixel in a given class, based on its previous state, will convert to other classes during a defined time period. In addition to the Markov Chain model outputs, in this study, a Multicriteria Evaluation (MCE) based on fuzzy logic was used to generate suitability maps for the land-use changes [55,56,57]. This method combines factors and constraints that affect urban growth.

The criteria considered for simulating future urban expansion include the following: the distance from rivers, the distance from main roads, the distance from existing built-up areas, and the distance from central business districts, as well as topography, including slope and elevation [57]. In the CA–Markov model applied in this study, we included several physical and infrastructural drivers of urban growth, such as slope, elevation, the distance to rivers, roads, existing built-up areas, and central business districts. While these factors can indirectly reflect some aspects of urban development, it is important to acknowledge that the model does not directly incorporate socioeconomic, regulatory, or market-related variables.

To ensure the consistency of data scales, all factor maps were normalized to the range of 0 to 1 using Min–Max linear transformation. The study area’s constraints include parks, rivers, water resources, and future road construction plans by the municipalities of each city, which were converted into Boolean maps (0 and 1). The weight of each criterion was determined using the Analytic Hierarchy Process (AHP) method based on expert opinions [55,56]. The experts included municipal employees, urban planning specialists, sustainable development experts, and spatial planning experts.

One fundamental spatial element influencing the dynamics of many change events is proximity: areas tend to be more likely to change to a class when they are near existing areas of the same class, i.e., the expansion phenomenon as reflected in the Tobler’s first law of geography. These phenomena can be effectively modeled using CA. CA is a discrete dynamic system where the state of each cell at time t + 1 is determined by the state of neighboring cells at time t, according to predefined transformation rules. As an analytical model with spatiotemporal dynamics, CA can simulate changes in high-resolution two-dimensional space.

In this study, this model was used to predict land cover for the year 2056 in different cities based on land-cover changes from 1990 to 2024. Finally, using the Cross-tab model, land-cover changes between 1985–2024 and 2024–2058 were analyzed.

3.2.3. LST Prediction

To predict the LST maps for future years based on the impact of projected land-cover changes, the results from the CA–Markov model were utilized. This approach was proposed in a study by Firozjaei et al. [23] and has been used in various studies [24,36,58]. For this purpose, the spatial relationship between the LST change maps and the land-cover changes over past time periods was assessed. Here, the effect of land-cover changes on the LST for the study period was also analyzed. The average LST changes for different land-use classes were calculated. The computed average LST changes were applied to the predicted land-cover change maps for the period 2024–2058, resulting in a map of LST changes due to land-cover changes for this period. Finally, by combining the 2024 LST maps with the LST change map due to land-cover changes for the period 2024–2058, the LST map for the year 2058 was generated.

3.2.4. LST Classes and DSUHII Changes

To assess the spatial and temporal changes in LST classes over different periods, the LST maps for various years were first normalized. Normalizing the LST enables a better and more precise comparison of the spatial distribution of the LST obtained from images under different seasonal and atmospheric conditions over a given period. For this purpose, cold pixels, including those with full and healthy vegetation, were identified for each date. In the second step, the average LST of the cold pixels was calculated. In the third step, the LST value of each pixel was subtracted from the average LST of the cold pixels [58,59]. Then, the mean (${\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{m}\mathrm{e}\mathrm{a}\mathrm{n}}$) and standard deviation (${\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{s}\mathrm{t}\mathrm{d}}$) of the normalized LST maps were calculated. The normalized LST maps for each year were classified into five relative classes: very low ($\mathrm{T}\le {\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{m}\mathrm{e}\mathrm{a}\mathrm{n}}-1.5{\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{s}\mathrm{t}\mathrm{d}}$), low (${\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{m}\mathrm{e}\mathrm{a}\mathrm{n}}-1.5{\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{s}\mathrm{t}\mathrm{d}}<\mathrm{L}\mathrm{S}\mathrm{T}\le {\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{m}\mathrm{e}\mathrm{a}\mathrm{n}}-{\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{s}\mathrm{t}\mathrm{d}}$), medium (${\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{m}\mathrm{e}\mathrm{a}\mathrm{n}}-{\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{s}\mathrm{t}\mathrm{d}}<\mathrm{L}\mathrm{S}\mathrm{T}\le {\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{m}\mathrm{e}\mathrm{a}\mathrm{n}}+{\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{s}\mathrm{t}\mathrm{d}}$), high (${\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{m}\mathrm{e}\mathrm{a}\mathrm{n}}+{\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{s}\mathrm{t}\mathrm{d}}<\mathrm{L}\mathrm{S}\mathrm{T}\le {\mathrm{T}}_{\mathrm{m}\mathrm{e}\mathrm{a}\mathrm{n}}+1.5{\mathrm{T}}_{\mathrm{s}\mathrm{t}\mathrm{d}}$), and very high ($\mathrm{L}\mathrm{S}\mathrm{T}>{\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{m}\mathrm{e}\mathrm{a}\mathrm{n}}+1.5{\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{s}\mathrm{t}\mathrm{d}}$). Next, the spatial and temporal changes in the different LST classes and the area of each class were analyzed and compared for different cities located in various climates. Finally, the DSUHII for different cities was calculated based on the difference in the mean normalized LST between urban and non-urban areas in the years 1990, 2024, and 2052, and the results were compared. In this study, the DSUHII was estimated using class-based land-cover averages. While this approach allows for a broad understanding of surface temperature patterns, it involves a simplification by not accounting for within-class variability or pixel-level temperature heterogeneity.

4. Results

4.1. Land-Cover Maps

The overall accuracy of the land-cover maps prepared for the cities of Babol, Rasht, Mashhad, Yazd, and Kerman in the years 1990 and 2024 is estimated to be 93% and 92%, 91% and 92%, 85% and 87%, 85% and 89%, and 86% and 85%, respectively. The Kappa coefficient for these cities is 0.91 and 0.90, 0.89 and 0.90, 0.83 and 0.86, 0.83 and 0.86, and 0.85 and 0.84, respectively. The Kappa coefficient is a statistical measure that assesses the accuracy of the model compared to a random model, with higher values indicating a high agreement between the observed data and model predictions. Given the challenges in accurately distinguishing built-up and barren land cover in the cities of Mashhad, Yazd, and Kerman, the classification accuracy of land cover in Rasht and Babol has been higher. As shown in Figure 4, the changes in land cover in the cities of Rasht, Babol, Mashhad, Yazd, and Kerman from 1990 to 2024 and projected for 2058 indicate a trend of expanding urban areas and decreasing green spaces. In Rasht and Babol, built-up areas and green spaces are identified as the dominant land covers, while in Mashhad, Yazd, and Kerman, built-up and barren lands are predominant. Notably, the absence of water bodies in Yazd is due to the region’s dry climate and scarcity of water resources. A visual analysis of the maps reveals a significant conversion of green spaces to urban areas over recent decades across all the studied cities. This trend, primarily driven by population growth and urban development, is likely to continue in the future, with various environmental and social implications. The reduction in green spaces leads to an increased air pollution, decreased quality of life for residents, and threats to biodiversity. Therefore, urban planning and environmental policy are crucial for managing these land-cover changes sustainably.

Figure 5 illustrates the area of different land-cover classes in the cities of Babol, Rasht, Mashhad, Yazd, and Kerman for the years 1990, 2024, and the projected 2058. In 1990, the area of built-up land in the cities of Babol, Rasht, Mashhad, Yazd, and Kerman was 12.65, 26.7, 120.6, 49.6, and 28.75 km², respectively. By 2024, these areas had increased to 28.35, 69.2, 278.6, 136.9, and 107.11 km², respectively. The area of built-up land has significantly increased across all the cities studied during the observed period. The highest growth rates in built-up land area are observed in Kerman (57%) and Yazd (49%), indicating a rapid urban development trend in these regions. Projected results suggest that the area of built-up land in these cities will reach 38.52, 94.3, 393.4, 205.1, and 168.49 km² by 2058, respectively. The area of green spaces and water bodies for Babol in 1990 was 53.31 and 0.7 km², respectively. By 2024, these values changed to 37.11 and 1.2 km². For Rasht, the area of green spaces, barren land, and water bodies in 1990 was 116.1, 27.3, and 6.5 km², respectively, which changed to 97.5, 2.5, and 7.5 km² by 2024. In Mashhad, the areas of these land covers in 1990 were 210.9, 443.7, and 0.1 km², which changed to 100.1, 395.7, and 278.6 km² by 2024. For Yazd, the area of barren land and green spaces in 1990 was 427.9 and 53.9 km², respectively, and decreased to 360.6 and 33.9 km² by 2024. For Kerman, these values were 294.5 and 102.7 km² in 1990, and changed to 253.7 and 65.2 km² by 2024. In contrast to built-up areas, the area of green spaces has decreased across all cities. This reduction indicates a prioritization of urban development over the preservation of green spaces. The most significant decrease in green-space area is observed in Mashhad (42%), highlighting the intense pressure of urban development on green spaces in this city. In some cities, such as Mashhad, Yazd, and Kerman, the area of barren land has decreased, suggesting a conversion of these areas to other uses, such as agriculture or construction. The area of built-up land in the cities of Babol, Rasht, Mashhad, Yazd, and Kerman is expected to increase by 36%, 36%, 41%, 49%, and 57%, respectively, between 2024 and 2058. During the same period, green spaces in these cities are projected to decrease by 27%, 24%, 42%, 31%, and 23%, respectively. The area of barren land in the cities of Mashhad, Yazd, and Kerman is expected to decrease by 18%, 16%, and 18%, respectively. These development trends, driven by rapid population growth and the need for new infrastructure, are occurring in most cities worldwide. Such changes can lead to increased pollution, reduced biodiversity, and other environmental issues. To address these challenges, there is a need for sustainable policies and more precise planning in the management of resources and urban green spaces.

4.2. Land-Cover Changes

The land-cover change maps for the studied cities over the period from 1990 to 2024 and projections for 2024 to 2058 are shown in Figure 6. The most significant conversion of green spaces and barren lands to built-up areas has occurred in the suburbs and surrounding urban areas. A substantial amount of green space within urban areas has been converted to built-up land from 1990 to 2024. This trend is expected to continue in the coming years. The most significant change from green space to built-up land has occurred in the northwest sections of Mashhad. Forecasted land-cover changes indicate that the cities of Yazd and Mashhad will lose a significant portion of their green spaces in the future. Overall, in all the studied cities, the predominant trend has been the conversion of natural lands (green spaces and barren lands) to built-up areas, reflecting the rapid urbanization and prioritization of economic development over environmental preservation. In Babol, between 1990 and 2024, 16.2 km² of green space were converted to built-up land. Forecasts suggest that in the next 32 years, 10.1 km² of green space will be converted to built-up land. For Rasht, between 1990 and 2024, 14.1, 28.2, and 0.13 km² of barren land, green space, and water, respectively, were converted to built-up land. Additionally, 10.98 and 1.59 km² of barren land and water were converted to green space. Only 1.01 km² of green space were converted to barren land. The most significant changes were related to the conversion of green space to built-up land. It is expected that in the next 32 years, 23.59 and 1.50 km² of green space and barren land in Rasht will be converted to built-up land. In Mashhad, 95.4 and 62.7 km² of barren land and green space, respectively, were converted to built-up land. Additionally, 72.7 km² of green space were converted to barren land. Over the past 32 years, 24.6 km² of barren land were converted to green space. Also, due to dam impoundment, 0.65 and 0.21 km² of barren land and green space were converted to water bodies. The most significant changes were related to the conversion of barren land to built-up land. It is anticipated that in the next 32 years, 29.26 and 81.42 km² of green space and barren land in Mashhad will be converted to built-up land. Due to physical growth in Yazd over the past 32 years, 61.95 and 25.37 km² of barren land and green space, respectively, were converted to built-up land. Additionally, 4.28 km² of green space were converted to barren land. During this period, 9.65 km² of barren land were converted to green space. As in Mashhad, the most significant changes were related to the conversion of barren land to built-up land. It is expected that in the next 32 years, 12.47 and 52.16 km² of green space and barren land in Yazd will be converted to built-up land. In Kerman, 56.65 and 21.70 km² of barren land and green space, respectively, were converted to built-up land. Additionally, 44.37 km² of green space were converted to barren land. During this period, 28.54 km² of barren land were converted to green space. It is anticipated that in the next 32 years, 12.01 and 49.37 km² of green space and barren land in Kerman will be converted to built-up land.

4.3. LST Maps

The LST maps for the studied cities for the years 1990, 2024, and 2058 are shown in Figure 7. In all three cities, due to specific environmental conditions, including the juxtaposition of built-up areas, green spaces, barren lands, and water bodies, and high spatial variability in topographic conditions, there are significant differences in maximum and minimum temperatures. The spatial distribution of LST varies and is heterogeneous across different years. For the cities of Babol and Rasht, the highest LSTs are found in the built-up areas within the urban limits. In Mashhad, Yazd, and Kerman, the highest LSTs are observed in the barren areas surrounding the cities. Visually and in interpreting thermal patterns on the LST maps, the temperature difference between urban and non-urban areas is higher in Babol and Rasht compared to Mashhad, Yazd, and Kerman. Warm clusters in Babol and Rasht are concentrated in urban areas and have increased in extent over the past 32 years. Additionally, forecast results for these two cities indicate that warm clusters will expand within their boundaries. LSTs in urban areas of Yazd, Kerman, and Mashhad are lower than those in the suburbs. In these three cities, especially Kerman and Yazd, areas with lower LSTs are located within the urban limits. Future LSTs in Kerman and Yazd are expected to be cooler compared to the suburbs. For Mashhad, an increase in LST is predicted, but the increase will be less than that for Babol and Rasht. For Rasht (Babol), the average LST of built-up areas for the years 1990 and 2024 was 36.1 °C (37.8 °C) and 40.3 °C (42.6 °C), respectively, which, like Babol, is higher than the average LST of barren lands, green spaces, and water bodies in this area. For Mashhad, the average LST of built-up areas, barren lands, green spaces, and water bodies for 1990 was 39.9 °C, 43.3 °C, 36.4 °C, and 28.7 °C, respectively. By 2024, these values changed to 39.7 °C, 43.4 °C, 36.7 °C, and 29.5 °C, respectively. For Yazd, the average LST of built-up areas, barren lands, and green spaces in 1990 was 43.3 °C, 47.2 °C, and 39.8 °C, respectively. By 2024, these values had changed to 41.8 °C, 48.8 °C, and 40.9 °C. In Kerman, the average LST of built-up areas, green spaces, and barren lands in 1990 was 47.9 °C, 46.9 °C, and 51 °C, respectively. These values changed to 40.1 °C, 45 °C, and 48.5 °C by 2024. The greatest spatial heterogeneity in LST values is observed for built-up areas in Babol, Rasht, and Yazd, while for Mashhad and Kerman, the greatest heterogeneity is related to green spaces. In Babol and Rasht, the highest average LSTs correspond to the built-up land class. In the other three cities—Mashhad, Yazd, and Kerman—the average LST of barren lands is higher than that of other land types. This is attributed to human activities and the physical, chemical, and geometric properties of these land types. Generally, sensible heat flux is higher in built-up and barren lands compared to other land types. The increase in commercial and industrial centers, high volumes of motor vehicle traffic, and heavy traffic due to connections with other cities contribute to a more rapid increase in LST in built-up areas compared to other lands. Water bodies, due to their high heat capacity, exhibit the lowest temperatures. LST is influenced by various surface conditions; areas with more vegetation cover tend to have lower LSTs compared to areas lacking vegetation. Vegetation creates a natural air conditioning system by absorbing solar energy and transpiring water through its leaves. Therefore, green spaces, with their higher transpiration due to lushness, have lower LSTs compared to barren and built-up lands.

4.4. LST Changes

The LST change maps for the cities of Rasht, Mashhad, and Yazd over the period 1990–2024 are shown in Figure 8. The most significant increases and decreases in LST in these regions over the past 32 years were 16 °C and −12 °C, respectively. The physical development of Babol and Rasht in recent decades has had considerable environmental impacts, including the transformation of green spaces both within and around the city into built-up areas. This has led to the replacement of natural surfaces with impermeable surfaces like buildings and roads, resulting in a reduction in vegetation cover, loss of natural cooling systems, and an increase in LST. In contrast, in the other three regions, especially Yazd and Kerman, the expansion of built-up areas into barren lands has resulted in a notable decrease in LST. However, in some parts of these areas, the conversion of green spaces to built-up areas has led to an increase in LST. Additionally, in Rasht, LSTs in areas converted from barren lands to green spaces or water bodies have decreased significantly. The highest increase in LST in built-up areas over the past 32 years was observed in Babol. On average, the LST in the built-up areas of this city increased by approximately 5.3 °C. In Rasht, the LST of built-up areas increased by about 3.5 °C. For Mashhad, the average increase in LST in these areas was 0.8 °C, while in Yazd and Kerman, the LSTs decreased by approximately 1.3 °C and 2.2 °C, respectively. The average increase in LST due to the conversion of green spaces to built-up areas in Babol, Rasht, Mashhad, Yazd, and Kerman was 6.5 °C, 5.9 °C, 3.7 °C, 3.1 °C, and 2.8 °C, respectively. Additionally, the conversion of water bodies to built-up areas in Rasht led to an average increase of 6.6 °C in LST. In Rasht, the conversion of barren lands to built-up areas caused an average LST increase of 3 °C. This change in Mashhad was accompanied by a decrease of 2.5 °C. For Yazd and Kerman, the LSTs decreased by 5.5 °C and 1.6 °C, respectively, due to the conversion of barren lands to built-up areas. The average decrease in LST due to the conversion of barren lands to green spaces in Rasht, Mashhad, Yazd, and Kerman was 4.1 °C, 4.7 °C, 5.1 °C, and 5.8 °C, respectively. The most significant decrease in LST due to the conversion of barren lands to water bodies was observed in Mashhad. The green spaces, barren lands, and water bodies that remained unchanged exhibited the least LST variation in these areas.

4.5. LST Classification Maps

The LST classification maps for the studied cities for the years 1990, 2024, and 2058 are shown in Figure 9. In Rasht and Babol, most areas with high and very high temperature classes are located within the urban boundaries. The cold and very cold temperature classes in these two cities are found in areas with green space and water bodies. The areas with high and very high temperature classes have increased significantly from 1990 to 2024. Forecast results indicate that this trend will continue until 2058. For Mashhad, Yazd, and Kerman, the high and very high temperature classes are located outside the urban areas. Pixels in urban areas are typically heterogeneous, including green spaces, shaded areas, and moisture, resulting in lower LSTs compared to the surrounding areas with barren land. These regions in Mashhad, Yazd, and Kerman primarily consist of cold and very cold temperature classes. Over the years, the conversion of barren lands to built-up areas has led to an expansion of areas with moderate, cold, and very cold temperature classes. Forecast results indicate that the physical expansion of areas with cold and very cold temperature classes in Mashhad and Yazd will continue. This trend is visibly observable on the predicted temperature classification map for Yazd. In northern Kerman, some areas with very low and low temperature classes in 1990 have been converted to high and very high temperature classes due to the transformation of green spaces into barren lands. The very low and low temperature class areas in the outskirts of Kerman in 1990 have moved into the urban area by 2024.

Figure 10 shows the area of temperature classes for the studied cities in the years 1990, 2024, and 2058. In the city of Babol, the area of high and very high temperature classes was 7.09 km² and 7.98 km², respectively, in 1990, increasing to 11.11 km² and 15.44 km² by 2024. The total area of low and very low temperature classes decreased from 40.25 km² in 1990 to 28.61 km² in 2024. Forecasts indicate that from 2024 to 2058, the area of very high temperature classes will increase by 43%, while the area of very low temperature classes will decrease by 27%. In Rasht, the area of very hot temperature classes increased from 13.8 km² in 1990 to 23.6 km² in 2024, and is expected to reach 35.3 km² by 2058. The area of very cold temperature classes was 4.2 km² in 1990, decreased to 0.05 km² in 2024, and is projected to further decrease to 0.02 km² by 2058. For Mashhad, the total area of hot and very hot temperature classes decreased by 69.4 km² from 1990 to 2024, with an expected additional decrease of 3.5 km² from 2024 to 2058. Conversely, the total area of cold and very cold temperature classes increased by 42.1 km² from 1990 to 2024, with an anticipated increase of 23.1 km² in the future. In Yazd, the area of hot and very hot temperature classes significantly decreased from 46.5 km² in 1990 to 12.2 km² in 2024, with a further decrease expected of 2.3 km² by 2058. The area of cold and very cold temperature classes was 28.3 km² in 1990, increased to 72.8 km² in 2024, and is expected to reach 91.3 km² by 2058. The area of very low temperature classes increased from 29.74 km² in 1990 to 67.59 km² in 2024, with expectations to reach 110.57 km² by 2058.

4.6. DSUHII Analysis

The DSUHII in the studied cities for the years 1990, 2024, and 2058 is shown in Figure 11. The estimated values for the DSUHII in the city of Rasht (Babol) are 2.9 °C (2.5 °C) in 1990, 5.5 °C (5.4 °C) in 2024, and 7.6 °C (7.8 °C) in 2058. The DSUHII in the cities of Babol and Rasht increased by 2.9 °C and 2.6 °C, respectively, from 1990 to 2024. Forecasts indicate that these values will increase by 2.4 °C and 2.1 °C, respectively, from 2024 to 2058. In contrast to Babo and Rasht, the trend in the DSUHII for the cities of Mashhad and Yazd is decreasing. The DSUHII in Mashhad, Yazd, and Kerman was −1.1 °C, −2.5 °C, and −2.8 °C, respectively, in 1990, which changed to −1.5 °C, −3.3 °C, and −5.1 °C in 2024. In these cities, there is an occurrence of a cold island effect. The results show that the DSUHII in Mashhad, Yazd, and Kerman decreased by 0.4 °C, 1.5 °C, and 2.3 °C, respectively, over the past 32 years. Forecasts indicate that the DSUHII in Mashhad, Yazd, and Kerman will reach −1.9 °C, −6.4 °C, and −7.1 °C, respectively, by 2058. The reduction in the DSUHII is less pronounced in Mashhad compared to Yazd and Kerman.

5. Discussions

In cities located in the dry–hot climate, the DSUHII during the day is typically negative, resulting in a phenomenon known as cool island [58]. This phenomenon occurs due to the specific characteristics of these climates, such as low humidity, low evaporation, and the open surfaces of barren lands compared to urban areas. In such conditions, the land surface rapidly absorbs heat and loses it at night, leading to a decrease in the city’s surface temperature compared to areas outside the city [60]. In contrast, in cities located in humid climates, the DSUHII is positive during the day [23,61]. This situation is due to the high humidity and high evaporation processes in the suburban areas, which lower the surface temperature in these regions. In these conditions, the land surface absorbs a lot of heat, and due to high evaporation, the surface temperature within the city is higher than in the surrounding areas, leading to an increase in urban temperature and a stronger DSUHII effect [62].

Our findings show that changes in the DSUHII in different years are influenced by climatic and geographical factors. In cities with a humid climate like Babol and Rasht, the reduction of green spaces and the increase in built-up areas have led to an increase in DSUHII and surface temperature. In contrast, in areas with a hot and dry climate like Yazd and Kerman, changes in land cover have focused on the conversion of barren lands into built-up areas, which has led to a decrease in the DSUHII. These differences are primarily due to the specific climatic characteristics and land-cover types in these areas. In humid climates, the increase in humidity and shading from vegetation helps reduce temperature, and the degradation of these areas causes an increase in DSUHII [36,63]. However, in drier regions, the lack of vegetation and green spaces leads to higher temperatures and a stronger DSUHII. The expansion of built-up areas in these regions, however, leads to a decrease in surface temperature and DSUHII. Although the use of class-based land-cover averages provides valuable insights into DSUHII patterns, it is important to acknowledge that this approach simplifies the complex spatial variability of surface temperatures. As a result, some finer-scale thermal differences within land-cover classes may not be fully captured.

One of the main reasons for the difference in surface temperature between barren and built-up areas is the moisture content of the land. Barren lands typically have lower moisture levels, absorb more solar radiation, and reflect less, which converts more energy into heat [58]. In contrast, in built-up areas, hard surfaces and construction materials like asphalt and concrete, which can absorb heat, may absorb less heat than barren lands due to their porosity and physical properties. Barren lands generally have a smooth, dark surface that can absorb solar radiation and tend to retain more heat. While surfaces such as asphalt and concrete may absorb heat, due to differences in structure and heat transfer, their temperatures may be lower than barren lands. Barren lands usually lack vegetation, which is a key factor in reducing temperature through the processes of evapotranspiration, especially on hot days [64]. In this context, urban and built-up areas typically have trees and plants that carry out this evaporative process, keeping temperatures lower [65]. In urban areas, various constructions like buildings, roads, parks, and other features can play a significant role in heat distribution. For example, buildings may provide shading and prevent direct sunlight from reaching the ground. Additionally, some construction materials may absorb heat and reflect it at night, whereas in barren areas, this heat storage process occurs more widely and without obstacles. In some regions, barren lands may experience severe dryness, hot winds, and intense solar radiation due to climatic changes and desertification, leading to higher surface temperatures compared to built-up or urban areas [66]. Therefore, barren lands, especially in deserts and hot conditions, tend to have higher temperatures than urban areas with vegetation and built infrastructure [64,67].

Geographical factors also have a significant impact on the DSUHII. For instance, proximity to water or topography can influence the intensity of surface temperature changes and DSUHII. In areas closer to water resources, a higher humidity and reduced temperatures are observed, while in drier areas, the temperature rises significantly.

Limitations and Future Recommendations

Despite the significant findings of this study, some limitations need to be addressed. One limitation is the use of satellite data for only one season of the year. This could overlook seasonal effects and limit the scope of the results obtained. Additionally, using only a few temporal data points may reduce the accuracy of future predictions. To better analyze long-term trends and changes, more data across different seasons and over time is needed. Furthermore, the models used in this study may require improvements and the use of more advanced algorithms to enhance the accuracy of the analyzes. For future research in this area, it is recommended that multi-seasonal data be utilized to better understand seasonal variations and their impact on DSUHII. Moreover, using machine learning-based models such as artificial/convolutional neural networks could improve the accuracy of predictions. Furthermore, conducting comparative studies across different urban regions and climates could help better understand the effects of land-use changes on surface temperatures.

It is important to acknowledge that the use of only two time points (1990 and 2024) for modeling future changes in LST and land cover introduces a certain degree of uncertainty in the results. The selection of these years was intended to represent long-term urban expansion trends, capturing the broader patterns of land-use transitions and their associated impacts on LST. However, the limited temporal scope may not fully account for all fluctuations or dynamic processes influencing urban development and surface temperature. The CA–Markov model used in this study provides useful insights into the general trajectories of land-use changes, but future research should consider incorporating additional historical time points and exploring more dynamic modeling techniques to improve the accuracy and robustness of predictions. Despite these limitations, the results offer valuable information on potential future trends in DSUHII across different climatic contexts, highlighting the need for more comprehensive data in future studies to address the inherent uncertainties. Also, a key limitation of this study is the lack of the integration of socioeconomic, regulatory, and market drivers in the CA–Markov model. While the physical and infrastructural variables used provide important insights, they do not fully capture the complexity of urban dynamics. Future research should explore hybrid approaches or agent-based models that incorporate these additional dimensions to improve the explanatory power and predictive accuracy of urban growth simulations.

One of the main limitations of this study is the reliance on class-based land-cover averages, which may reduce the precision of DSUHII estimations in areas with high spatial heterogeneity. Future research is encouraged to explore pixel-level calculations or the use of reference zones to enhance the accuracy and detail of DSUHII assessments. Accurately predicting LST and the DSUHII requires considering a comprehensive set of natural and anthropogenic factors that simultaneously influence surface temperature and the distribution of urban heat. In addition to LULC changes, which serve as a practical proxy for reflecting surface biophysical modifications, dynamic biophysical properties such as albedo, soil moisture, vegetation cover, and building materials play significant roles in regulating surface energy balance. Furthermore, human activities, including building energy consumption, vehicular traffic, air conditioning operations, industrial activities, and air pollution, act as major sources of anthropogenic heat emissions and strongly impact DSUHII. Topography also contributes to surface warming by affecting the distribution of solar radiation and airflow patterns. Given the complexity of the interactions among these factors and the challenges of independently predicting biophysical changes under variable climatic conditions, future research would benefit from employing multivariate models that integrate LULC characteristics, surface biophysical indicators such as albedo, vegetation cover, surface moisture, and impervious surface area (ISA), along with climatic parameters and anthropogenic effects, at the pixel level and within reference buffer zones around urban areas. This approach could lead to more accurate and realistic analyzes of urban thermal dynamics.

In this study, we focused on five Iranian cities representing a variety of climate types, including humid subtropical, semi-arid, and hot/cold arid climates. While this selection enhances the internal comparative validity of the study, it is important to recognize that the extrapolation of these findings to regions with different climate types, such as tropical or temperate oceanic climates, should be done with caution. The climatic and geographic variability across regions can significantly influence the outcomes of urban expansion studies. We acknowledge that further research should expand this approach by considering a wider range of climatic conditions and urban contexts to provide more comprehensive insights into the thermal impacts of urban growth across various climate zones.

Additionally, the role of urban policies, construction regulations, and green space management strategies in reducing DSUHII should be explored in greater depth in future research. This study provides several practical insights for policymakers and urban planners. Specifically, in cities facing increasing DSUHII, expanding greenery and using vegetation can play a key role in reducing surface temperatures. Furthermore, in cities with dry climates, the installation of shade structures, planting tall trees, and using drought-resistant vegetation can help mitigate DSUHII [68]. Furthermore, the use of shiny stone facades, e.g., granite, due to their high albedo effect, which amplifies the DSUHII effect, in favor of transparent surfaces such as glass facades, should be investigated. Sustainable urban development policies that focus on reducing energy demand by improving housing insulation and improving environmental conditions can reduce DSUHII and enhance urban quality of life [69]. It is also important to develop simulation models to forecast the pick effects of DSUHII that are fatal to certain socioeconomic groups [70].

6. Conclusions

This study investigates the spatial and temporal changes of the DSUHII in some climate regime diverse cities of Iran. The results showed that the rapid expansion of built-up areas and the reduction of green spaces in these regions have had a direct impact on the intensity of urban heat islands. In cities with humid climates such as Rasht and Babol, a significant increase in urban areas and the decrease in green spaces have led to a remarkable rise in DSUHII. In contrast, in cities with hot and dry climates like Yazd and Kerman, although urban development has been significant, the DSUHII has decreased, and these cities have transformed into “relatively cooler areas” due to the slower rate of urban warming compared to the surrounding rural areas. While the urban areas are indeed warming, their temperature increase is less pronounced than that of the rural surroundings, leading to a relative decrease in the surface temperature difference between the two. Predicted changes in land cover and surface temperature indicate concerning outcomes for Rasht and Babol. According to the findings, the expansion of built-up land in these two cities will continue at a faster rate in the coming years, leading to increased human activities, a rise in high and very high temperature zones, and consequently a rise in DSUHII in different parts of Rasht and Babol. These two cities are located in a humid climate, where the land cover around the cities consists of agricultural land and green spaces, which have low surface temperatures due to their greenery and high humidity. Urban expansion is turning these areas into built-up and impervious land, causing a significant increase in surface temperature. However, barren lands on the outskirts of Mashhad, Yazd, and Kerman have higher surface temperatures than the built-up areas. As a result, replacing these barren lands with built-up areas due to the physical growth of cities leads to a decrease in surface temperature. Although converting green spaces to built-up areas in cities located in hot and dry climates significantly increases surface temperature, causing thermal impacts, the overall results indicate a reduction in the area of high and very high temperature zones and a decrease in DSUHII in these regions. Therefore, for these climates, the physical growth of cities does not face serious thermal concerns or has fewer thermal negative effects compared to cities in humid climates. This result is particularly noticeable for Yazd and Kerman, located in hot and dry climates. However, the physical growth of cities in any climate can have various negative effects, including environmental damage, which cannot be overlooked and requires further study. The findings emphasize the necessity of implementing green space-based solutions and climate modifications, especially in cities with humid climates. It is recommended to promote novel urban planning and regulatory frameworks, e.g., urban regeneration and urban greenery, green roofs, and a temporal shift of public and private services to mitigate the cascading effects of DSUHII.