Geospatial Analysis of Urban Warming: A Remote Sensing and GIS-Based Investigation of Winter Land Surface Temperature and Biophysical Composition in Rajshahi City, Bangladesh

Abstract

This study investigates urban warming in Rajshahi City, Bangladesh, by examining changes in land surface temperature (LST) from 1990 to 2023 and exploring its relationship with key biophysical factors. LST was derived from Landsat thermal imagery, and both spatial and temporal variations were analyzed using Geographic Information Systems (GIS). Key biophysical indices, including Normalized Difference Vegetation Index (NDVI), Normalized Difference Built-up Index (NDBI), Normalized Difference Water Index (NDWI), Normalized Difference Moisture Index (NDMI), and Normalized Difference Bareness Soil Index (NDBSI), were calculated using corresponding Landsat satellite sensors, and they evaluated the impact of LULC types (vegetation, water, soil, and built-up areas) on thermal variations. LULC was derived following the Support Vector Machine classification technique. The Urban Thermal Field Variance Index (UTFVI) was employed to assess surface urban heat island (SUHI) effects, warming conditions, ecological stress, and thermal comfort zones. Spatial trend and hotspot analyses of LST change were performed using spatial trend analysis and the Getis-Ord Gi* statistic, respectively. Linear regression analysis examined the relationship between LST and biophysical indices. Results show that winter mean LST increased by 2.66 °C during the 33-year period, with maximum LST rising by 4.29 °C. The most significant warming occurred in central-northern, central-western, and south-eastern zones. The rise in LST and the growing intensity of SUHI effects are largely due to urban growth, especially where green spaces and water bodies have been replaced by impervious surfaces. Hotspot analysis identified clusters of high-temperature zones, while UTFVI analysis confirmed a marked expansion of strong heat island conditions, especially in central urban areas. Linear regression results showed notable links between LST and key biophysical variables, where higher LST values were commonly linked to greater built-up density and declines in vegetation cover and surface water. Overall, the results highlight the need for better urban planning approaches such as increasing green cover, using permeable materials, and adopting strategies that can adapt to climate impacts. This study presents a framework for analyzing urban climate dynamics that can be adapted to other rapidly growing cities, aiding efforts to promote sustainable development and build urban resilience.

1. Introduction

Air and surface temperature patterns in urban areas often differ from those of rural settings, largely due to changing patterns in land use and land cover (LULC). Such variations play a crucial role in influencing the urban climate. As cities expand, the natural landscape is replaced with roads, buildings, and other impervious surfaces, altering the local climate and energy balance [1,2,3]. The dominance of these anthropogenic surfaces alters energy balance and local climate, significantly impacting both land surface temperature and air temperature. The urban climate is shaped by the built environment of the city and differs notably from surrounding rural areas due to surface modifications and human activities. Urbanization, coupled with large-scale alterations in surface and atmospheric properties, contributes to a measurable rise in urban temperatures. Recent studies suggest that many cities are experiencing warming at rates up to twice or more the global average (0.20 °C per decade), often due to reduced vegetation and increased energy use [4,5,6]. For example, Phoenix (USA) has seen a temperature rise of about 0.86 °C per decade, while Shanghai (China) warms at roughly 0.37 °C per decade due to rapid urbanization. In western Sydney, the rate is around 0.65 °C per decade. On a broader scale, megacities across Asia have experienced average warming rates of 0.69 °C per decade [4,5,6,7,8]. These trends highlight the intensifying impact of urban heat driven by reduced vegetation and increased energy use. During winter, the urban heat island effect often intensifies as vegetation cover decreases and human-generated heat becomes a larger contributor to local warming. Cities such as Beijing, New York, and London have documented temperature differences of 3–5 °C compared to nearby rural areas [4,5,6,7]. In the United States, urban air temperatures have been found to exceed those of nearby natural land cover by up to 5.6 °C, with New York showing differences above 5 °C [8]. These differences in surface temperature, known as the surface urban heat island (SUHI) effect, can influence energy demand, air pollution, and thermal stress in city environments [9,10,11,12,13,14,15]. LST, measured by satellites or ground sensors, is a key metric used to analyze heat patterns across urban areas. The measure indicates the temperature of the land’s exterior layer, covering vegetation, soil, urban structures, and various surface elements [16,17]. Remote sensing technologies enable LST estimation by capturing infrared radiation emitted by surface materials [18,19], making it an indispensable tool for analyzing urban thermal dynamics and supporting climate-resilient planning.

Monitoring LST is essential not only for understanding urban heat but also for modeling broader climate-related processes, especially as they relate to land changes, particularly concerning warming trends and the greenhouse effect. Understanding and tracking urban climate change are essential for comprehending the broader urban environment, given the intricate spatial and temporal variations inherent within urban areas [20]. Recent research has placed a significant emphasis on observing and monitoring LST and SUHI effects across diverse landscapes, owing to their profound impacts on local weather and climate patterns, particularly within urban settings [21,22,23,24,25]. However, accessing comprehensive historical climate data, including temperature, precipitation, and humidity, remains challenging, especially in developing countries like Bangladesh, where large-scale and time series observational datasets are scarce. Given the limitations in observed data availability, LST emerges as a potent proxy for monitoring urban climate change, as historical spatial data of LST can be readily obtained from thermal sensors aboard satellites such as Landsat, the Moderate Resolution Imaging Spectroradiometer (MODIS), the National Oceanic and Atmospheric Administration (NOAA), and the Television Infrared Observation Satellite (TIROS) [26,27]. Sensors like the Landsat Thematic Mapper (TM), the Enhanced Thematic Mapper Plus (ETM+), and the Operational Land Imager (OLI) are widely utilized for exploring LST dynamics and assessing their associations with other urban components [28]. The significance of LST transcends mere climate monitoring; it also plays a pivotal role in regulating the energy and water balance within a given area, thus exerting influence over its local climate. Advances in remote sensing technologies and analytical methodologies enable the effective calculation of LST, facilitating its monitoring over extended periods [29]. Consequently, continuous monitoring of LST through remote sensing offers crucial insights into urban climate change conditions and the surface physical properties that drive various environmental processes [30,31].

Urban expansion induces variations in LST, resulting in elevated temperatures in specific locations known as the SUHI effect. LST dynamics have been extensively scrutinized across numerous cities worldwide, with a wealth of studies documenting these phenomena over recent decades [21,32,33,34,35,36,37,38,39,40,41]. Pioneering investigations into SUHI, utilizing satellite-derived LST data, were conducted by the authors of [42] to examine temperature patterns along the mid-Atlantic coast. Subsequently, Refs. [43,44] utilized NOAA data to identify urban heat island (UHI) effects. Since then, a plethora of studies have delved into SUHI and LST patterns, leveraging data collected from a variety of remote sensing sources, such as satellite systems, aerial platforms, and ground-based monitoring devices [38,45,46,47]. In contemporary research, Landsat sensor data have emerged as a cornerstone for analyzing SUHI patterns, urban LST dynamics, and elucidating the interplay between LST and LULC [41,46]. Similarly, studies focusing on spatio-temporal analyses of LST using remote sensing thermal data in Bangladeshi cities have surfaced, predominantly utilizing Landsat satellite data to examine LULC changes and LST patterns, given the significant transformations in urban landscapes [25,40,47,48,49,50,51,52,53,54]. Despite these efforts, studies investigating LST and SUHI as pivotal climate parameters for monitoring urban climate change and discerning the influence of biophysical parameters on such changes, particularly in Bangladeshi cities, remain sparse [25,47,48]. The escalating impact of urban heat has spurred heightened interest among researchers and urban planners, fueled by extensive infrastructural and economic growth in urban areas. However, the scientific understanding and comprehensive analysis of urban thermal phenomena and related climate changes remain limited [55]. In particular, studies focusing on LST variations and warming dynamics during the winter season are scarce. Yet, investigating winter warming is critical for gaining a holistic understanding of urban ecological processes and environmental conditions. The rise in urban winter temperatures has become an increasingly pressing issue in the context of climate change, largely driven by the UHI effect. In densely built environments, impervious surfaces absorb and retain heat, causing winter temperatures in cities to be noticeably higher than in nearby rural areas. This warming trend not only alters local climatic conditions but also disrupts ecological stability within urban ecosystems. One notable consequence is the shortening of the winter season, a pattern now evident in many tropical cities. Warmer winters also create favorable conditions for the year-round survival of pests and the spread of disease vectors, while posing challenges to urban vegetation and wildlife. In the broader framework of climate change, this phenomenon underscores the critical need for climate-responsive urban planning, the integration of green infrastructure, and the promotion of adaptive strategies to reduce heat stress and enhance the resilience and livability of cities.

Therefore, this study underscores the importance of LST and SUHI as critical phenomena for monitoring urban climate change, particularly within the context of urban warming scenarios in winter and in response to biophysical compositions, focusing on the Rajshahi metropolis. Particularly, the aim is to analyze LST alongside biophysical components for urban warming change and SUHI dynamics, primarily utilizing satellite remotely sensed data. Also, this study seeks to enhance the comprehension of urban warming processes and research findings-based planning for long-term urban sustainability by highlighting these critical phenomena. This investigation focuses on achieving the following specific aims: (i) examine the nature of urban warming through changes in LST from 1990 to 2023; (ii) identify the spatio-temporal patterns of LST within the city area; (iii) analyze the spatial relationship between LST and selected biophysical indices over time; and (iv) detect spatial trends and hotspots of LST changes. These objectives were achieved by quantifying LST, analyzing warming scenarios, investigating LULC changes, assessing multiple biophysical indices, and detecting spatial trends and hotspots. The uniqueness of the present study lies in its inclusive exploration of the spatio-temporal patterns of urban warming and changing patterns in Rajshahi City, focusing on recent LST dynamics and related phenomena, which have not been previously studied in the area. By analyzing urban warming through changes in LST and finding its relationship with important biophysical urban components, this study indicates the significance of proper planning to maintain adequate green spaces and surface water features and to adopt climate-adaptive urban strategies. The approach and analytical framework of this study are transferable to other urban areas in developing countries to support effective urban planning aimed at mitigating the adverse impacts of warming and promoting sustainable urban development.

2. Data Used and Methods

2.1. Study Area

Rajshahi City Corporation (RCC), situated in the north-west region of Bangladesh, is designated as the focal area of the present research endeavor. Encompassing an area of 48 km², the city is strategically positioned between the latitude and longitude 24°12′ to 24°42′ N and 88°15′ to 88°50′ E, respectively [56] [Figure 1]. Because of vast urban growth and urbanization during the last few decades and to know the LST pattern, this city was purposely selected. Serving as a divisional headquarters, RCC comprises 30 administrative wards. The city is nestled along the banks of the Padma River amidst verdant landscapes and fertile plains. The climate of the city is tropical monsoon, featuring an average annual temperature of 24.5 °C and rainfall measuring 1448 mm per year. In winter, temperature ranges from 9 °C to 30 °C with almost no rainfall in this season. Furthermore, the region experiences significant variability in annual rainfall patterns, a factor that can potentially exacerbate temperature fluctuations in the area [57,58,59]. It is important to note that the study area is not prone to flooding, and the analysis was conducted for the month of November, which typically experiences very little to no rainfall. Therefore, the influence of precipitation or flood-related effects on LST during the study period is considered negligible. As a divisional headquarters, Rajshahi has undergone accelerated urban growth in recent decades, becoming home to approximately one million inhabitants. The population has exhibited an annual growth rate of 2.88% from 1990 to 2022, as reported by Bangladesh Bureau of Statistics [56]. This urban development has noticeably altered the city’s landscape, with open and agricultural lands being converted into built-up areas, thereby impacting the urban climate. The unchecked growth and unrestricted development of built-up areas, along with the reduction in urban green spaces, water bodies, fallow lands, and croplands, have adversely affected the city’s environmental quality and overall resilience, particularly leading to warming trends in urban areas [59]. Consequently, detailed information on urban climate dynamics, particularly surface temperature variations and associated biophysical compositions, is indispensable for informed city development and sustainable expansion efforts. The present study is meticulously designed to analyze the dynamics of land surface temperature, the level of urban warming, and its geospatial patterns over the period spanning 1990 to 2023 in the Rajshahi City Corporation of Bangladesh, aiming to shed light on the evolving urban climate dynamics and contribute to informed decision-making processes for the city’s sustainable future.

2.2. Data Used

To extract and analyze the LST of the study area, Landsat TM, ETM+ and OLI sensors’ multispectral images were acquired for the years 1990, 2000, 2010, and 2023. The details of the remote sensing datasets are presented in

Table 1. Details of satellite datasets.

Sensor	Sensor	Data Level	Date	Band Used	Radiometric Resolution	Spatial Resolution
Landsat 5	TM	Level 1	14 November 1990 and 5 November 2010	Blue (b1) Green (b2) Red (b3) NIR (b4) SWIR-1 (b5) Thermal (b6) SWIR-2 (b7)	0.45–0.52 µm 0.52–0.60 µm 0.63–0.69 µm 0.76–0.90 µm 1.55–1.75 µm 10.40–12.50 µm 2.08–2.35 µm	30 30 30 30 30 120 (30) * 30
Landsat 7	ETM+	Level 1	17 November 2000	Blue (b1) Green (b2) Red (b3) NIR (b4) SWIR-1 (b5) Thermal (b6) SWIR-2 (b7)	0.45–0.52 µm 0.52–0.60 µm 0.63–0.69 µm 0.76–0.90 µm 1.55–1.75 µm 10.40–12.50 µm 2.08–2.35 µm	30 30 30 30 30 120 (30) * 30
Landsat 8	OLI	Level 1	3 November 2023	blue (b2) Green (b3) Red (b4) NIR (b5) SWIR-1 (b6) SWIR-2 (b7) TIRS-1 (b10)	0.45–0.51 µm 0.53–0.59 µm 0.64–0.67 µm 0.85–0.88 µm 1.57–1.65 µm 2.11–2.29 µm 10.6–11.19 µm	30 30 30 30 30 30 100 (30) *

* Resampled to 30 m.

. All data were sourced from the United States Geological Survey (USGS) and downloaded using the GloVis portal (http://glovis.usgs.gov (accessed on 10 September 2024)). Additionally, other spatial data, including city boundaries, ward boundaries, and field data, were obtained from the Rajshahi City Corporation office, the Center for Environment and Geographic Information Services (CEGIS), and the Local Government Engineering Department (LGED), collected using Global Positioning System (GPS) and observation techniques for the study. For spatial analysis, datasets were resampled to a 30 m resolution using ArcGIS 10.6.1.

2.3. Methods

2.3.1. Pre-Processing of Landsat Data

Geometrically corrected Level-1 Landsat imagery was utilized and further processed to enhance the accuracy of the analysis. As the initial step in pre-processing, atmospheric correction was applied to all visible bands using the Dark Object Subtraction (DOS) technique. This image-based method, developed by [60], is extensively applied in remote sensing to mitigate the effects of atmospheric scattering, particularly haze, by estimating and subtracting the path radiance component. However, Albedo, which influences land surface temperature by affecting the amount of solar radiation absorbed or reflected by the surface, is implicitly accounted for through the surface reflectance data obtained using the DOS technique. The DOS method assumes that certain surface features, such as deep-water bodies or shadowed areas, have near-zero reflectance in the visible spectrum. By identifying the darkest pixels in each band, the atmospheric path radiance is estimated and subtracted from all pixels in that band. DOS is favored for its simplicity, ease of application, and independence from in situ atmospheric measurements. It is a widely accepted approach within the geospatial community for correcting light scattering effects [61]. In this study, the DOS module of the ENVI 5.1 image processing software was employed for atmospheric correction. For LST calculations, we followed the methodology outlined in the Landsat User Guide, which recommends using Digital Number (DN) values and converting them to radiance for the radiometric correction of the thermal bands [62]. The detailed procedure is discussed in the following section. The DOS method is not well-suited for thermal bands, and alternative atmospheric correction techniques—such as the Radiative Transfer Equation—require numerous input parameters, making accurate correction of thermal data more complex. Given that the study area features relatively flat terrain and the Landsat imagery was acquired during the dry winter season, typically associated with lower atmospheric water vapor, atmospheric interference in the thermal infrared bands is hence expected to be minimal [63,64]. Therefore, we adopted the methodology recommended by [62], which has been commonly used in earlier studies to estimate LST in diverse geographic settings [25,29,30,31,36,37,38,51,54,65]. In this study, atmospherically corrected imagery was used to derive surface emissivity and other biophysical indices. Incorporating surface emissivity into LST calculations is essential, as it corrects for the radiative transfer between the Earth’s surface and the satellite sensor, thereby enhancing the accuracy of temperature estimates. Further details on the LST computation from thermal data are provided in the following section.

2.3.2. Retrieval of LST from Thermal Landsat TM and ETM+ Images (for 1990, 2000, and 2010)

To compute LST using Landsat TM, ETM+, and OLI data for the years 1990, 2000, 2010, and 2023, the following five steps were employed:

• Step 1: Convert DN (Digital Number) Value to Radiance ($L_{\mathsf{\lambda }}$)

In this step, the digital number (DN) values from the Landsat data were converted into spectral radiance. For Landsat TM and ETM+, spectral radiance was calculated using Equation (1) [62], while for Landsat OLI, Equation (2) [66] was used.

$$L_{\mathsf{\lambda }}={\displaystyle \frac{\left(L_{max \mathsf{\lambda }}-L_{min \mathsf{\lambda }}\right)}{\left(Q_{cal max}-Q_{cal min}\right)}}\times \left(Q_{cal}-Q_{cal min}\right)+L_{min\mathsf{\lambda }}$$

(1)

where $L_{\lambda }$ = spectral radiance (watts/(m² × sr × μm) [watts per square meter per steradian per micrometer], $L_{max \lambda }$ = radiance maximum for the thermal band, $L_{min \lambda }$= radiance minimum for thermal band, $Q_{cal}$= DN value of the thermal band (Band 6 for Landsat TM 5 and ETM⁺), $Q_{cal max}$= the maximum quantized calibrated pixel value of the thermal band, and $Q_{cal min}$= the minimum quantized calibrated pixel value of the thermal band. $L_{min \lambda }$, $Q_{cal max}$, and $Q_{cal min}$ values can be found in the metadata file.

$$L_{\mathsf{\lambda }}={MSF}_{\mathsf{\lambda }}\times Q_{cal}+{ASF}_{\mathsf{\lambda }}-O_i$$

(2)

where $L_{\lambda }$ = spectral radiance (watts/(m² × sr × μm) [Watts per square meter per steradian per micrometer], ${MSF}_{\lambda }$ = radiance multiplicative scaling factor for the band (RADIENCE_MULTI_BAND_10), ${ASF}_{\lambda }$= radiance additive scaling factor for the band (RADIENCE_ADD_BAND_10), $Q_{cal}$= DN value of the thermal band (here Band 10), and $O_i$= correction value (0.29) for Band 10 of Landsat OLI [67,68]. ${MSF}_{\lambda }$ and ${ASF}_{\lambda }$ values can be found in the metadata file.

• Step 2: Convert Radiance to At-sensor Brightness Temperature (°C) (${\mathsf{\tau }}_{\mathit{c}}$)

Here, to change radiance to At-sensor Brightness Temperature in degrees Celsius, Equation (3) was applied [65,68,69,70].

$${\mathsf{\tau }}_c={\displaystyle \frac{k_2}{Ln\left(\frac{k_1}{L_{\mathsf{\lambda }}}+1\right)}}-273.15$$

(3)

where ${\tau }_c$ = At-sensor Brightness Temperature, in °C, $k_1$= pre-launch calibration thermal constants for the band (K1_CONSTANT_BAND_6 for TM and ETM+ and K1_CONSTANT_BAND_10 for OLI, obtained from the metadata), $k_2$= pre-launch calibration thermal constants for the band (K2_CONSTANT_BAND_6 for TM and ETM+ and K2_CONSTANT_BAND_10 for OLI, obtained from the metadata), and $L_{\lambda }$= spectral radiance.

• Step 3: Calculation of NDVI and Proportion of Vegetation ($P_v$) for Emissivity Correction

An emissivity correction is crucial for accurately deriving LST from the remotely sensed data, such as Landsat, as highlighted in the studies [27,65,71]. The calculation of the NDVI holds significance because it enables the determination of vegetation proportion ($P_v$), which is closely linked with NDVI, and subsequently influences surface emissivity ($\mathsf{\epsilon }$), as established by [30,68]. Therefore, NDVI and NDVI-based PV were computed using atmospherically corrected Landsat data (surface reflectance) and Equations (4) and (5), respectively.

$$NDVI={\displaystyle \frac{b_{NIR}-b_{RED}}{b_{NIR}+b_{RED}}}$$

(4)

where $NDVI$ represents the Normalized Difference Vegetation Index; $b_{NIR}$ and $b_{RED}$ denote the surface reflectance of the near-infrared band (Band 4 for TM and ETM+, Band 5 for OLI) and red band (Band 3 for TM and ETM+, Band 4 for OLI), respectively. Afterwards, using the minimum NDVI (${NDVI}_{min}$), and maximum NDVI (${NDVI}_{max}$) values observed across the entire scene, the proportion of vegetation ($P_v$) was calculated using Equation (7) [59,72,73].

$$P_v={\left({\displaystyle \frac{NDVI-{NDVI}_{min}}{{NDVI}_{max}-{NDVI}_{min}}}\right)}^2$$

(5)

• Step 4: Estimation of Emissivity (ε)

Here, surface emissivity is computed utilizing the vegetation proportion ($P_v$) and Equation (6) [68,72]. Surface emissivity represents an intrinsic characteristic of the land, facilitating the conversion of surface heat energy into radiant energy.

$$\mathsf{\epsilon }=0.004\times P_v+0.986$$

(6)

where $\mathsf{\epsilon }$ represents surface emissivity and $P_v$ denotes the proportion of vegetation.

• Step 5: LST in Degrees Celsius ($\mathsf{\theta }$)

Finally, the retrieval of LST, specifically the emissivity-corrected LST (θ), is accomplished by computing Equation (7).

$$\mathsf{\theta }={\displaystyle \frac{{\mathsf{\tau }}_c}{\left[1+\left(\mathsf{\lambda }\times \frac{{\mathsf{\tau }}_c}{\mathsf{\rho }}\right)\times Ln\left(\mathsf{\epsilon }\right)\right]}}$$

(7)

where $\mathsf{\theta }$ = LST in degrees Celsius (°C), ${\tau }_c$= At-sensor Brightness Temperature, in °C, $\lambda$ = wavelength (emitted radiance, $\lambda$= 11.45 for TM/ETM+ and 10.985 for OLI thermal band), $\mathsf{\epsilon }$ = surface emissivity, ρ = 1.438 × 10⁻² m. K = 14,388, since $\rho =h{\displaystyle \frac{c}{\sigma }}$, $h$ is Plank’s constant (6.626 × 10⁻³⁴ J s), $\sigma$ is the Boltzmann Constant (1.38 × 10⁻²³ J/K), and $c$ is the velocity of light (2.988 × 10⁸ m/s) [68].

2.3.3. Urban Thermal Field Variance Index (UTFVI)

To comprehend the incidence of the UHI phenomenon and its impact on urban climate and to evaluate the ecological aspects of UHI zones in the area, the UTFVI was assessed using Equation (8) [74,75]. Subsequently, the UTFVI was categorized into five classes to represent surface urban heat island (SUHI) intensity, heat distribution, ecological evaluation index (EEI), and urban thermal comfort zones in the area, in accordance with

Table 2. UTFVI value range for UHI, ecological evaluation, and thermal comfort.

UTFVI	Urban Heat Island (UHI) Phenomenon/Heat Distribution	Ecological Evaluation Index (EEI)/Thermal Comfort
≤0.005	None/Weak	Excellent/Good
0.005–0.01	Middle	Normal
0.01–0.015	Strong	Bad
0.015–0.02	Stronger	Worse
≥0.02	Strongest	Worst

[76,77,78].

$$UTFVI={\displaystyle \frac{{LST}_x-{LST}_{\overline{x}}}{{LST}_{\overline{x}}}}$$

(8)

where $UTFVI$ represents the Urban Thermal Field Variance Index, ${LST}_x$ is the LST in degrees Celsius (°C), and ${LST}_{\overline{x}}$ is the mean LST in (°C) observed across the entire scene.

2.3.4. LULC and Biophysical Compositions

In this research, LULC data were derived from Landsat satellite imagery through a digital image classification approach. The pre-processing stage involved geometric correction of the images using accurately distributed ground control points (GCPs), ensuring a root mean square error (RMSE) below 0.15. The corrected images were resampled to match the sensor’s pixel resolution, employing the Universal Transverse Mercator (UTM) coordinate system, the nearest neighbor resampling method, and polynomial transformation. Image classification was carried out using the Support Vector Machine (SVM) algorithm, as outlined in references [79,80]. SVM is a supervised machine learning method that enables the system to learn patterns from data autonomously. The classification process emphasizes support vectors—the key training samples located near the decision boundaries of the SVM model—which are essential for identifying the optimal separating hyperplane and disregarding less critical samples [80]. Training and validation sites were created based on the standard SVM classification workflow. The classification employed four LULC categories: vegetation, water bodies, built-up areas, and bare land, integrating field-verified reference data for accuracy. A radial basis function (RBF) kernel was selected for the SVM due to its strong performance in handling complex, nonlinear data distributions [81]. Comprehensive explanations of the mathematical foundations of the SVM and RBF kernel are available in [82,83]. For the LULC classification task, the ‘Train Support Vector Machine Classifier’ tool within ArcGIS 10.6 was used with the prepared training dataset. Model accuracy was assessed by comparing the resulting maps against an independent validation dataset and constructing an error matrix to evaluate classification performance.

Furthermore, in this study, to find the relationship of LST with biophysical compositions, NDVI, NDBI, NDWI, NDMI, and NDBSI were calculated using Equations (4), (9), (10), (11) and (12), respectively. All of these indices were derived using atmospherically corrected Landsat data (surface reflectance).

$$NDBI={\displaystyle \frac{b_{SWIR1}-b_{NIR}}{b_{SWIR1}+b_{NIR}}}$$

(9)

where $NDBI$ represents the Normalized Difference Built-up Index and $b_{SWIR1}$ and $b_{NIR}$ denote the shortwave-infrared and near-infrared bands, respectively [84,85,86].

$$NDWI={\displaystyle \frac{b_{GREEN}-b_{NIR}}{b_{GREEN}+b_{NIR}}}$$

(10)

where $NDWI$ represents the Normalized Difference Water Index and $b_{GREEN}$ and $b_{NIR}$ denote the green and near-infrared bands, respectively [45,84].

$$NDMI={\displaystyle \frac{b_{NIR}-b_{SWIR1}}{b_{NIR}+b_{SWIR1}}}$$

(11)

where $NDMI$ represents the Normalized Difference Moisture Index [87].

$$NDBSI={\displaystyle \frac{\left(b_{SWIR2}+b_{RED}\right)-(b_{NIR}+b_{BLUE})}{\left(b_{SWIR2}+b_{RED}\right)+(b_{NIR}+b_{BLUE})}}$$

(12)

where $NDBSI$ represents the Normalized Difference Bare Soil Index and $b_{SWIR2}$, $b_{BLUE}$, and $b_{RED}$ denote the shortwave-infrared-2, blue, and red bands, respectively [49,73].

2.3.5. Spatial Trend and Hotspot Analysis of LST Change

To comprehensively examine the spatial dynamics of LST and urban climate change, we conducted both spatial trend analysis and hotspot analysis to discern patterns of change. Spatial trend analysis was performed using the Land Change Modeler (LCM) integrated within TerrSet 2020. The trend analysis module of LCM identifies both change-prone and stable regions, assigning numerical values—1 for change and 0 for stability—to enable accurate quantitative assessment. To illustrate spatial change patterns between two time points, the model produced a best-fit polynomial trend surface, following the approach detailed in [59,88]. Specifically, a 7th-order surface trend was applied to capture and analyze the spatial trend of change among various LST classes. Conversely, we also conducted a hotspot analysis to identify significant changes within LST classes. Leveraging the Spatial Getis-Ord Gi* (G-i-star) technique pioneered by [89], we pinpointed clusters of high and low values denoting significant hotspots. The results of the spatial G-i-star analysis, based on z-scores and p-values, revealed clusters of high-value spatial clustering, indicated by high z-scores and low p-values, signifying hotspots of change, and vice versa [59,90]. The G-i-star statistics, expressed through Equations (13-15), provided a quantitative measure of spatial clustering. For hotspot mapping of LST changes, ArcGIS (version 10.6; Environmental Systems Research Institute, Inc., California, USA) was employed, ensuring precise visualization and analysis of the identified hotspots.

$$G_i^{\ast }={\displaystyle \frac{{\sum }_{j=1}^n{w}_{i,j}{x}_j-\overline{X}{\sum }_{j=1}^n{w}_{i,j}}{S\sqrt{\frac{\left[{n{\sum }_{j=1}^n{w}_{i,j}^2-\left({\sum }_{j=1}^n{w}_{i,j}\right)}^2\right]}{n-1}}}}$$

(13)

$$\overline{X}={\displaystyle \frac{{\sum }_{j=1}^n{x}_j}{n}}$$

(14)

$$S=\sqrt{{\displaystyle \frac{{\sum }_{j=1}^n{x}_j^2}{n}}-{\left(\overline{X}\right)}^2}$$

(15)

where $x_j$ represents the attribute value of feature $j,$ $w_{i, j}$ is the spatial weight between feature $i$ and $j$, and $n$ is the total number of features.

3. Results

3.1. Land Use/Land Cover (LULC): Pattern and Changes

Figure 2 and Figure 3 illustrate the spatial distribution and temporal changes in LULC patterns over the study period. From Figure 2, it is evident that the most significant expansion of built-up areas occurred in the central, northern, central-western, south-eastern, and southern parts of the study area. The statistical analysis presented in Figure 3 indicates substantial LULC changes between 1990 and 2023. During this period, built-up areas expanded markedly from 9.73 km² to 22.99 km², reflecting rapid urban growth. In contrast, vegetation cover experienced a notable decline of 12.48 km², indicating a shift from natural to developed land uses. Bare land showed minor fluctuations in area, with a slight overall increase of 0.66 km², while water bodies decreased by 1.44 km², suggesting pressure on natural water resources. Overall, the built-up area increased by approximately 28%, whereas vegetated and water-covered areas decreased by 26% and 3%, respectively, over the 33-year period.

3.2. Urban Climate: LST Distribution and Dynamics

This study investigated the pattern of urban warming in the RCC area by analyzing LST using Landsat thermal imagery. Key LST metrics—including minimum, maximum, and mean values—were derived and are presented in

Table 3. Minimum, maximum, and mean LST and their changing patterns in RCC.

LST (°C)	Year				Change (°C)
	1990	2000	2010	2023	1990–2000	2000–2010	2010–2023	1990–2023
Minimum	20.62	21.10	21.70	22.82	0.48	0.60	1.12	2.20
Maximum	26.67	27.82	29.10	30.96	1.15	1.28	1.86	4.29
Mean	23.32	23.95	24.90	25.98	0.63	0.95	1.08	2.66

. The findings show a progressive rise in mean LST: from 23.32 °C in 1990 to 23.95 °C in 2000, 24.90 °C in 2010, and reaching 25.98 °C in 2023. This corresponds to mean LST increases of 0.63 °C (1990–2000), 0.95 °C (2000–2010), and 1.08 °C (2010–2023) and a total increase of 2.66 °C over the full 1990–2023 period. On average, the mean temperature rose by approximately 0.08 °C per year over these 33 years. Additionally, LST values ranged between 20.62 °C and 26.37 °C in 1990, 21.10 °C and 27.82 °C in 2000, 21.70 °C and 29.10 °C in 2010, and 22.82 °C and 30.96 °C in 2023 (Table 3 and Figure 4). Both the minimum and maximum LST values increased from 1990 to 2023. The minimum LST increased by 0.48 °C, 0.60 °C, 1.12 °C, and 2.20 °C, while the maximum LST increased by 1.15 °C, 1.28 °C, 1.86 °C, and 4.29 °C between the periods 1990–2000, 2000–2010, 2010–2023, and 1990–2023, respectively. These findings indicate a significant increase in maximum temperatures, which has contributed to the overall rise in mean temperatures in the area (Figure 4).

To analyze and interpret the spatial distribution of and temporal changes in land surface temperature (LST), we categorized LST values into four zones: 20–22 °C, 22–25 °C, 25–28 °C, and 28–31 °C (Figure 5). The spatial patterns of these zones over time reveal that the higher temperature ranges have increasingly concentrated in the central, western, and south-central parts of the city in recent decades (Figure 5). Figure 6 illustrates that from 1990 to 2023, there was a sharp change in LST distribution and levels of LST increased, particularly in the central and western parts of the city. Figure 4 denotes that in 1990 and 2000, the largest proportion of the city area, 36.50% and 29.73%, respectively, fell within the 22–25 °C LST zone. However, in 2010 and 2023, the majority of the area (20–22.50% of the total area) was within the 25–28 °C and 28–31 °C LST zones. During these years, a small proportion of the total area (≤ 6%) remained in the 20–22 °C and 22–25 °C zones. Furthermore, Figure 6 shows that the area within the 25–28 °C and 28–31 °C LST zones increased sharply from 2000 onward, while the area within the 22–25 °C zone decreased significantly during the same period. This indicates a notable shift, with a large proportion of the area originally under the 22–25 °C zone transitioning to higher-LST zones, namely, 25–28 °C and 28–31 °C.

To gain a clear understanding of the changing patterns of LST zones and the warming trend over time, the gain and loss areas of LST for each zone were analyzed between specific years. Figure 7 illustrates the conversion patterns of LST zones, highlighting areas of gain and loss. The analysis reveals a significant loss in the 22–25 °C LST zone between 1990–2000 and 2000–2010, accounting for 23.33% (11.20 km²) and 52.33% (25.12 km²) of the total area, respectively. In contrast, there was a notable increase in the areas of the 25–28 °C and 28–31 °C LST zones during 2000–2010 and 2010–2023, representing 75.40% (35.19 km²) and 34.75% (16.68 km²) of the total area, respectively. Overall, from 1990 to 2023, a large portion of the area under the 22–25 °C LST zone decreased by 71.25% (34.20 km²), while the areas under the 25–28 °C and 28–31 °C LST zones increased significantly, covering 83.10% (39.89 km²) of the total area. During this period, LST increased in about 87% of the total area. These findings indicate an overall warming trend in the urban area, primarily due to the transformation of low-LST zones to higher-LST zones (Figure 7).

To determine the level of LST increase and the spatial pattern of warming in the area, the spatial distribution of increasing LST patterns was analyzed, and the results are shown in Figure 8 and

Table 4. Spatial increasing patterns of LST.

Increase in LST (°C)	1990–2000		2000–2010		2010–2023		1990–2023
	Area (km2)	%	Area (km2)	%	Area (km2)	%	Area (km2)	%
<1	33.67	70.15	3.95	8.23	43.35	90.31	9.62	20.04
1–2	12.40	25.83	19.44	40.50	4.53	9.44	17.03	35.48
2–3	1.25	2.60	21.08	43.92	0.04	0.08	16.80	35.00
>3	0.68	1.42	3.53	7.35	0.08	0.17	4.55	9.48
Total	48.00	100.00	48.00	100.00	48.00	100.00	48.00	100.00

. Table 4 shows that during 1990–2000, most of the area (70.15%) experienced an increase of less than 1 °C, while another 25.83% of the area saw a 1–2 °C increase in LST. The spatial distribution of LST increases during this period exhibited a pepper-and-salt pattern, indicating an insignificant warming trend (Figure 8). In contrast, a significant change in LST increase was observed between 2000 and 2010. During this period, 40.50% and 43.92% of the total area experienced a 1–2 °C and 2–3 °C increase in LST, respectively, and 7.35% of the total area saw an increase of more than 3 °C. These statistics indicate that from 2000 to 2010, LST increased by 1–3 °C in about 84.42% of the total area, with some areas experiencing an increase of more than 3 °C (Table 4). The spatial pattern of LST increase reveals that areas with a 2–3 °C and more than 3 °C increase were concentrated in the central, north-eastern, and northern parts of the city, while areas with a 1–2 °C increase were spread throughout the city (Figure 8). On the other hand, during 2010–2023, a much smaller increase in LST was observed, with about 90.31% of the total area experiencing an increase of less than 1 °C and 9.44% of the area seeing a 1–2 °C increase. The slower increase in LST from 2010 to 2023, compared to the 2000–2010 period, can be attributed to changes in land use and urbanization patterns. During 2010–2023, vegetation and water bodies in the fringe areas of the city were gradually converted into paved surfaces. However, these fringe areas exhibited less compact urban development, characterized by lower building density and more open spaces. This less intensive urbanization likely mitigated the rate of LST increase, as the sparse development allowed for better heat dissipation compared to the rapid and dense urbanization observed in the earlier period. The analysis of the increasing pattern of LST between 1990 and 2023 reveals a concerning trend. Over the past 33 years, urban land surface temperature increased by 1–3 °C in 70.48% of the total area, with 9.48% of the total area experiencing an increase of more than 3 °C. The spatial distribution of LST increase from 1990 to 2023 clearly highlights that the central-northern, central part of the western zone, and south-eastern parts of the city experienced an increase of more than 3 °C. Meanwhile, the central, northern, western, and central-southern parts of the city experienced a 2–3 °C increase in LST (Figure 8). These findings indicate a dramatic increase in LST over the past three decades, with a particularly significant rise observed during 2000–2010 in the city area.

To analyze the micro-level urban climate, particularly the distribution and dynamics of urban LST, we examined the LST distribution at the Ward level, which is the basic urban administrative unit of the city. This analysis included the minimum, maximum, mean, and the range (minimum–maximum difference) of LST. The Ward-wise LST distribution showed fluctuations; however, in 1990 and 2000, both minimum and maximum LST were relatively steady. In contrast, a sharp increase in both minimum and maximum LST was observed across all wards in 2010 and 2023 (Figure 9). The average LST distribution in each ward also indicated a continuous increase in LST over the study period (Figure 9c). Figure 9a,b further reveal that the highest increases in minimum temperature occurred in Wards 20, 21, and 22, while the highest increases in maximum temperature were observed in Wards 4, 19, and 26. Conversely, the lowest minimum LST levels were found in Wards 4, 7, 25, 28, and 29, and the lowest maximum LST levels were in Wards 1, 5, 6, 8, 11, 24, and 29 (Figure 9a,b). The differences between minimum and maximum LST showed the following variations: 1.73 to 5.61 in 1990, 2.53 to 7.53 in 2000, 1.82 to 9.26 in 2010, and 1.69 to 5.52 in 2023. Both the lowest and highest variations decreased over time, and a low variation between minimum and maximum LST was observed in most wards, particularly those that are located in the central parts of the city (Figure 9d). This reflects thermal discomfort in these areas. The primary cause of these changes is the reduction in green and blue spaces due to uncontrolled and unplanned urban growth. As urban expansion extended outward from the city center, green and blue spaces in the urban fringe areas continually decreased. To improve the city’s climatic conditions, proper city planning and the implementation of greening policies are essential. These measures should aim to increase green areas or at least maintain them at steady levels for better urban climate management.

3.3. Zonal and Directional Analysis of LST

To determine the zonal and directional changes in LST, a combined directional and zonal study was conducted. This study analyzed LST changes across different LST zones by considering eight quadrants, each at a 45° angle from the city center, serving as the reference point for directional analysis (Figure 10). This combined zonal and directional analysis measures the directional change in LST zones, helping to establish spatial relationships between the LST zones and their dynamics relative to the city center [91]. Over the last 33 years, the highest area reduction in the 22–25 °C LST zone was observed in the NEE direction at a rate of 27.12 ha y⁻¹. From 1990 to 2023, the second and third highest area reductions were found in the NNE and NWW directions, at rates of 24.27 and 21.97 ha y⁻¹, respectively, for the 22–25 °C LST zone. A significant reduction in the area of the 22–25 °C LST zone was also found towards the SEE direction during this period (Figure 10). In contrast, from 1990 to 2023 the highest area increase was noted in the NWW and NNE parts of the city at rates of 19.39 and 16.58 ha y⁻¹, respectively, for the 28–31 °C LST zone. The NEE direction saw an increase of 17.70 ha per year for the 25–28 °C LST zone during the period 1990–2023 (Figure 10). A remarkable increase in the 28–31 °C LST zone was also observed in the NNW direction at a rate of 11.06 ha y⁻¹.

Figure 10 further reveals that between 1990 and 2000, the highest proportion of the city area was under the 22–25 °C LST zone, predominantly in the NEE, NNE, NWW, and NNW directions. By 2010 and 2023, an insignificant amount of the area remained under the 22–25 °C LST zone in all directions from the city center. Conversely, the area under the 25–28 °C LST zone increased gradually from 1990 to 2023, with a higher rate of increase in the NEE, NNE, and NWW directions, particularly in 2010 and 2023. The 28–31 °C LST zone (the warmest zone) saw a notable increase only in 2010 and 2023, concentrated in all directions from the city center except the south and south-west. The mighty river Padma, flowing in the southern part of the city, acted as a barrier to urban expansion, resulting in a lower increase in LST in this region. The directional analysis indicates that areas with low-LST zones, particularly the 22–25 °C range, were converted to higher-LST zones (25–28 °C and 28–31 °C) from 1990 to 2023. This conversion began primarily around 2000, with significant increases observed in 2010 and 2023. Notably, substantial growth in areas with higher LST was seen in the north, north-east, north-west, and south-east directions of the study region (Figure 10). These findings indicate the warming of the urban climate in the study area, primarily in recent decades. Rapid urban expansion, suitable land for settlement, an increase in road networks and infrastructure development, rising land values, and rapid population growth have exerted tremendous pressure on the city area, facilitating mobility and urban expansion and causing an increase in LST.

3.4. Urban–Rural Gradient of LST

The urban–rural gradient approach is utilized to examine the spatio-temporal variation of environmental variables in relation to distance within a city [85,92]. This gradient encompasses a range of conditions and characteristics, transitioning from highly concentrated urban hubs to low-density rural regions, including variations in population density, infrastructure development, economic activities, lifestyles, and environmental conditions. In this study, we apply this approach to analyze the spatial dynamics of LST at specific intervals from the city center to the rural periphery, investigating how it changes across the urban landscape [92]. The urban–rural gradient for LST shows how surface temperatures vary from urban center to rural peripheries. Urban areas, characterized by dense infrastructure, concrete, and asphalt, tend to have higher LST because of the urban heat island effect. This effect occurs due to buildings and roads absorbing and retaining heat, less vegetation, and more human activities, leading to higher daytime temperatures than in rural areas. As you move along the gradient towards suburban and rural areas, the LST gradually decreases.

To determine the pattern of LST concerning the urban–rural gradient, we measured and analyzed the areas within different LST zones using 0.5 km interval buffer zones from the city center (Figure 11a). We examined changes in the areas under various LST zones across the urban–rural gradient for the years 1990, 2000, 2010, and 2023. The results are presented in Figure 11b–e. The figure shows that the area distribution under the 20–22 °C LST zone (cooler zone) was relative to the distance from the city center, and in all the selected study years, less than 11% of each buffer area fell within this zone, with no specific pattern (Figure 11b). For the 22–25 °C LST zone, in 1990 and 2000, the area was relatively small near the city center and gradually increased towards the city periphery. However, in 2010 and 2023, both near the city center and farther from it, very small proportions of each buffer area were under this LST zone (Figure 11c). Overall, the urban–rural gradient pattern of the area under the 22–25 °C LST zone indicates that it was more dominant in 1990 and 2000, whereas in 2010 and 2023, very small areas were under this LST zone near the city center (Figure 11c). The distribution of area under the LST zones 25–28 °C and 28–31 °C from the city center to the periphery clearly shows that in 2010 and 2023, the areas dramatically increased (occupying about 60–90% of each buffer zone area) near the city center compared to the periphery (Figure 11d,e). Interestingly, for the area distribution of the 28–31 °C LST zone, in 1990 and 2000, very negligible areas were under this zone throughout the city center to the periphery. However, in 2010 and 2023, a significant amount of area fell under this zone near the city (Figure 11e). These patterns likely represent the expansion of built-up areas in the city’s core and the greater presence of vegetation near its outskirts.

3.5. Spatio-Temporal Variation of Surface Urban Heat Island Intensity Using UTFVI

The Urban Thermal Field Variance Index (UTFVI) is a measure used to assess the intensity and distribution of urban heat islands (UHIs) within a city. It evaluates the variation in land surface temperature (LST) to identify areas with notable thermal anomalies, typically caused by dense infrastructure, limited vegetation, and high levels of human activity. By calculating the UTFVI, researchers can identify hotspots of warming, which are often associated with negative environmental and health impacts. With the increasing effects of climate change, the UTFVI becomes even more important as cities are likely to face more intense and frequent heatwaves, making it crucial to understand and mitigate UHI effects.

In this study, the UTFVI was derived from LST data, classified, and analyzed to quantitatively explain the UHI effect in the study area. A classified UTFVI offers a clear spatio-temporal indication of the condition of environment and thermal comfort in the city (Figure 12). Figure 12a illustrates the spatial pattern of UTFVI distribution in the RCC area over four different years from 1990 to 2023. The study detected changes in UHI intensity over the years, corresponding with the evolution of several thermal comfort categories in the region. By analyzing the UTFVI, it was found that 58.56%, 62.67%, 54.48%, and 55.35% of the total area in 1990, 2000, 2010, and 2023, respectively, fell under the none/weak level of UHI and good/excellent level of ecological condition, indicating these areas as thermal comfort zones (

Table 5. UTFVI, UHI phenomenon, and ecological conditions.

UTFVI	Area (km2)				% of the Total Area				Urban Heat Island Phenomenon (UHI)	Ecological Evaluation Index (EEI)
	1990	2000	2010	2023	1990	2000	2010	2023
≤0.005	28.11	30.09	26.15	26.57	58.56	62.67	54.48	55.35	None/Weak	Excellent/Good
0.005–0.01	-	-	-	2.27	-	-	-	4.73	Middle	Normal
0.01–0.015	-	-	11.40	2.24	-	-	23.75	4.67	Strong	Bad
0.015–0.02	10.69	10.39	-	2.22	22.27	21.65	-	4.63	Stronger	Worse
≥0.02	9.20	7.53	10.44	14.70	19.17	15.69	21.77	30.63	Strongest	Worst
Total	48	48	48	48	100	100	100	100	-	-

and Figure 12b). In all analyzed years, the area under the middle and strong levels of UHI, corresponding to normal and poor ecological conditions, was very low. However, in 2010 and 2023, the area under the stronger and strongest levels of UHI, indicating worse ecological conditions, showed an increasing trend, with the notably worst condition observed in 2023. This highlights a thermal discomfort zone affected by heat stress in 2023 (Table 5 and Figure 12b). Over the past 33 years, 11.46% of the city area has transitioned into the strongest level of UHI, reflecting a significantly worse ecological condition compared to the strongest UHI level in 1990. Moreover, the evolving pattern of the UTFVI reveals a warming trend in the city, especially in the central, central-western, and central-northern regions (Figure 12a). This trend correlates with increased urban development, particularly the expansion of paved areas, resulting in these regions becoming increasingly unfavorable due to heightened thermal discomfort.

3.6. Spatial Trend of LST Change

This analysis examines the spatial trends in land surface temperature (LST) changes across different zones from 1990 to 2023. The periods analyzed include 1990–2000, 2000–2010, 2010–2023, and the entire period from 1990 to 2023. The spatial trend analysis, as depicted in Figure 13, provides insight into the patterns of temperature changes and simplifies the trend of these changes [93]. During these periods, the changes in LST from 20–22 °C to other levels were generally not significant, with trends mostly ranging from low to medium (Figure 13). In contrast, the trends in LST changes from 22–25 °C to other levels were more pronounced. From 1990 to 2000 and 2000 to 2010, significant increases were observed, especially in the central, northern, central-western, and eastern parts of the city. This trend continued from 2010 to 2023, extending notably towards the northern, eastern, and western areas of the city. Over the entire period from 1990 to 2023, the spatial pattern of LST changes from 22–25 °C to other levels indicated that temperature increases predominantly occurred in the central, western, northern, and south-eastern parts of the city. The increase in LST began in the central areas and gradually spread towards the western, northern, and eastern regions (Figure 13).

For LST changes from 25–28 °C to other levels, high trends were observed in the central-eastern and western parts of the city from 1990 to 2000. In the period from 2000 to 2010, this high trend was concentrated in the central part of the city (Figure 13). However, from 2010 to 2023, the high changing trends were primarily located in the south-eastern and south-western parts of the city. Overall, from 1990 to 2023, the most extreme LST increases were concentrated in the central part of the city (Figure 13). Finally, the trend analysis of LST zones reveals that the central, northern, western, and south-eastern parts of the city have experienced significant increases in LST, making them more vulnerable in terms of eco-urban structure and urban thermal comfort. This vulnerability is largely due to urban development, the reduction in vegetation, and the loss of water bodies in these areas. For sustainable urban development and improved urban thermal comfort, it is crucial to prioritize these areas for restoration efforts. This includes reintroducing greenery, protecting existing green and blue spaces, and possibly expanding water bodies to mitigate the effects of rising land surface temperatures.

3.7. Hotspot Analysis of LST Change

To identify significant hotspots of LST changes across different zones, a hotspot analysis was executed using Getis-Ord Gi* (G-i-star) statistics. Specifically, we analyzed hotspots for the changes in LST zones of 22–25 °C to 25–28 °C, 22–25 °C to 28–31 °C, and 25–28 °C to 28–31 °C. As shown in Figure 14, for the LST change from 22–25 °C to 25–28 °C, hotspots with a 99% confidence level were observed sparsely distributed across the city, with only a few patches identified during the periods 1990–2000, 2000–2010, and 2010–2023. Overall, significant hotspots for this LST change during 1990–2023 were mainly distributed in the northern and north-eastern parts of the city (Figure 14a). During 2000–2010, hotspots for the LST change from 22–25 °C to 28–31 °C were predominantly observed in the north-western and north-eastern parts of the city. For the extreme LST change from 25–28 °C to 28–31 °C, hotspots with a 99% confidence level were concentrated in the central, central-southern, and, to some extent, central-western parts of the city (Figure 14a). In addition to analyzing the zonal distribution of LST hotspots, we also examined the hotspots based on all LST zones combined. The observed spatial patterns of LST change hotspots are presented in Figure 14b. The figure illustrates the evolving pattern of these hotspots over time. From 1990 to 2000, only a few significant hotspot patches were identified. However, during 2000–2010 and 2010–2023, significant LST change hotspots became more concentrated in the central, central-southern, and central-northern parts of the city. Considering the entire period from 1990 to 2023, significant hotspot areas of LST change were identified as compact patches in the central, eastern, northern, and central-western parts of the city (Figure 14b). This indicates significant warming in these areas over time.

3.8. Relationship Between LST, LULC, and Biophysical Component

While the analysis of LST changes, trends, and hotspots provides valuable insights, understanding the relationship between LST variations, LULC, and key biophysical urban components is crucial for identifying the drivers of LST increase. In urban areas, the spatial dynamics of LST are strongly influenced by the eco-environment and landscape structure. Major factors affecting LST include the extent of built-up areas, vegetation cover, presence of water bodies, bare soil exposure, and soil moisture levels. To explore these relationships, we analyzed the spatial patterns of LST increase alongside LULC distribution (Figure 15) and conducted correlation and regression analyses to examine the associations between LST and several critical biophysical variables (Figure 16). Figure 15 reveals that LST is highest in built-up and bare land areas and has increased more rapidly in these zones compared to vegetated regions and water bodies. The expansion of built-up areas between 1990 and 2023 (Figure 3) significantly contributed to local warming, with average surface temperatures rising from 23.64 °C to 26.85 °C, an increase of 3.21 °C. A significant portion of this rise (2.43 °C) is linked to the loss of vegetation cover, which formerly helped moderate surface temperatures. The reduction in water bodies during this period also contributed to a smaller but notable increase of 1.25 °C. Overall, the study period saw an average surface temperature rise of 2.66 °C, highlighting the substantial impact of urbanization and LULC changes on the local climate. On the other hand, Figure 16 illustrates a strong positive correlation between LST and both the NDBI and the NDBSI, indicating that built-up areas and bare soil are major contributors to elevated surface temperatures. In contrast, LST shows a significant negative correlation with the NDVI and the NDMI, reflecting the cooling effects of vegetation and soil moisture. An insignificant positive relationship is observed between LST and the NDWI. These findings emphasize that low levels of vegetation, soil moisture, and surface water, combined with high proportions of impervious surfaces and exposed soil, substantially contribute to LST increases. In essence, areas lacking greenery, moisture, and water exhibit higher surface temperatures, and ongoing declines in these components exacerbate the warming trend. Conversely, the expansion of built-up and bare areas correlates strongly with higher LST, reinforcing the role of urbanization and landscape transformation in driving local temperature increases.

4. Discussion

Land surface temperature is an important phenomenon, with its dynamics playing a strong role in understanding urban climate patterns and the impacts of urbanization. The findings contribute meaningful understanding into the winter LST dynamics in Rajshahi City, Bangladesh, using Landsat data and GIS technology for the years 1990, 2000, 2010, and 2023. The findings indicate that LST has shown a marked increase during the 33-year study period, with the mean LST rising by approximately 2.66 °C (0.08 °C per year) from 1990 to 2023. Specifically, the minimum and maximum LST increased by 2.20 °C and 4.29 °C, respectively, during this period, highlighting a pronounced trend in the maximum LST increase during the study period. This indicates a faster warming of the LST in the city, primarily due to the noticeable rise in maximum LST over the past three decades. Similar trends, i.e., rises in LST, have been observed in various South Asian cities and other parts of the world [24,51,58,75,85,94,95]. For instance, studies have shown significant LST increases in cities such as Dhaka and Chattogram in Bangladesh. In Dhaka, the mean LST increased by about 7.6 °C from 1989 to 2009 [51], while in Chattogram, Ref. [75] reported a 9.60 °C increase in mean LST over the last 30 years. Ref. [50] also found a 6.5 °C increase in mean LST in Chattogram during the period 1998–2018. In Bogra, the mean LST increased by 0.62 °C from 2001 to 2020 [49], and in Khulna, the mean LST rose by 5.54 °C in urban areas and 4.14 °C in suburban areas during the period 1998–2018 [96]. The variation in LST increases across different cities can be attributed to regional weather patterns, levels of urban development, population growth rates, and land use patterns. However, all these studies consistently indicate an overall increase in LST in urban areas over time. Moreover, similar findings of LST increases have been reported in other parts of the world [21,37,78,92], underscoring the global relevance of this issue. This study’s results align with these broader trends, emphasizing the need for further research and policy measures to address the impacts of rising LST in urban environments. Moreover, the 2.66 °C rise in temperature over 33 years (0.80 per decade) is significantly higher than the global average increase of about 0.2–0.3 °C per decade, according to the IPCC [9]. In cities like Rajshahi, local factors such as rapid urban growth, loss of vegetation, and more paved surfaces are likely speeding up this warming. This highlights the importance of city-specific strategies to reduce the impacts.

During the study period, the spatial patterns of LST increases were analyzed in Rajshahi City. The central-northern, central part of the western zone, and south-eastern parts of the city experienced an increase of more than 3 °C. Additionally, the central, northern, western, and central-southern parts of the city experienced a 2–3 °C increase in LST (Figure 8). These dramatic increases in high-LST areas account for about 45% of the total study area, primarily due to the transformation of 22–25 °C LST zones into 25–28 °C and 28–31 °C LST zones (Figure 7), indicating an overall warming trend in the urban area. Areas experiencing winter LST increases of ≥3 °C may face significant environmental, social, and economic consequences. Such warming can alter local microclimates, reduce seasonal cooling relief, and contribute to thermal discomfort, thereby diminishing urban livability and increasing the risk of heat-related stress. According to [97], if the LST is between 27 and 29 °C, the greater part of the community in that area is expected to feel discomfort, and if it is between 29 and 31 °C, the total inhabitants are likely to feel uneasy. Therefore, the LST zones of 25–28 °C and 28–31 °C, where LST has increased by 3 °C or more, can be identified as discomfort thermal zones of the city. These areas have shifted from lower to higher risk levels over time, which correlates with the increase in the built-up area and contraction in the vegetated area. Additionally, due to similar circumstances, Wards 19, 20, 21, 22, and 26, which are primarily located in the central part, can be identified as vulnerable to rising LST and increased thermal discomfort. Particular focus should be given to these wards for effective management. Several cities across low- and middle-income countries have observed that local warming has increased due to urbanization, resulting in deteriorated urban thermal comfort alongside the rise in LST [98].

Rising temperature in urban areas is influenced by changes in LULC, particularly the expansion of built-up areas and the reduction in water bodies. In Rajshahi City, the built-up area dramatically increased from 9.73 km² in 1990 to 22.99 km² in 2023, a gain of 13.26 km² (Figure 3 and Figure 7). The most significant increases in built-up areas occurred in the central, southern, and western parts of the city (Figure 5), which correspond to areas with a 3 °C or greater rise in LST (Figure 8). Consequently, the highest LST increases were observed in the wards located in these parts of the city (Figure 9). LST distribution by LULC classes shows that LST is higher in built-up and bare land areas and has increased more rapidly compared to vegetated areas and water bodies (Figure 15). Built-up areas, dominated by impervious surfaces, absorb and emit solar radiation quickly, leading to higher temperatures [58]. In contrast, vegetation absorbs most of the solar radiation and acts as an intercepting surface, reflecting very little radiation. Water bodies have similar properties. The increase in built-up areas at the expense of vegetation and water has contributed notably to rising LST in Rajshahi. It is worth noting, however, that this analysis emphasizes surface temperatures and does not include air temperature, which also plays a key role in urban heat dynamics. Additionally, sensible heat represents only one element contributing to surface energy dynamics. The role of latent heat, which includes the effects of evaporation and transpiration, should also be considered to gain a more inclusive sense of the warming in the area. These factors present opportunities for further research and could serve as valuable directions for future studies. The highest increases in LST were observed in areas where land use changed from vegetated and water areas to built-up areas. Area under vegetation and water play a crucial role in controlling temperature in urban areas. To reduce LST, maintaining a good proportion of vegetation and water bodies is essential.

Again, zonal and directional analysis in this study reveals that low-LST zones progressively transformed into higher-LST zones over time, with significant increases observed in 2010 and 2023. Notably, higher LST was concentrated in the northern, north-eastern, north-western, and south-eastern areas of the city. This rise in LST can be attributed to rapid urban expansion and population growth in these regions. The transformation of non-urban areas into urban zones indicates that greater urbanization is closely linked to higher LST. Since the rate of urban conversion was highest near the city center, these areas experienced the most significant rise in LST over the study period. Consequently, LST gradually decreased along the gradient from the city center toward suburban and rural areas. The spatial distribution of high-LST zones from the city center to the periphery (urban–rural gradient) over time shows that the concentration of high-LST areas increased closer to the city center, particularly in recent decades. These LST patterns also reflect changes in land use, characterized by dense urban development in the city center and higher levels of vegetation in the outer zones. Both factors significantly contribute to LST distribution and urban warming. The findings from this zonal and urban–rural gradient analysis of LST changes offer valuable insights for urban planning. They can support climate resilience strategies, improve public health and urban thermal comfort, and promote the development of green infrastructure and sustainable practices to mitigate heat stress within urban areas.

In urban areas, the dynamics of LST, particularly its spatial patterns, largely depend on the urban eco-environment and landscape. The connection between LST change and biophysical urban components reveals a strong positive relationship between LST and the NDBI and the NDBSI. Conversely, an inverse relationship was found between LST and both the NDVI and the NDMI. Thus, low levels of greenery, soil moisture, and water bodies, along with high levels of impervious surfaces and bare soil, remarkably contribute to rising LST. The findings of this study and the observed associations between LST and selected biophysical components are consistent with several previous studies. For instance, Ref. [99] found a sharp increase in extremely high-temperature zones over time, largely due to the decline in agricultural lands and water bodies, which were gradually replaced by construction areas. Similarly, Ref. [47] observed that built-up areas exhibited the highest mean LST, followed by bare soil and vegetation. Ref. [100] noted significant increases in LST within urban areas, particularly in paved regions, while [101] confirmed that the expansion of built-up areas contributed to a rise in LST over time. Furthermore, Ref. [102] identified a positive correlation between NDBI, NDBSI, and LST, whereas NDVI and NDWI were inversely related to LST. This suggests that a higher presence of water bodies, healthy vegetation cover, ample open spaces, and reduced concrete surfaces are crucial for maintaining lower LST in urban environments.

Moreover, urban climate change, as reflected in the UTFVI, reveals a significant increase in urban heat island intensity as categorized by the strongest UHI over the past three decades. This UHI zone was increased by 11.46% of the total area, indicating a critical ecological deterioration and a concerning warming trend in LST, particularly in the central, central-western, and central-northern parts. These regions correspond to areas of rapid urban development, characterized by the expansion of paved surfaces, which exacerbate thermal discomfort and make these areas increasingly unfavorable for living conditions. Urbanization alters the local climate by replacing natural surfaces with impervious ones, leading to increased heat absorption and reduced evapotranspiration. This change contributes to higher local temperatures and exacerbates the UHI effect. Refs. [48,78,103,104] highlighted that the gradual increase in LST contributes to the intensification of UHI. UTFVI levels tend to be higher in areas that are significantly warmer than their surroundings. These studies also emphasized the notable impacts of elevated UTFVI, including adverse effects on local wind patterns, humidity, air quality, and overall comfort levels, which indirectly result in economic losses. Furthermore, Ref. [78] observed a mutual linkage between the UTFVI microclimate thermal zones and urban land cover patterns. Areas with vegetation cover corresponded to optimal microclimates (UTFVI < 0.01), whereas thermal discomfort zones (UTFVI > 0.01) were associated with impervious surfaces. Furthermore, through spatial trends and hotspot analysis of LST changes, we observed that over the past 33 years, LST has predominantly increased in the central, western, northern, and south-eastern parts. The rise in LST initially began in the central areas and gradually expanded towards the western, northern, and eastern regions. Significant LST hotspots during this period were identified as compact clusters in the central, eastern, northern, and central-western parts. These findings highlight remarkable warming in these areas over time, designating them as zones of heightened thermal discomfort. Therefore, proactive measures are necessary to mitigate urban heat island effects, foster sustainable urban development, and enhance the overall quality of life in urban areas.

5. Conclusions

This study examined urban climate dynamics in Rajshahi City, Bangladesh, focusing on land surface temperature (LST) changes and heat island intensity in relation to selected biophysical compositions over the period 1990 to 2023. Using Landsat thermal data and GIS techniques, LST changes were analyzed geospatially. The results revealed a significant increase in urban LST during this period, with the mean LST rising by 2.66 °C (0.08 °C per year), accompanied by increases of 2.20 °C in minimum LST and 4.29 °C in maximum LST. These findings indicate a pronounced warming trend in the urban area. The geospatial analysis highlighted that areas in the central-northern, central-western, and south-eastern parts of the city saw LST rises exceeding 3 °C, accounting for approximately 10% of the total area. while another 35% saw increases of 2–3 °C, encompassing regions in the central, northern, western, and central-southern parts. This alarming rise in LST has resulted in expanding thermal discomfort zones, posing significant environmental and public health risks. Furthermore, the observed LST increases (>2.0 °C) in these urban zones have already surpassed the United Nations Framework Convention on Climate Change (UNFCCC) aim to cap temperature increases below the 2.0 °C threshold [105]. Hence, this evidence reinforces the call for resilient urban design and heat mitigation measures to manage the growing threat of urban warming. The primary driver of this warming is rapid urban development and expansion, which has transformed areas under vegetation and water bodies into impervious surfaces, intensifying the UHI effect. This was validated by a strong positive correlation between LST and indices such as the NDBI and NDBSI and a strong negative correlation between LST and the NDVI, NDWI, and NDMI. The results highlight that reductions in green and blue spaces, coupled with expanded paved surfaces, directly contribute to increased urban temperatures. If this trend persists, further reductions in green and blue spaces, coupled with an increase in paved areas, will likely result in even greater urban warming in the future.

The spatial analysis of LST changes revealed that higher LST areas corresponded to regions of intense urbanization, with a growing concentration of high-LST zones near the city center in recent decades. This aligns with the transformation of land uses in central areas, where non-urban land has been replaced by built-up infrastructure, in contrast to the peripheral zones. The UTFVI further highlighted the expansion of strong UHI effects, spanning around a third of the urban landscape, primarily clustered in the core, central-western, and central-northern regions. The UTFVI and UHI analyses serve as vital tools for urban planners, aiding in the strategy development to mitigate LST impacts and encourage sustainable urban expansion. Moreover, hotspot analysis identified compact clusters of LST in central, eastern, northern, and central-western areas, marking zones of significant thermal discomfort and ecological vulnerability. These findings reinforce the urgent need for climate-responsive urban planning. Preserving vegetation, enhancing water bodies, promoting permeable surfaces, and adopting climate-adaptive urban design are essential for mitigating heat stress. The integration of Landsat and GIS in this study also provides a scalable framework for analyzing urban thermal patterns in other rapidly urbanizing regions. However, it is important to note that this study relied on single-date thermal satellite data due to the limited availability of consistent, high-resolution thermal imagery over extended time periods. This constraint represents a limitation of the dataset. Nevertheless, the adopted approach remains methodologically sound and is widely supported in the literature [31,100,101]. Future studies could strengthen the analysis by incorporating multi-temporal or seasonal data as they become more readily available. Finally, to mitigate UHI effects and urban warming, it is crucial to prioritize urban greening initiatives, incorporate permeable surfaces, and adopt climate-resilient urban designs. Regular monitoring of LST and UHI patterns using satellite data, along with fostering community engagement in green infrastructure projects, is essential for creating sustainable urban environments. Rajshahi City Corporation (RCC) has recently initiated the ‘Zero Soil Project’, aimed at reducing bare soil and converting it to vegetation to combat UHI impacts and improve the city’s ecological conditions. By addressing the challenges identified in this study and implementing the recommended strategies, Rajshahi City can alleviate the urbanization-induced climate stress and move toward a more resilient and sustainable cityscape. Furthermore, combating urban climate change on an international scale requires collaborative efforts to share best practices, improve technological accessibility, and promote green infrastructure globally. Policies emphasizing sustainable urban development, coupled with continuous monitoring and global knowledge exchange, can help cities worldwide adapt to and mitigate the effects of urban climate change, fostering more resilient and livable urban landscapes.