1. Introduction
Urbanization has become a major driver of environmental change, significantly altering local climates through the formation of urban heat islands (SUHIs)—a phenomenon where urban areas exhibit higher temperatures than their rural surroundings, and which has become a central focus in urban climate research [
1] This temperature difference is largely attributed to human activities, altered land surfaces, and the modification of natural landscapes to built environments [
2,
3]. The primary factors contributing to SUHIs include the replacement of natural vegetation with buildings, roads, and other infrastructure, which alters the thermal properties and radiative balance of the surface [
4]. Impermeable surfaces such as asphalt and concrete absorb and retain heat more efficiently than natural landscapes, leading to elevated temperatures in urban areas [
5]. Additionally, human activities such as transportation, industrial processes, and air conditioning contribute to increased anthropogenic heat production, further intensifying the SUHI effect [
6,
7,
8]. Urban infrastructure, such as vehicles and buildings, continuously emits heat, especially during peak usage times, contributing to the persistent temperature increase in urban centers [
9]. The modification of natural landscapes into built environments not only affects surface temperatures but also impacts the local and regional climate. A reduction in vegetation cover decreases the natural cooling effect provided by evapotranspiration, while the concentration of buildings can obstruct wind flow, reducing natural ventilation and leading to stagnant air conditions [
10]. Additionally, while green areas tend to be more resilient to temperature fluctuations due to their natural cooling effects and evapotranspiration processes, urban areas are significantly more vulnerable to heatwaves [
11]. Urban canyons, formed by tall buildings, can trap heat and limit airflow, exacerbating the SUHI effect [
12]. The degree of urbanization in a region is also critically important, as the SUHI effect manifests with significant variations across urban, suburban, and rural areas, with urban regions experiencing more pronounced temperature increases due to dense infrastructure and human activities [
13]. These changes can exacerbate a range of environmental and health issues, including increased energy consumption for cooling, elevated air pollution levels, and increased incidences of heat-related illnesses and mortality [
14,
15,
16].
The accurate classification of land use and land cover (LULC) is essential for understanding the spatial and temporal dynamics of urbanization, vegetation, and land management. Among the various machine learning techniques employed in remote sensing, the random forest (RF) algorithm has emerged as a highly effective and robust classifier due to its ability to handle high-dimensional data, reduce overfitting, and maintain strong predictive performance even with noisy or correlated inputs [
17]. Recent studies have successfully implemented RF for LULC classification using multi-temporal Landsat imagery, often within cloud-based platforms like Google Earth Engine (GEE), to monitor long-term environmental and urban changes [
18,
19]. RF’s ensemble learning approach, which aggregates the results of multiple decision trees, enhances classification accuracy across diverse landscapes and land cover types, including built-up areas, vegetation, and water bodies [
20]. In addition, integrating spectral indices (e.g., NDVI, NDBI), auxiliary data such as digital elevation models (DEM), and pan-sharpened imagery has been shown to further improve classification outcomes, achieving overall accuracies exceeding 90% in several case studies.
Recent studies have highlighted the spatial and temporal dynamics of SUHIs, revealing that the intensity of SUHIs can vary significantly across different urban areas and seasons [
21,
22,
23]. For instance, summer months often exhibit the most pronounced SUHI effects due to higher solar radiation and longer daylight hours [
24]. The application of remote sensing technologies has been particularly effective in monitoring and analyzing SUHI effects. For instance, satellite-derived land surface temperatures (LSTs) have been widely used to assess the spatial extent and intensity of SUHIs [
25,
26]. Advanced techniques, such as the use of thermal infrared imagery from satellites such as Landsat and MODIS, enable precise mapping of temperature variations and SUHI hotspots [
27]. Furthermore, the integration of remote sensing data with the geographical information system (GIS) allows for the comprehensive analysis of SUHI patterns in relation to land use and land cover changes [
28]. This integration facilitates the identification of critical areas that require mitigation efforts and supports the development of targeted strategies to combat SUHI effects [
29]. In the literature, numerous studies have focused on analyzing land use and land cover (LULC) changes using machine learning approaches, as well as calculating various temperature and vegetation-related indices. Ref. [
30] reported that between 2013 and 2020 in Greater Arba Minch, Ethiopia, LST was significantly correlated with changes in LULC, the normalized difference vegetation index (NDVI), the normalized difference water index (NDWI), and the normalized difference built-up index (NDBI), with notable increases in settlement and barren land, decreases in forest and water bodies, and significant positive changes in NDBI values. Using Landsat 8 median composite satellite images and the Google Earth Engine for Aurangabad, India, between 2015 and 2020, Ref. [
31] found that urban climate change due to land cover changes resulted in a 2 °C increase in LST, a decrease in vegetation cover and NDVI values, and an increase in built-up land, wasteland, and water bodies. In another study, Ref. [
32] reported that between 2016 and 2021, urbanization and population growth in Yogyakarta led to the rise of slum settlements with dense, low-quality buildings and misaligned land use, resulting in high environmental criticality in downtown areas due to dense built-up land, low vegetation, and high surface temperatures, with slums concentrated along major waterways such as the Gadjah Wong, Code, and Winongo Rivers, as determined using LST, built-up index (BU), and modified normalized difference water index (MNDWI) to map the environmental criticality index (ECI). Ref. [
33] used Google Earth Engine (GEE) and median satellite products to study vegetation coverage on Zhoushan Island from 1985 to 2022, and reported that the average NDVI decreased from 0.53 to 0.46, low vegetation areas increased from 28.84 km
2 to 67.29 km
2, and extremely high vegetation areas decreased from 197.96 km
2 to 146.32 km
2, with significant NDVI clusters and hot spots in the island’s interior and cold spots along the coast. Finally, Ref. [
34] found that using Landsat satellite images and the support vector machine (SVM) method to map LST, urban thermal field variance index (UTFVI), and SUHI index from 1995 to 2016 in Ahvaz, Iran, revealed the highest temperatures in bareland (42.93 °C) and residential areas (40.06 °C) in 2017, a 50% reduction in green spaces from 14% to 7%, and the worst UTFVI in the hottest locations, emphasizing the need for urban planning to mitigate SUHI intensification.
In Turkiye, Ref. [
35] emphasized the importance of assessing LULC changes in urban planning, showing that increases in urban built-up areas and agricultural lands in Sivas from 1989 to 2015 led to higher LST and positive SUHI intensity, and remote sensing (RS) and GIS were used to analyze these changes. The study by [
36] evaluated the impact of the surface urban heat island (SUHI) effect on Istanbul’s coastal zone using Landsat TM/ETM+ data from 1984 to 2011, highlighting that urbanization increases SSUHI intensity and emphasizing the importance of incorporating coastal management strategies to address the rising rate of impervious surfaces (e.g., roads and buildings, contribute significantly to heat storage in urban environments) and their effect on land surface temperature. Ref. [
37] reported that between 2013 and 2022 in Kayseri, urban heat island effects strongly negatively correlated with the NDVI and strongly positively correlated with the NDBI, highlighting the need for strategic urban planning and SUHI mitigation. Finally, Ref. [
38] examined the SUHI effect in Samsun, revailing a significant increase in SUHI intensity along the coastline over 20 years using the UTFVI with LST data from 2000 ETM+ and 2020 OLI/TIRS Landsat images.
The aim of this study is to examine both thermal and vegetative indices such as BT, LST, NDVI, NDBI, BUI, ECI, SUHI, and UTFVI together, as reported in the literature, and to investigate their changes not only in urban areas but also in green spaces, water bodies, and barren lands. In this context, the effects of various land use and cover types on temperature, vegetation, and structural characteristics will be analyzed in detail, revealing the role of changes in different land types on the urban heat island effect.
Given the intensifying SUHI effect in coastal urban areas and the increasing availability of high-resolution satellite data, this study aims to assess the spatial and temporal patterns of SUHI in Samsun using multiple thermal and ecological indices. We seek to answer: (1) How have different land cover types influenced LST and SUHI intensity over the past decade? (2) What are the spatial correlations between vegetation loss, built-up expansion, and thermal stress? (3) How can remote sensing tools like Google Earth Engine support monitoring and mitigation of urban climate risks in mid-sized cities? By addressing these questions, the study offers methodological and empirical contributions to urban climate research.
5. Conclusions
This study highlights the significant role of urban expansion and increased impervious surfaces in intensifying surface urban heat island (SUHI) effects, resulting in elevated land surface temperatures (LSTs) and heightened environmental stress, particularly in densely developed areas. Conversely, green spaces emerged as vital thermal regulators, effectively mitigating surface temperatures and enhancing environmental quality—evident from the stable or improved values of vegetative indices such as the NDVI and BUI.
The SUHI intensity increase observed in this study (from 0.47 to 1.33) aligns with similar trends documented in other cities (e.g., Kayseri, Istanbul, Tehran), reinforcing the representativeness and external validity of our findings. Moreover, the spatial analysis revealed up to 12 °C differences in LST between industrial zones and large green spaces, emphasizing the importance of urban vegetation in regulating local microclimates.
Methodologically, the study demonstrates the utility of integrating Google Earth Engine (GEE) with the random forest (RF) algorithm and multi-index composites to monitor urban thermal dynamics over time. Although classification accuracy was generally high, confusion was observed between bareland and urban areas, and between vegetated zones and shallow lakes with similar spectral signatures. Future studies could address this by expanding the number and distribution of training points, and by incorporating additional spectral or texture features to improve class separability.
Looking ahead, future research may incorporate nighttime thermal data (e.g., from MODIS or ECOSTRESS) to better capture diurnal variations in SUHI patterns. Additionally, integrating anthropogenic heat flux data and urban activity metrics (e.g., traffic density, energy consumption) could further enhance the explanatory power of SUHI analyses. Finally, applying this framework to different city sizes and climate zones would provide broader insights into the scalability and adaptability of remote sensing-based UHI-monitoring approaches.