Evaluating Urban Heat Islands Dynamics and Environmental Criticality in a Growing City of a Tropical Country Using Remote-Sensing Indices: The Example of Matara City, Sri Lanka

Abstract

Urbanization has undeniably improved human living conditions but has also significantly altered the natural landscape, leading to increased Urban Heat Island (UHI) effects. While many studies have examined these impacts in other countries, research on this topic in Sri Lanka remains limited. This study aimed to evaluate the effects of changes in built-up areas (BAs) and Vegetation Cover (VC) on UHI and environmental criticality (EC) in Matara cityCity, Sri Lanka, utilizing Landsat data. This study employed the commonly used remote-sensing (RS) indices such as the land surface temperature (LST), the UHI Index, and the Environmental Criticality Index (ECI). Various techniques were utilized including supervised image classification, Urban–Rural Gradient Zone (URGZ) analysis, grid-based analysis, UHI profiles, and regression analysis. The results revealed that built-up areas increased by 12.21 km², while vegetation cover decreased by 9.94 km², and this urban expansion led to a 2.7 °C rise in mean LST over 26 years. By 2023, newly developed BA showed the highest LST and environmental criticality, with mean LST values ranging from 25 °C to 21 °C in URGZs 1 to 15 near the city center, and lower values of 15 °C to 16 °C in URGZs 40 to 47 further from the core. The correlation analysis highlighted a strong positive relationship between the NDBI and LST, underscoring the significant impact of BA expansion on LST. Consequently, high-density built-up areas are experiencing high environmental criticality. To minimize these effects, planning agencies should prioritize green urban planning strategies, particularly in high LST and environmental criticality zones. This approach can also be applied to other cities to assess the UHI and LST phenomena, with the goal of protecting the natural environment and promoting the health of urban dwellers.

1. Introduction

Rapid population growth and an accelerating shift from rural to urban environments have made urbanization a global phenomenon. By 2018, 55% of the world’s population resided in cities, and it is projected that 68% will live in urban areas by 2050 [1,2,3]. In developing countries, the urban population is expected to double by 2050 [1]. Sri Lanka has been undergoing rapid urbanization in recent decades, with the urban population expected to increase to 60% by 2030, up from 14% in 2010 [1].

Urbanization is a major driver of land use changes, often resulting in the conversion of non-urban areas to urban uses [4,5]. While urbanization is generally associated with economic development, unchecked and unplanned urban expansion can disrupt natural and social systems, making it a potentially destructive process. To accommodate urban growth, infrastructure such as roads, bridges, and residential buildings are developed, leading to significant changes in land use patterns. These changes in Land Use Land Cover (LULC) are crucial issues in urban areas [6,7], affecting the microclimate and ecology. Recent studies have shown that LULC variations contribute to changes in urban temperature, leading to the formation of urban heat islands (UHIs) [4,6].

UHI is a significant environmental concern with adverse effects on both the natural environment and human health. UHI refers to the phenomenon where urban areas experience higher temperatures compared to their rural surroundings. UHIs can be categorized into two types: (i) atmospheric UHI, which is based on air temperature measurements, and (ii) surface UHI (SUHI), which is derived from land surface temperature (LST) [6,8]. LST is a crucial factor in controlling the surface energy balance and influences various physical, chemical, and biological processes on Earth’s surface. While both surface and atmospheric UHIs are important, they are driven by LST and air temperature, respectively [6]. Surface UHI can be generally observed both day and night, but are typically more pronounced during the day when solar radiation is at its peak [7]. High LST values are commonly found within urban heat islands, particularly in urban hot spots [9].

The extensive growth of built-up areas (BAs) and reduced vegetation cover (VC) have significantly raised LST and intensified UHI. Variations in urban LULC further exacerbate UHI, with temperature increases ranging from 2 °C to 8 °C in urban areas [10]. Consequently, green urban planning has become a critical approach and effective strategy for mitigating the adverse effects of UHI. Together, Geographical Information Systems (GIS) and remote sensing (RS) offer comprehensive tools to understand and solve complex spatiotemporal geographical issues by providing detailed spatial analysis and continuous monitoring capabilities.

Given the recent proliferation of UHI in urban areas globally, there has been a noticeable increase in systematic investigations using modern RS data and techniques, integrated with GIS, to address and analyze these spatial phenomena. RS data and techniques are crucial and reliable for examining LST and its impacts on UHI [8]. The changes in LULC often exacerbate UHI in urban and suburbs. Technological advancements have enabled the acquisition of both thermal and optical data from the sensor platforms, enhancing research on temperature conditions and urban LULC [6]. Satellite imagery at various temporal scales allows for the assessment of UHI magnitude in urban areas. As a moderate-resolution satellite, Landsat is effective for analyzing LST in urban areas [9,10].

Recent scholarly studies have extensively examined the impact of LULC dynamics on UHI using remote-sensing techniques in various countries. For example, an analysis of LST in Narayanganj City, Bangladesh, revealed a significant correlation between LST and the NDBI [11]. According to studies in Wuhan City, the expansion of built-up areas resulted in a 6.05 °C increase in LST from 2007 to 2020 [11,12]. A study in Yerevan, Armenia, using Landsat TM/ETM+/OLI-TIRS images, also highlighted that declines in LULC contributed to greater UHI impacts [13]. Similarly, a study in Kolkata showed that LST increased rapidly in the urban core due to the expansion of built-up areas over vegetation [9]. Additionally, a recent study employing Multifractal Detrended Fluctuation Analysis (MDFA) with space-borne sensor data in the Indian Himalayan foothills confirmed that urbanization has accelerated UHI [9,14].

Despite being an island nation with limited land resources, Sri Lanka experienced a 6.42% annual increase in urban area growth from 1995 to 2017, making it one of the least urbanized countries globally [15,16]. Various studies have investigated UHI phenomena in Sri Lanka, predominantly focusing on the Colombo [16,17,18] and the Kandy City areas [19]. Additionally, studies have examined UHIs and their effects on the ecology and socio-economic environment in peripheral urban areas such as Gampola [20]. Given that smaller cities in Sri Lanka are also experiencing UHI effects, it is crucial to extend UHI studies to those urban areas across the country.

The Southern Province of Sri Lanka includes three administrative districts: Galle, Matara, and Hambantota. Covering 4946 km² hectares and with 2.5 million residents [21,22], this province has experienced notable changes in LULC due to increased urbanization and infrastructure development over recent decades. Urbanization and increasing population density have markedly altered the landscape in recent decades. The Matara Divisional Secretariat Division (DSD) was selected for this study due to its status as one of Sri Lanka’s municipalities and its strategic location connecting major cities like Galle and Hambantota. The area has seen substantial urban expansion driven by growing demand for residential, commercial, and recreational facilities, influenced by its popularity as a tourist destination.

No previous research has focused on the relationship between LST, UHI, and urban expansion in Matara. Thus, our research aimed to address this gap by analyzing the relationship between landscape dynamics LST and UHI. We derived commonly used remote-sensing indices, including LST, UHI, Normalized Difference Vegetation Index (NDVI), Normalized Difference Built-up Index (NDBI), and Environmental Criticality Index (ECI) using Landsat 5, 8, and 9 images for the years 1996, 2007, 2016, and 2023. To gain a deeper understanding of the situation, we analyzed the relationship between mean LST, NDVI, NDBI, and ECI within Urban–Rural Gradient Zones (URGZs). Also, a grid-based analysis and linear regression were performed among the Fraction of Built-up Area (FBA) and the Fraction of Vegetation Cover (FVC). Further, in this study, we focused on the ECI in relation to the expansion of FBA and the decline of FVC within different URGZs. The findings are intended to support sustainable urban planning, mitigate the impacts of LULC changes, and enhance the ecosystem and human health in Matara City.

The implications of this research extend significantly into the realm of climate change adaptation and the achievement of various Sustainable Development Goals (SDGs). The findings underscore the urgent need for sustainable urban planning to mitigate the urban heat island effect, which directly impacts climate resilience (SDG 13: Climate Action). By demonstrating the adverse effects of urban expansion on vegetation cover and surface temperatures, the research highlights the importance of maintaining green spaces to promote ecosystem health (SDG 15: Life on Land). Additionally, integrating green infrastructure can enhance urban livability and well-being (SDG 11: Sustainable Cities and Communities), fostering healthier environments for residents. The correlation between built-up areas and rising temperatures also emphasizes the need for inclusive urban policies that address socioeconomic disparities, aligning with SDG 10: Reduced Inequalities. Ultimately, this research serves as a critical resource for policymakers, providing evidence-based strategies that support both environmental sustainability and human health, vital for navigating the complexities of urbanization in a changing climate.

The social impact of this study is significant, as it provides critical insights into how urbanization in Matara affects local climate conditions, particularly through the UHI effect. By highlighting the relationship between LULC changes and temperature variations, the findings can inform urban planning strategies aimed at improving public health and quality of life. Additionally, this study emphasizes the importance of preserving green spaces, which can foster community well-being and environmental sustainability. Ultimately, this research supports informed decision-making that benefits residents and enhances resilience to climate change impacts.

2. Materials and Methods

2.1. Study Area: Experimental Site

Matara City, located on Sri Lanka’s southern coast at 6.10° N and 80.13° E, is the administrative capital of the Matara District and a pivotal commercial center in the province. The city has been experiencing rapid urbanization, underscored by its classification as a first-order urban center in the National Physical Planning Department’s regional structure plan. This growth is characterized by substantial infrastructure development, including the construction of new roads, commercial buildings, and residential complexes. Additionally, Matara City has seen enhancements in amenities such as shopping centers, schools, and healthcare facilities. Consequently, the city has witnessed increased population density, expanded economic opportunities, and significant overall urban development.

According to the 2001 census, Matara experienced a 40% increase in population, leading to the incorporation of 46 surrounding villages into the city’s administrative boundaries [21,22]. The Urban Development Authority (UDA) of Sri Lanka declared Matara as a municipality in 2002 [22]. As a major commercial hub on Sri Lanka’s southern coast, Matara has the potential to drive regional development. The Nilwala River, which flows through the city from north to south, is a notable geographic feature. Matara enjoys a tropical climate, with an average annual precipitation of approximately 2147 mm and an average annual temperature of 26.8 °C. The elevation of the DSD ranges from −6 m to 127 m above Mean Sea Level (MSL), with slopes varying from 0 to 30 degrees. Our study focuses on the Matara DSD, encompassing the Matara Municipal Council area and its periphery (Figure 1).

2.2. Materials

The satellite images were taken during the wet and dry seasons and were downloaded via the United States Geological Survey (USGS) website [23,24]. For analysis, four images with minimal cloud cover/less than 10%/completely cloud-free were downloaded. The whole study area belongs to path 56 and row 141. A summary of all the data about the two sensors is included in

Table 1. Description of the satellite datasets.

Satellite	Path/Row	Date	Time (GMT)	Cloud Cover	Thermal Conversion Constants
					K1	K2
Landsat 5 TM	141/056	24 February 1996	04:02:15	2.00	607.76	1260.56
Landsat 5 TM	141/056	3 March 2007	04:49:17	3.00	607.76	1260.56
Landsat 8 OLI	141/056	28 November 2016	04:54:36	0.27	774.8853	1321.0789
Landsat 9 OLI	141/056	23 December 2023	04:54:36	7.18	774.8853	1321.0789

Note: K₁ is the calibration constant related to the spectral radiance; K₂ is the constant used in the exponential part of Planck’s equation to calculate the satellite temperature.

[23,24]. Thermal bands (TM Band 6 and OLI Band 10) were utilized for retrieving LST. The boundaries of the study area, as well as the digital map layers for Sri Lankan districts and DSDs, were sourced from the digital data layers provided by the Sri Lanka Survey Department, at a scale of 1:20,000. The spatial data were processed using ArcMap 10.8 software (ESRI 2020, USA).

2.3. Methods

Our study employs a novel integration of established remote-sensing techniques and innovative statistical analyses to examine LULC changes and their implications for UHI. While the methods of supervised classification and various indices (NDVI, NDBI, ECI) are widely utilized in remote-sensing research, our approach innovatively combines these techniques with a detailed urban–rural gradient analysis and grid-based assessments to derive spatial relationships between LST, vegetation, and impervious surfaces. The scientific logic behind this innovation lies in the need for a comprehensive understanding of urban dynamics and environmental impacts in rapidly urbanizing regions. By systematically evaluating LULC changes combined with climatic variables, our methodology enhances the accuracy of urban planning and environmental management strategies. The integration of regression analysis with grid-based spatial assessments allows for a nuanced understanding of the interplay between urbanization and local climate, providing valuable insights for sustainable development initiatives. This multifaceted approach not only addresses gaps in traditional methodologies but also contributes to the existing body of knowledge on urban environmental dynamics in the Global South. Several key steps were undertaken to create LULC maps and RS indices, including image preprocessing, classification, accuracy assessment, change detection, calculation of RS indices, and performing statistical analysis for the target period. A broad technical flowchart illustrating the research process encompassing four phases is illustrated in Figure 2.

2.3.1. Image Preprocessing

The data preprocessing must be performed before image classification and change detection to establish a more thorough link between the data acquired and the biophysical properties on the ground. The data preprocessing involved several essential steps to ensure cleanliness, accuracy, and readiness for analysis. First, radiometric calibration was carried out by converting raw digital numbers (DN) into radiance or reflectance values, which helped correct sensor-specific biases. The spatial consistency of all images was maintained by re-projecting them into the WGS84/UTM 44N coordinate system. During the image sub-setting step, the images were clipped to the areas of interest (AOI) using the Matara DSD boundary polygon. Then, the spatial resolution of the images was adjusted using the nearest neighbor resampling technique, ensuring a consistent 30 m resolution.

2.3.2. Image Classification

LULC categorization is crucial for understanding and managing environmental change, which supports sustainable development [25,26]. The categorization of satellite images, particularly in urban areas, is often complex due to spectral heterogeneity. Various approaches have been employed by researchers to classify satellite image pixels into distinct classes [27,28]. Past researchers have mapped satellite image pixels into different LULCs utilizing remote-sensing data through various classification approaches [29,30,31,32,33].

In the present analysis, we classified LULC in 1996, 2007, 2016, and 2023 by applying the Maximum Likelihood Classification (MLC) algorithm. MLC is a widely used statistical approach in remote sensing and image analysis for classifying land cover and other features in multispectral and hyperspectral data. It operates under the assumption that data follow a normal (Gaussian) distribution and assigns each pixel to the class with the highest probability based on mean and variance parameters. While robustness in handling diverse datasets, effective complexity management, and flexibility are significant advantages of the MLC method, its reliance on the Gaussian assumption and heavy dependency on accurate training data are notable weaknesses. When training data are chosen well, MLC may achieve a high classification accuracy, making it one of the most commonly employed methods for LULC mapping. Given said advantages, we also used this algorithm in our study. The spectral characteristics gleaned from training samples are used in supervised image classification. To manage all of the spectral signatures from the training regions for the images that are being classed, the signature editor was used [4,5]. Also, this is commonly utilized because regional land cover mapping may be achieved at a reasonable cost using medium-resolution satellite remote-sensing data [10,25]. It depends on how likely a pixel is to belong to a particular class. Supervised classification methods in LULC mapping offer several advantages, primarily due to their reliance on training data, which enhances classification accuracy. By using known samples to train the model, these methods can effectively distinguish between different LULC types based on their spectral characteristics. This results in higher accuracy rates compared to unsupervised methods, particularly in complex landscapes where spectral signatures overlap [31,32]. Furthermore, supervised classification allows for more precise customization of the classification scheme, enabling the incorporation of expert knowledge to define classes that are relevant to the specific study area. Overall, these advantages make supervised classification a preferred approach for accurate LULC mapping in diverse environments. In our analysis, the LULC was categorized into four classes: built-up, vegetation, water and marshlands, and homesteads (

Table 2. Details of LU/LC classes in the study area.

Refs.	LU/LC	Details
[4]	Built-up	Buildings, roads, and all other impervious surfaces
[5,16]	Vegetation	Forest, grassland, and other green areas
[19]	Water and marshlands	Sea, river, wetland areas, swamps, irrigation, and drainage canals
[26]	Homesteads	All other LULC types except to the above

).

2.3.3. Accuracy Assessment

The equal random sampling technique was employed to assess the accuracy of the LULC classification. To compare the pixels in the identified images with reference data for each year, these samples were used. Independent training points were utilized to evaluate accuracy. For each year, 400 points were generated, covering all LULC categories [34]. Google Earth served as the reference data for accuracy evaluation. Results were statistically analyzed using a confusion matrix [28,35,36], which organizes numbers into rows and columns to display the different sample points assigned to each LULC class compared to the actual ground conditions [37].

The kappa coefficient was used to assess the accuracy of the land use classification. To measure classification accuracy, the user accuracy, producer accuracy, overall accuracy, and kappa coefficient were calculated [38,39,40,41]. Equations (1)–(4) were used to compute producer accuracy (1), user accuracy (2), overall accuracy (3), and the kappa coefficient (4) [26,27,28,38,40,41].

$$OverallAccuracy={\displaystyle \frac{\sum CCP\left(Diagonal\right)}{\sum CRP}}\times 100$$

(1)

where the corrected reference pixels are CRP, and the corrected categorized pixels are CCP (diagonal);

$$UserAccuracy={\displaystyle \frac{\sum CCP\left(Category\right)}{\sum CPC\left(Row\right)}}\times 100$$

(2)

where CCP (category) is the corrected classified pixels (category) and CPC (row) is the classified pixels in that category (the row total);

$$ProducerAccuracy={\displaystyle \frac{\sum CCP\left(Category\right)}{\sum CPC\left(Column\right)}}\times 100$$

(3)

where CCP (category) is the corrected classified pixels (diagonals) and CPC (column) is the classified pixels in that category (the column total);

$$KappaCoefficient={\displaystyle \frac{N{\sum }_{i=1}^r{X}_{ii}-{\sum }_{i=1}^r{(X}_{i+}\times X_{+i})}{N^2-{\sum }_{i=1}^r{(X}_{i+}\times X_{+i})}}$$

(4)

where N is the total samples; r is the number of rows in the error matrix; X_ii is the total corrected samples in the ith row and column; N₂ is the square of total samples; X_i+ is the column total; and X_+i is the row total [27,28,29].

2.3.4. Deriving LST

Landsat 5 TM (1996/2007), Landsat 8,9 Optical Land Imager (OLI), and Thermal Infrared Sensor (TIR) data were utilized to retrieve LST using the Split-window (SW) technique. The process involved converting the DNs from the thermal bands into spectral radiance values [28]. This conversion was achieved using thermal conversion constants, which allowed for the calculation of the satellite brightness temperature in degrees Kelvin (°K) [27,28,42,43].

Initially, the raw data from the thermal bands (Band 6 in Landsat TM and Band 10 in Landsat 8) were transformed into spectral radiance values using Equation (5). These values were then used to compute LST [9,44,45].

$$L_{\lambda }=M_L\times Q_{Cal}+A_L$$

(5)

where L_ϕ is the Top of Atmosphere (TOA) spectral radiance (watts/(m²·sr·µm)), M_L is the multiplicative rescaling factor based on a specific band from the metadata, Q_Cal is the quantized and calibrated pixel values of standard product, and A_L is the additive rescaling factor based on a specific band from the metadata. Thermal bands including brightness temperatures expressed in Kelvin were employed.

Equation (6) was used to calculate the land surface emissivity of the LST before retrieval [43,46].

ε = mPV + n

(6)

where ‘ε’ is land surface emissivity, n = 0.004 [8] and m = 0.986 [8], and PV is the proportion of vegetation. Equation (7) was then incorporated for the extraction of the proportion of vegetation (PV) [47,48,49] as follows:

PV = ((NDVI − NDVImin)/(NDVImax − NDVImin))²

(7)

where PV is the proportion of vegetation and NDVI is calculated using Equation (10) as in Section 2.3.6.

Finally, the study used Equation (8) to convert the LST value (Kelvin) into (°C).

LST (°C) = LST (°K) − 273.15

(8)

2.3.5. UHI Retrieval

The UHI Index was used to analyze the UHI phenomenon in more detail within the MC. Typically, high UHI Index values indicate warmer regions within a city. The UHI Index was calculated using the following Equation (9) as described by Sultana and Satyanarayana (2018) [9].

LST > μ + þ 0.5 × δ

(9)

where μ and δ are the mean and standard deviation of LST in the study area, respectively.

2.3.6. NDVI Calculation

Spectral analysis results are often distilled into vegetation indices, which correlate reflectance values from two or more wavelength bands in satellite imagery. These indices are particularly useful for applications like precision agriculture and vegetation assessment. NDVI is the most widely used vegetation index, commonly utilized to assess chlorophyll content and overall vegetation health. NDVI was employed to assess vegetation cover on the surface and to determine the presence of vegetation at specific locations. NDVI represents the vegetation conditions of the surface, which influence local climatic variables such as precipitation [46,47]. The NDVI values range from −1 to 1, with negative values typically indicating water bodies [49]. In contrast, an NDVI score close to +1 signifies areas with dense vegetation. NDVI is commonly computed using the normalized difference between the near-infrared band and the red band of the image. The following Equation (10) was used to calculate NDVI [28,49].

NDVI = (NIR − RED)/(NIR + RED)

(10)

where the NIR band represents Band 4 in Landsat TM (0.76–0.90 µm (wavelength)) and Band 5 in Landsat OLI (0.85–0.88 µm (wavelength)), respectively, and the RED band represents Band 3 in Landsat TM (0.63–0.69 µm (wavelength)) and Band 4 in Landsat OLI (0.64–0.67 µm (wavelength)), respectively.

2.3.7. Deriving NDBI

NDBI (Normalized Difference Built-up Index) is a widely used index in remote sensing to identify and monitor urban and built-up areas. It is calculated using the reflectance values from satellite imagery, typically leveraging the contrast between near-infrared (NIR) and short-wave infrared (SWIR) bands. While this is a widely used, simple, and effective method for monitoring urban growth and classifying urban LULC, with the ability to integrate with vegetation indices, it has weaknesses such as spectral overlap, data dependency, and limited contextual analysis capabilities. Positive NDBI values indicate the presence of built-up areas, while negative values typically represent vegetation. The NDBI values generally range from −1.0 to +1.0. Vegetation is usually associated with negative NDBI values, whereas built-up areas are indicated by positive values. NDBI is derived using the mid-infrared and near-infrared bands, as outlined in Equation (11) [48,49].

NDBI = (MIR − NIR)/(MIR + NIR)

(11)

where the MIR band represents Band 5 in Landsat TM (1.55–1.75 µm (wavelength)) and Band 6 in Landsat OLI (1.57–1.65 µm (wavelength)), respectively, and the NIR band represents Band 4 in Landsat TM (0.76–0.90 µm (wavelength)) and Band 5 in Landsat OLI (0.85–0.88 µm (wavelength)), respectively.

2.3.8. Calculating ECI

The environment is in a critical state, as indicated by the Environmental Criticality Index (ECI), due to rising LST and declining NDVI. LST shows a direct correlation with ECI, while NDVI demonstrates an inverse relationship [17,18]. The ECI was used to identify environmentally critical areas by analyzing the ratio between LST and NDVI [14,15,27]. Additionally, the Built-up (BU) Index, which is based on the differences between NDBI and NDVI, was employed to distinguish urban areas. This method helps to differentiate urban locations more effectively by minimizing the influence of vegetation reflections. The BU Index was calculated using Equation (12) [15,27].

BU = (NDBI) − (NDVI)

(12)

The retrieved built-up (BU) layers were initially normalized using histogram equalization, adjusting pixel values to a range between 1 and 255. Following this, Equation (13) was employed to estimate the ECI [16,17,30].

ECI = LST × BU (1 − 255 stretched)

(13)

The pixel values of the derived images were normalized to a range between 0 and 1 using raster normalization techniques. Higher ECI values denote more environmentally critical areas. Prior to identifying these critical areas, water bodies and green spaces were excluded through masking. The spatial variation in ECI across the study area was categorized as follows: very high (75–100), high (50–75), moderate (40–50), low (25–40), and very low (<25). This classification method is consistent with the approach used by [16].

2.3.9. Urban–Rural Gradient Analysis

The Urban–Rural Gradient Zone (URGZ) analysis was used to explore the relationship between mean LST, FBA, and FVC, as the distance from the city center varies. This method has been applied in previous studies of single-core cities to assess spatial variations in LST, impervious surfaces (ISs), and green space (GS) [16,18,19].

To derive the URGZs, a series of methodical steps were followed. First, 210 m buffers were created and overlaid on the LST, NDBI, and NDVI maps. An amount of 48 URGZs were then established, considering the maximum distance from the city center to the city boundary (10.2 km), with the center of Matara as the reference point (0 grid). For each zone, the mean LST, FBA, and FVC were calculated using the zonal statistics tool. Regression analysis was subsequently used to examine the data and reveal spatial variations along the urban–rural gradient [25,26,27].

2.3.10. Grid Analysis

This method is also commonly employed in research on LST and LULC change analysis [16,17]. To analyze the spatial relationships among mean LST and FBA, a 210 m × 210 m (7 × 7 pixels) grid was utilized. Previous studies in this field have demonstrated that this grid size provides more accurate correlations [16,19,26].

2.3.11. Statistical Analysis

To analyze the relationship between LST, mean LST, and other indices, regression analysis was conducted using MS Excel. Scatter plots were then created to visualize the relationships between LST, FBA, FVC, and other indices for the years 1996, 2007, 2016, and 2023. For this analysis, pixel values for LST, NDVI, NDBI, FBA, and FVC were converted into data points.

3. Results

3.1. LULC Changes

The combined accuracy rates for the four periods exceeded 80%, reflecting a reliable classification of LULC and a strong correlation between the reference and classified maps. The kappa coefficients were 0.89 in 1996, 0.96 in 2007, 0.96 in 2016, and 0.92 in 2023, as shown in

Table A1. Statistics of accuracy assessments of LULC classes.

	LULC	1996	2007	2016	2023
User accuracy (%)	water	88	100	100	92
	built-up	92	92	92	96
	homestead	88	96	96	92
	vegetation	100	100	100	96
Producer accuracy (%)	water	100	100	96	95
	built-up	95	100	100	96
	homestead	100	92	96	95
	vegetation	78	96	96	88
Overall accuracy (%) and kappa		92	97	91	94
		0.89	0.96	0.96	0.92

.

The classified land cover maps for Matara DSD for the years 1996, 2007, 2016, and 2023 are illustrated in Figure 3, while the changes in LULC are detailed in

Table A2. LULC statistics in study area, 1996–2023.

LULC Classes	1996–2007		2007–2016		2016–2023		1996–2023		1996–2016		2007–2023
	km2	%	km2	%	km2	%	km2	%	km2	%	km2	%
Built-up	6.4	50.19	5.42	47.7	−0.35	4.72	12.21	48.8	11.82	50.02	5.81	33.2
Water and Marshland	−0.02	0.15	−0.66	5.8	0.95	112.8	0.27	1.08	−0.68	2.87	0.29	1.65
Homestead	−5.16	40.47	0.24	2.11	2.37	32	−2.55	10.21	4.92	20.8	2.61	14.9
Vegetation	−1.17	9.17	−5.04	44.36	−3.37	50.4	−9.94	39.8	−6.21	26.2	−8.77	50.1
Total	12.75	100	11.36	100	7.4	100	24.97	100	23.63	100	17.48	100

. The results indicate a gradual increase in built-up areas from 1996 to 2023. Figure 3 highlights the rapid urban development pattern observed from 1996 to 2023, progressing outward from the city center. This spatial pattern of LULC is most prominent in the southwest area. Concurrently, vegetation cover has steadily decreased due to urban expansion, which has been driven by population growth.

According to Table A2, significant changes in LULC have been observed from 1996 to 2023. Built-up areas increased substantially, with a total rise of 12.21 km² (48.8%) over the period.

This growth was most pronounced between 1996 and 2016, when built-up areas expanded by 11.82 km² (50.02%), although there was a slight decrease of 0.35 km² (4.72%) from 2016 to 2023. This reflects a period of rapid urban expansion followed by a minor stabilization.

In contrast, water and marshland areas experienced a net increase of 0.27 km² (1.08%), despite an overall decline of 0.68 km² (2.87%) from 1996 to 2016. This category saw a significant recovery with a 0.95 km² (112.8%) increase between 2016 and 2023, indicating a partial reversal of earlier losses. Homestead areas saw a notable decrease of 2.55 km² (10.21%) from 1996 to 2023, with a considerable drop of 5.16 km² (40.47%) from 1996 to 2007. However, there was some recovery from 2007 to 2016 and a modest increase of 2.37 km² (32%) from 2016 to 2023, reflecting a fluctuating trend in homestead development. Vegetation cover faced a continuous decline, with a total loss of 9.94 km² (39.8%) over the period. The most significant reductions occurred from 2007 to 2016 (−5.04 km², −44.36%) and from 2016 to 2023 (−3.37 km², −50.4%), highlighting a persistent trend of vegetation cover loss driven by built-up area and homestead expansion (Figure A1 and Figure A2). Overall, the data illustrate a pattern of substantial built-up area growth, coupled with a consistent decrease in vegetation cover and a variable trend in water and marshland areas. The changes reflect ongoing urbanization pressures and their impact on the natural environment over nearly three decades.

3.2. Spatiotemporal Pattern of LST Changes

Figure 4 illustrates the distribution of LST in the Matara City area for the years 1996, 2007, 2016, and 2023. The distribution patterns for each year reveal how temperature ranges and variability have evolved over time.

Table A3. Descriptive statistics of LST (°C).

Year	Minimum	Maximum	Mean	Standard Deviation
1996	22.37	29.18	24.66	0.82
2007	24.11	31.64	26.36	1.07
2016	15.46	23.31	19.70	1.22
2023	14.05	27.21	20.06	3.63

provides a summary of the descriptive statistics for the LST values across these years, highlighting key measures such as the minimum, maximum, mean, and standard deviation.

According to Table A3, from 1996 to 2023, LST data for Matara show notable trends. Initially, in 1996, temperatures were relatively stable, with a narrow range and low variability. By 2007, temperatures rose with increased variability. In 2016, the minimum temperature dropped significantly, and the mean decreased, though variability slightly increased. By 2023, the minimum temperature fell further while the maximum increased, leading to a higher mean and a substantial rise in variability, indicating more extreme temperature fluctuations and suggesting greater climatic instability over the period.

In 2023, the LST in Matara was notably higher in the city core and its surrounding wards, particularly in the northern and eastern areas. This pattern reflects the influence of urban expansion, as evidenced by the development of road networks and growth nodes. The increase in surface temperatures in the core area correlates with urban growth and a reduction in vegetation cover, especially in the southern coastal region. Additionally, LST has risen in densely populated areas and surrounding urban patches, highlighting the impact of urbanization on temperature increases.

3.3. Behavior of Urban Heat Island Index

Figure 5 illustrates the variations in the UHI effect in the Matara City area for the years 1996, 2007, 2016, and 2023. The UHI stack profile depicts the changes in UHI values from the city center to the surrounding hinterland areas. The data for Matara City from 1996 to 2023 shows a significant upward trend in the UHI Index, reflecting an intensification of the urban heat island effect over the years.

In 1996, the UHI Index was relatively low, with a high UHI value of −0.8 and a low UHI value of −7.6, indicating that urban areas were cool. This suggests that, at that time, the impact of urban development on local temperatures was minimal. By 2007, the high UHI Index had risen dramatically to 7.11, and the low UHI Index had moved closer to zero at −0.41. This shift indicates an initial phase of significant warming in urban areas, likely due to increased construction, reduced green spaces, and greater energy consumption associated with urban growth.

The trend continued into 2016, with the high UHI Index slightly increasing to 7.19, while the low UHI Index improved marginally to −0.67. Though there was a slight rise in the high UHI Index, the marginal change in the low UHI Index suggests that rural areas were also warming, but urban areas were experiencing a more pronounced heat effect.

By 2023, the high UHI Index surged to an alarming 21.64, reflecting a substantial increase in the temperature difference between urban and rural areas. Concurrently, the low UHI Index had risen to 8.53, indicating that rural areas had also experienced warming, though the disparity between urban and rural temperatures had widened significantly. This dramatic rise in the high UHI Index points to a marked intensification of the urban heat island effect, driven by accelerated urbanization, expansion of impervious surfaces, and other factors that exacerbate heat retention in urban areas.

Overall, the data reveal a clear and troubling trend: urban wards in Matara City have become increasingly warmer compared to their rural surroundings over the past few decades. This growing UHI effect underscores the need for targeted urban planning and mitigation strategies to address the impacts of urban heat and enhance environmental sustainability.

3.4. Spatiotemporal Pattern of NDVI and NDBI

Figure A3 illustrates the spatial pattern of NDVI during four distinct periods, namely 1996, 2007, 2016, and 2023. In NDVI maps, healthy vegetation typically appears in shades of dark green, while sparse or stressed vegetation may appear in shades of light green. NDVI maps can reveal spatial patterns of vegetation distribution across the mapped area. Clusters of high NDVI values indicate areas with dense vegetation cover, while low NDVI values may highlight regions with limited vegetation. According to Figure A3, in 1996, high vegetation cover was at the Matara City area, especially in the northeastern and western parts of the city. Then, with the rapid urban development in 2007, vegetation cover of the northwestern part of the city has been decreasing. Rich vegetation cover can be seen near the areas adjacent to beaches, marshlands, and water bodies. In 2016, moderate vegetation cover can be identified around the city center and when considering the year 2023, it could be observed low-density vegetation cover around the city area and moderately density vegetation in the northeastern and western part of the city.

Table A4. Descriptive statistics in NDVI.

Year	Minimum	Maximum	Mean	Standard Deviation
1996	−0.155	0.450	0.281	0.071
2007	−0.123	0.459	0.290	0.073
2016	−0.179	0.557	0.345	0.090
2023	−0.191	0.559	0.340	0.094

outlines the trends in the NDVI for Matara City from 1996 to 2023. Over this period, the NDVI values reveal a notable shift in vegetation characteristics. Analyzing these minimum values reveals a trend of decreasing vegetation in the least vegetated areas. Specifically, the minimum NDVI values have worsened from −0.155 in 1996 to −0.191 in 2023. This decline indicates that the areas with the lowest vegetation cover have become even less vegetated over time, suggesting a reduction in green spaces or degradation of vegetation in the most sparsely vegetated regions of the city. This trend could reflect urban expansion or environmental changes leading to reduced vegetation in these specific areas.

Figure A4 demonstrates the NDBI variations in the Matara City area in 1996, 2007, 2016, and 2023. Due to the presence of a densely built-up area with the road network, the highest distributed NDBI was reported around the city center along the coastal belt, central, and southern parts of the city area from 1996 to 2023, whereas the lowest NDBI was reported on the northeastern side of the city due to the presence of dense vegetation.

Table A5. Descriptive statistics in NDBI.

Year	Minimum	Maximum	Mean	Standard Deviation
1996	−0.312	0.159	−0.133	0.065
2007	−0.303	0.137	−0.138	0.065
2016	−0.328	0.204	−0.137	0.068
2023	−0.325	0.259	−0.138	0.080

provides an analysis of the NDBI for Matara City across the following selected years: 1996, 2007, 2016, and 2023. The data show a gradual increase in both the minimum and maximum NDBI values over this period. The minimum NDBI values have improved slightly from −0.312 in 1996 to −0.325 in 2023, and the maximum values have increased from 0.159 to 0.259, indicating a growth in built-up areas and possibly more extensive urbanization. The mean NDBI has remained stable at approximately −0.138, suggesting that the average intensity of built-up areas has not fluctuated significantly. However, the standard deviation has increased from 0.065 in 1996 to 0.080 in 2023, which points to growing variability in built-up areas throughout the city. This suggests that while the overall extent of built-up areas has increased, there is more variation in how these areas are distributed within Matara City.

3.5. Correlation Between LST, NDVI, and NDBI

Figure 6 illustrates the correlations between different environmental variables in Matara from 1996 to 2023, specifically focusing on the relationships between the NDVI and LST, NDVI and the NDBI, and NDBI and LST. For NDVI and LST, the correlation values show considerable fluctuation over the years. In 1996, the correlation was quite low at 0.1075, indicating a weak relationship between vegetation health and land surface temperature. This relationship strengthened to 0.3271 in 2007, suggesting a moderate positive association where higher vegetation was somewhat linked with higher temperatures. However, by 2016, the correlation decreased slightly to 0.2743 and fell further to 0.0754 in 2023, reflecting a weakening and nearly negligible association between NDVI and LST over time.

In contrast, the correlation between NDVI and NDBI, which measures built-up areas, consistently remained high across the years, starting at 0.6559 in 1996 and reaching 0.7027 by 2007. But, it decreased a little by reaching 0.6721 in 2023. Although a high correlation indicates a strong and relatively stable negative relationship between vegetation and built-up areas, higher vegetation density is consistently associated with lower levels of built-up land throughout the observed period.

The correlation between NDBI and LST demonstrates significant variability as well. It started at 0.2053 in 1996, suggesting a weak relationship between built-up areas and land surface temperature.

By 2007, this correlation increased to 0.5283, indicating a moderate to strong positive relationship, where more built-up areas were associated with higher temperatures. The correlation decreased to 0.2943 in 2016 but then rose again to 0.3374 in 2023, reflecting a moderate association between built-up areas and temperature with some fluctuation over time.

Overall, while the relationship between NDVI and LST weakened over the years, indicating less influence of vegetation on temperature, the correlation between NDVI and NDBI remained strong, pointing to a consistent inverse relationship between vegetation and built-up areas. The connection between NDBI and LST showed moderate variability, suggesting a complex interplay between urbanization and temperature changes.

3.6. Landscape Dynamics Along Urban–Rural Gradient Zones

The derived 47 URGZ map, which illustrates the distribution of FBA and FVC within a 210 m radius from the city center in the Matara DSD area (Figure 7), highlights a clear pattern: FBA decreases with increasing distance from the city core, while FVC rises as one moves away from the center. The highest mean LST values, ranging from 29 °C to 27 °C, were observed in URGZs 1 to 10, close to the city center. In contrast, the lowest mean LST values, between 24 °C and 23.7 °C, were recorded in URGZ 40 to 47, situated further from the urban core.

Figure 8 shows the spatial pattern of mean LST and the FBA and FVC along 47 URGZs. From 1996 to 2023 in the Matara DSD area, the data on FVC and FBA demonstrate evolving urban and vegetation patterns. In 1996, central zones were highly urbanized with very low FVC (0% to 0.2%) and high FBA (up to 5.9%). Vegetation cover increased to a peak of 5.0% in intermediate zones but decreased again in the outermost areas to 1.2% and lower. By 2007, FVC remained low, ranging from 0.0% to 5.5%, with the highest cover in less urbanized outer zones. FBA was notably high in the central areas (up to 8.0%) and decreased sharply towards the periphery. In 2016, FVC showed a slight decrease, from 0.0% to a maximum of 4.5%, with minimal cover in central zones (often around 0.0% to 0.3%) and higher values in less urbanized areas. FBA was high in the central zones (up to 7.7%) and lower in outer areas.

By 2023, FVC slightly increased to a range of 0.0% to 4.9%, with the highest values still in the less urbanized outer zones. Central zones continued to have very low FVC (below 1%), while FBA decreased overall, with the highest values in central zones (up to 7.2%) and minimal built-up areas in the periphery. This indicates a trend of reduced urbanization intensity in outer zones and a modest increase in vegetation cover over the years. The data from 1996 to 2023 for the Matara DSD area reveal a dynamic relationship between the FBA and FVC. The FBA values show an initial increase from 0.5594 in 1996 to 0.6847 in 2007, indicating growing urbanization. Concurrently, FVC rises from 0.0057 to 0.2513, reflecting a modest increase in vegetation despite urban expansion.

However, the trend shifted by 2016, with FBA decreasing to 0.4532 and FVC significantly rising to 0.5383, suggesting a period of reduced urban intensity and increased vegetation cover. The reason behind this may be attributed to climate change, as 2016 was possibly a cooler and rainy year compared to 2007. This cooler climate could have positively influenced vegetation condition and health, leading to an increase in the area of vegetation cover, which is reflected as an increased FVC in 2016 since FVC was extracted upon the NDVI. By 2023, FBA increases again to 0.5626 while FVC drops to 0.1141, showing a return to higher urbanization with a slight reduction in vegetation. This pattern suggests a generally inverse relationship between FBA and FVC, where higher urbanization typically corresponds with lower vegetation cover. The fluctuations in the data indicate that while this inverse relationship holds over time, there are periods where both FBA and FVC increase simultaneously or where the trend reverses, reflecting a complex interplay between urban development and vegetation dynamics. Figure 9 illustrates the distribution patterns of mean LST, mean NDVI, mean NDBI, and mean ECI across gradient zones for the years 1996, 2007, 2016, and 2023. A relatively higher mean LST was observed from the urban core, extending up to 1.2 km outward. The highest mean LST was recorded in 1996 and 2007, followed by a slight decrease in 2016, before rising again by 2023. Across all years, there was a significant drop in LST around 6.5 km from the core, which then increased further outward.

The mean NDVI also showed a notable increase from 0.20, extending up to 1.2 km in all years, and reached approximately 0.35 in the rural periphery, which is around 9.6 km from the urban core. The mean NDBI increased each year in almost all gradient zones, with significant growth observed in 2023. The highest mean NDBI was reported from the urban core extending up to 1.2 km outward. From the outskirts toward the city center, the mean ECI decreased in 2006 and 2016. However, in 2016 and 2023, it showed a slight increase up to 9 km from the city core. Overall, the mean ECI exhibited a slight decreasing trend by 2023 compared to 1996.

3.7. Dynamics of LST and FBA Within Urban Core

Figure 10 illustrates the spatial distribution and concentration patterns of mean LST and FBA (%) in the urban core across the 210 m × 210 m grid from 1996 to 2023. The highest mean LST was concentrated in the urban core area and displayed an increasing trend toward the southeast from the city center. The results revealed that the highest mean LST, reaching 31 °C, was concentrated in the southeastern portion of the city core in 2007, while the northwestern portion reported a lower LST. The northern and southern grid areas consistently showed comparatively lower mean LST across all years. The most notable observation is that a small vegetation patch was identified in the central part of the urban core grid in the years 2016 and 2023. Comparatively high FBA was observed in the years 1996 and 2007.

3.8. Environmental Criticality in Matara DSD

The environmental criticality maps for the Matara MC area, derived by integrating LST and NDVI images, are depicted in Figure 11, and descriptive statistics are in

Table A6. Descriptive statistics in ECI.

Year	Minimum	Maximum	Mean	Standard Deviation
1996	27.5	100	42.1	11.7
2005	22.7	100	40.0	13.4
2014	14.5	100	43.1	13.7
2022	10.4	100	35.2	16.4

. In creating the ECI maps, areas covered by vegetation and water bodies were excluded, as these regions are considered to have no criticality [21,22]. This exclusion ensures that the criticality maps accurately reflect the environmental pressures in urbanized and potentially vulnerable areas, providing a clear assessment of environmental criticality in regions where vegetation and water bodies do not influence the results.

The ECI for 1996 across the 47 URGZ in Matara DSD exhibits a notable gradient of environmental stress. The highest ECI value of 75, observed in the most urbanized central zones, signifies significant environmental criticality in areas of dense development. Moving outward from the city center, there is a general decline in ECI values, with many zones showing values between 30 and 54. This decreasing trend indicates lower levels of environmental stress in less developed, peripheral areas. The clustering of values around 36 to 54 in the outer zones reflects moderate environmental pressures, suggesting that while these areas are less stressed compared to the urban core, they still face considerable environmental challenges. Overall, the data highlight a gradient of environmental criticality, with the most severe pressures found in the urban core areas and decreasing towards the outskirts.

In 2007, the ECI across the 47 Urban Gradient Zones in Matara DSD presented a noticeable shift in environmental stress compared to previous years. The highest ECI value of 71 is found in the most urbanized central zones, indicating continued significant environmental criticality in these areas. However, there is a marked reduction in ECI values as one moves away from the city center. Peripheral zones exhibit lower ECI values, with many falling between 26 and 36. This decline suggests an improvement in environmental conditions in the less developed outer areas. The clustering of values around 30 to 35 in these zones reflects relatively lower but still notable levels of environmental stress. Overall, while central urban zones continue to experience high criticality, there is a clear trend towards reduced environmental pressure in the outer zones, indicating some progress in mitigating environmental issues in less developed areas.

In 2016, the ECI for the 47 URGZs in Matara DSD indicated a pattern of moderate environmental stress across the region. The highest ECI value of 67 is observed in the central zones, reflecting substantial environmental challenges in the most urbanized areas. Moving outward, there is a gradual decline in ECI values, with many zones exhibiting values between 30 and 35, suggesting reduced criticality in less developed areas. The peripheral zones show lower ECI values, with some as low as 27 to 28, indicating relatively better environmental conditions compared to the central areas. This distribution highlights a continued concentration of environmental stress in the urban core, while peripheral areas experience less criticality, reflecting a trend of improving environmental conditions outside the most densely built-up zones.

In 2023, the ECI across the 47 Urban Gradient Zones in Matara DSD shows a generally stable pattern with some notable variations. The ECI values in the central zones remain relatively high, with values of around 60 to 64, indicating ongoing environmental stress in the most urbanized areas. Moving outward, there is a slight decrease in ECI values, with many peripheral zones showing values between 31 and 45, suggesting moderately improved environmental conditions as urbanization decreases. A few areas at the extreme periphery report lower values, dropping to 28, reflecting better environmental conditions. This distribution underscores a trend of maintaining high environmental criticality in the core urban areas while experiencing slight improvements in the less densely developed zones, indicating ongoing environmental challenges in central zones but with some progress in less urbanized areas.

Over the years, from 1996 to 2023, the ECI for the 47 URGZs in Matara DSD demonstrates a trend of persistent environmental stress in urban areas with gradual improvements in less developed zones. In 1996, the highest ECI values were observed in the central urban zones, indicating significant environmental criticality. As one moved outward, the ECI values decreased, showing lower levels of stress in peripheral areas. By 2007, the central zones continued to show high ECI values, but there was a noticeable reduction in environmental stress in the outer zones, suggesting some progress. In 2016, while central zones still exhibited high ECI values, indicating substantial environmental challenges, peripheral zones showed further improvement with lower ECI values. By 2023, the central areas maintained high ECI values, reflecting ongoing environmental stress, while peripheral zones saw modest improvements, with some areas reporting notably lower ECI values. Overall, the data reflect a consistent pattern of high environmental criticality in central urban zones and improving conditions in less developed areas over time.

4. Discussion

4.1. Urban Growth and Environmental Health

From 1996 to 2023, significant LULC changes were observed, characterized by rapid urban expansion primarily in the northeast area, with built-up areas increasing by 12.21 km² (48.8%). Vegetation cover continuously declined, losing 9.94 km² (39.8%), driven by urban and homestead growth. These trends reflect the ongoing pressures of urbanization and its impact on the environment over nearly three decades. This trend aligns with previous research in the Galle MC [26] and other cities such as Colombo [16,17,18] and Kandy [19,27]. Similar patterns of landscape transformation were noted, including the rapid conversion of green spaces (GS) and other land use types into impervious surfaces (IS) over the past two decades [8,50]. These studies found minimal changes in water bodies, with most GS and other land use categories being absorbed by IS due to the growth of residential, administrative, and commercial areas. In the Matara DSD, this shift has led to significant urban expansion, resulting in adverse effects such as the formation of urban heat islands and increased environmental concerns.

From 1996 to 2023, LST data for Matara reveal a significant upward trend. Initially stable with low variability, temperatures began to rise and show increased variability by 2007. By 2016, the minimum temperature decreased while the mean temperature also fell, though variability increased slightly. By 2023, the minimum temperature had further declined, while the maximum temperature and mean temperature both increased, leading to substantial temperature fluctuations and greater climatic instability. This trend is particularly pronounced in the city core and surrounding areas, especially in the north and east, where urban expansion and reduced green cover have intensified surface temperatures. The rise in LST in densely populated and urbanized areas reflects the impact of urban development on the local climate.

The study area has experienced a rise in LST, driven by urban population growth, increased transportation, the expansion of impervious surfaces, and building density. Over 27 years, declining vegetation cover has led to a critical environmental condition, as measured by the NDVI and ECI. Urban expansion in Matara City is expected to increase overall LST and UHI intensity in the near future. This trend is supported by a declining NDVI and an increasing NDBI. While LST and NDVI have a low negative correlation, indicating that NDVI partially explains changes in LST, the inverse correlation between LST and NDBI suggests that IS directly contributes to higher LST and UHI intensity. Distribution maps of NDBI and ECI show that BA areas with elevated LST are more vulnerable to environmental stress. The addition of new impervious surfaces in the area is likely to exacerbate LST and ECI vulnerabilities [51].

4.2. Impacts of Changing Landscape Structure on Urban LST

The results found that the growth of the BA came at the cost of reduced VC from 1996 to 2023. The extent of VC decreased from 5.6 km² (32.7%) in 1996 to 1.7 km² (10.9%) in 2023. A 3.57 km² area covered by VC and a 3.8 km² area under other categories were solely transformed into BA within the concerned period. This trend is comparable to the previous research findings observed in the Galle MC [26] as well as in other cities like Colombo [16,17,18], Kandy [19], and Kurunegala [50]. Similar landscape-changing patterns were also observed by Dissanayake et al. (2020), who demonstrated the rapid transformation of GS and other areas into IS over the last two decades [26,27]. The findings indicated that there was minimal or no significant alteration in water bodies. The increase in residential, administration, and commercial establishments leads to the absorption of most GS and other LULC types by the IS. The urban growth in the Galle MC was driven by the swift transformation of GS into IS between 1996 and 2023 [26]. This has resulted in the emergence of detrimental consequences, such as the formation of urban heat islands with increasing EC.

The results of this study provided the evidence of increasing UHI intensity. Between 1996 and 2023, the UHI intensity index showed a considerable increase (14.58 °C) during 2007–2023. The URGZ analysis results also demonstrated that UHI intensity is concentrated within a 1 km² radius from the city core. The spatial and temporal distribution patterns of mean LST and FBA provide a clear image of the relationship between UHI intensity expansion and BA expansion between 1996 and 2023. But, on the other hand, the mean NDVI has increased by 0.059 within 26 years. The cause of this phenomenon might be attributed to the recent green covers [26]. However, urban expansion in the Matara area may lead to a subsequent rise in UHI intensity further in the near future. The inclining NDBI has further supported this trending pattern. Though LST and NDVI have a low negative relationship and values indicate that NDVI can partially explain changes in LST, the inverse correlation between LST and NDBI indicates that BA has a direct effect on raising the overall LST and UHI intensity.

4.3. Planning Initiatives for Safeguarding Environment and Human Health

Given the expansion of BA and the corresponding decrease in VC, city planning agencies such as the Urban Development Authority (UDA) and the Municipal Council (MC) should adopt sustainable urban planning strategies. The proposed UDA development plan for Matara MC (2019–2030) recommends the establishment of green belts along the beachside adjacent to the Matara–Tangalla new road, as well as within the congested urban areas in the city core [22]. The UDA building regulations include provisions for urban agriculture and rooftop greening on high-rise buildings. Based on these findings, we propose enhancing urban greening activities, particularly in core areas, through initiatives such as rooftop agriculture and the establishment of green belts alongside main roads. Previous scholars also have advocated for integrating green concepts into urban development [16,18,19,26,27]. Recommended measures include establishing green belts and promoting urban agriculture, as VC can mitigate LST through cooling effects and evapotranspiration [16,19,20,26]. These strategies should target high LST and UHI intensity hotspots, particularly along busy roads and dense settlement zones in the city core. Additionally, specialized planning initiatives should be implemented to reduce temperatures around the main road network affected by automobile emissions.

4.4. Constraints and Future Perspectives

While this research focused on quantifying the linear relationship between LST and urban expansion, further investigation is needed to explore the nonlinear dynamics between LST and BA, as well as the inverse correlations involving NDVI, NDBI, and ECI. Future studies should incorporate additional data on precipitation and temperature. High-resolution thermal remote-sensing images should be utilized to generate more accurate LULC maps and RS indices enhancing reliability.

Moreover, incorporating socioeconomic factors into the analysis will allow researchers to understand how demographic changes and economic activities influence urban growth patterns and environmental outcomes. Conducting comparative studies in other rapidly urbanizing regions of Sri Lanka could broaden the understanding of urbanization’s effects across different contexts. Lastly, developing predictive models using machine learning techniques could help urban planners forecast future LULC changes and their potential environmental impacts, thereby supporting proactive planning efforts aimed at sustainability and resilience. These recommendations will not only enrich the existing body of research but also provide actionable insights for urban planning and environmental management in rapidly developing cities.

5. Conclusions

• This study examined the spatial and temporal changes in LST and UHI intensity related to LULC alterations in the Matara DSD over 27 years. It highlighted a significant expansion of BA and a considerable decrease in VC from 1996 to 2023. And also, elevation distribution shows that Matara City’s development is largely influenced by low terrain, as most impervious surfaces have emerged in low-lying areas.
• We found that BA increased by 12.21 km² (48.8%), while VC declined by 9.94 km² (39.8%). This rapid urban growth extended into rural peripheries, contributing to a rise in UHI, which increased by 14.8 °C over 27 years. By 2023, newly developed BA exhibited the highest environmental criticality, with mean LST values reaching 25 °C in URGZs within 1 km² near the city core.
• Regression analysis showed a strong positive correlation between the NDBI and LST, emphasizing the significant impact of built-up area expansion on LST. To mitigate the adverse effects of UHI and environmental criticality, green urban planning strategies are crucial, particularly in identified Urban Gradient Zones.
• Urban land use planning and regulations should be updated based on these findings, and this methodology, utilizing high-resolution satellite data and validation techniques, could be applied to other urban areas in Sri Lanka. Given the scarcity of similar research, this study provides valuable insights into urban planning enhancing human security and ecosystem health.