The coastal city of Galle, the capital of the southern province, of Sri Lanka is bounded to the south and southwest by the Indian Ocean which is located in 6°.10ʹ north latitude and 80°.13ʹ east longitude. It serves as both the district and provincial capital and is regarded as the largest city in the Southern Province. Since the southern province is home to the primary business hub, Galle is a bustling city where many inheritances live and work (UDA, 2019). In the study, Galle MC (Figure 1) was selected and it lies within the wet zone with an average temperature of 28–30°C and 1,525–1900 mm average rainfall. Except for a few isolated peaks of about 60–160 feet, other areas consist of flat terrain. Also, from the agro-climatological perspective, the MC belongs to the wet zone (Dissanayake, 2020). Especially, Galle is the third municipal council in Sri Lanka, founded following the terms of the urban council statute of 1865 (Abeyweera and Kaluthanthri, 2018). In 1988, Galle City was named a World Heritage City in Sri Lanka by the United Nations.

The Landsat satellite images with a minimum cloud cover of less than 5% or without any clouds during the wet and dry seasons were downloaded via the United States Geological Survey (USGS) website (USGS, Accessed on 12 November 2023). All images were projected into WGS84/UTM 44N to maintain spatial consistency. The whole study area belongs to path 56 and row 141. The summary of all the data in the two sensors is included in Table 1.

To establish a thorough link between the acquired data and the biophysical properties of the landscape, preprocessing must be done before image classification and change detection (Land-resources, 2023). The identification of minute variations in surface reflectance and LULC is necessary for this investigation. The thermal bands (TM band 6 and OLI band 10) were employed for the retrieval of LST after radiation and atmospheric corrections (Landsat 8 (L8), 2016). To process and analyze the data, Arc Map 10.8 (ESRI, 2020), MS Office Excel (Microsoft, 1988), and Origin Pro data analysis and graphing software (Origin Lab Corporation, 2023, MA, United States) were used.

Several steps were followed to create LU/LC maps including image preprocessing, classification, accuracy checking, and change detection. Retrieval of NDVI, NDBI, NDWI, LST, UHI, and ECI was performed using standardized equations. The technical flowchart explaining the combination of image preprocessing, image classification, retrieval of RS-based indices, and other geospatial and statistical analysis is presented in Figure 2.

Among the four LULC classes, the findings indicate that impervious surface has shown the most significant expansion over the past 26 years. Throughout the study period, impervious surface areas grew mostly gradually, from 5.5 km² (31.7%) in 1996 to 12.8 km² (74.1%) in 2022 with indicating a 7.3 km² growth. This rapid growth of impervious surface occurred at the continuous expense of green surface, which decreased from 5.6 km² to 1.7 km² (−3.9 km²) during the concerned period. In addition to green surface, 3.4 km² areas including paddy lands and homesteads categorized under “other” have also been converted into other LULC over the last 26 years. Considerable changes in the water bodies, could not be detected since the MC area consists only of a few canals and marshes. The overall accuracy rates exceeded the recommended minimum standard of 75%, indicating a reliable classification of LULC for each year. The Kappa coefficient was 0.84 in 1996, 0.82 in 2005, 0.80 in 2014, and 0.78 in 2022 as Supplementary Table A1. The classified LULC maps in the Galle MC area for 1996, 2005, 2014, and 2022 are visualized in Figure 3 and quantified in Table 3.

Impervious surface comprised 5.5 km² of the total land area in 1996 and by 2014 it has gradually increased up to 9.7 km². However, by 2022, it can be seen how new impervious surface have been created in areas with green surface and other LU types as expanding impervious surface into 12.8 km². The LULC transition during the study period is visualized in Supplementary Figures A2, A3. Based on those figures and statistics, approximately 3.5 km² of green surface and 3.8 km² of other land use types were converted into impervious surface over the last 26 years. The rapid expansion can be detected in the central and southern areas of MC that are near to city core. These spatial-temporal variations in the LULC classes impacted the ecosystem health and LST and UHI intensity during the study period.

Figure 4 depicts the spatial distribution of LST in the Galle MC area for the years 1996, 2005, 2014, and 2022. In 1996, high LSTs were concentrated in the city core. LST varied from 21.06°C to 32.4 C in 1996, with a mean of 26.4 C. However, by 2005, areas with high LST had substantially grown towards the eastern and central parts and along the coastline, following the expansion pattern of impervious surface. During this period, the mean LST has increased to 28.9°C, showing an increase of 2.5 C. This trend continued until 2022, with the maximum LST increasing from 33.2°C to 33.7 C and the mean temperature rising from 28.9°C to 29.1 C between 2005 and 2022, although the minimum LST fluctuated slightly around 2014 as shown in Supplementary Table A2. High LST began to extend in and around the city center area in 2014. In 2022, the city core and its surrounding areas had higher LST values, which were dispersed mostly to the northern and eastern parts of MC. The link between LST and the pattern of impervious surface expansion is well demonstrated along road networks and growth nodes. LST increased in the core areas, while the southern beach area experienced significant LST changes due to impervious surface expansion and a loss of green surface. Throughout the entire period, most green surface in the northeastern and northwestern regions have maintained comparatively low LST.

Results further revealed that the UHI intensity in the MC area has also increased significantly over the 26 years as shown in Figure 5. The data on UHI intensity over the years indicates a substantial increase in both the minimum and maximum UHI intensity values, reflecting a growing severity of the UHI effect. In 1996 the UHI intensity ranged from a minimum of −2 C to a maximum of 1°C. This suggests that surrounding areas of the Impervious Surface core were cooler than the city areas (negative UHI intensity), while other areas experienced only a slight UHI effect. The UHI intensity increased significantly by 2005, with a minimum of 4.5 C and a maximum of 9 C. This indicates that most areas within the city were warmer than their rural counterparts, with a marked increase in the intensity of the UHI effect compared to 1996. In 2014 minimum UHI intensity further increased to 6°C, while the maximum remained at 9 C.

This demonstrates a continuing trend where the entire impervious surface is experiencing higher temperatures relative to rural areas, with the minimum intensity rising closer to the maximum intensity of the previous period. The minimum UHI intensity reached 7 C by 2022, and the maximum increased to 10.5 C. This data suggests that the UHI effect has become even more pronounced, with both the minimum and maximum intensities showing significant increases. The impervious surface is now consistently experiencing higher temperatures, and the peak intensity of the UHI effect has reached new highs. Overall, this progression indicates a consistent and substantial increase in UHI intensity over the 26 years, highlighting the growing impact of impervious surface and the associated changes in LULC on local climate conditions. Figure 6A [1, 2, 3, 4], (c) [8, 9, 10] further demonstrates that the majority of weak UHI areas and low LST areas consist of existing green surfaces from 1996 to 2022. Additionally, areas with high LST and changes in the HUI correspond to high-density impervious surface development during this period, as shown in Figure 6B [5, 6, 7], (d) [11, 12, 13]. Flooring, roofing, and concrete structures have disrupted the surface energy balance by reducing evapotranspiration and increasing sensible heat in the city core. This is the main reason for the day and night UHI intensity anomalies between impervious surface and other LULC types.

Supplementary Figure A4 illustrates the derived NDVI maps during four distinct periods, namely, 1996, 2005, 2014, and 2022. Supplementary Table A3 also summarizes descriptive statistics. The minimum NDVI values have increased from −0.53 in 1996 to −0.11 in 2022. This suggests a reduction in the number of areas with very low or negative vegetation indices. The maximum NDVI values have slightly decreased from 0.71 in 1996 to 0.51 in 2022. This indicates that the areas with the highest green surface or healthiest green surface have seen a decline, suggesting potential degradation of the most vegetated areas. The mean NDVI values have decreased from 0.41 in 1996 to 0.26 in 2022. This overall decline in the mean NDVI indicates a general reduction in environmental health and green surface over the years. The results suggest that over the period from 1996 to 2022, there has been a noticeable decrease in vegetation health and green surface density. Overall, mean and maximum NDVI values have declined, pointing to a general degradation in green surface quality and coverage as a result of the expansion of impervious surface.

The minimum NDBI values (Supplementary Figure A4; Supplementary Table A4) have increased from −0.57 in 1996 to −0.37 in 2022. This suggests that the areas with the lowest NDBI values. The maximum NDBI values have decreased slightly from 0.46 in 1996 to 0.19 in 2022. The NDBI data indicates that from 1996 to 2022, there has been a complex pattern of changes in impervious surface. The minimum NDWI values have increased from −0.62 in 1996 to −0.49 in 2022 (Supplementary Table A5). This suggests that the driest areas have become slightly less dry over time, potentially indicating a decrease in extreme water scarcity in certain areas. The maximum NDWI values have decreased from 0.58 in 1996 to 0.21 in 2022. This indicates that the areas with the highest water content or surface water presence have seen a significant reduction, suggesting a decrease in the extent or presence of surface water bodies.

Regression results derived R ² 0.1962 between LST and NDVI for year 1996 as shown in Supplementary Figure A5. This suggests a weak positive relationship, indicating that only about 19.62% of the variation in LST can be accounted for by the NDVI. But, in 2005 the relationship is stronger with R ² 0.3241 compared to 1996, with about 32.41% of the variation in LST being explained by NDVI. This suggests a moderate positive correlation. By 2014 The relationship is quite weak with R ² 0.1084, with only 10.84% of the variation in LST explained by NDVI, indicating a low positive correlation. In the year 2022 with R ² 0.2571 there is a moderate relationship, with 25.71% of the variation in LST being explained by NDVI, which is stronger than in 2014 but weaker than in 2005. Overall, these R ² values suggest that the relationship between LST and NDVI varies over time, being strongest in 2005 and weakest in 2014. The values indicate that NDVI can partially explain changes in LST.

The regression results derived R ² 0.4570 between LST and NDBI for the year 1996. This indicates a moderate positive relationship between LST and NDBI. About 45.70% of the variation in LST can be explained by NDBI. In 2005 also the relationship remains moderately strong, with 45.20% of the variation in LST being explained by NDBI with R ² 0.4520. This is very similar to the relationship in 1996. In year 2014 with R ² 0.3519 the relationship is weaker compared to 1996 and 2005, indicating a moderate but decreased positive correlation. In year 2022, the relationship is moderately strong, with R ² 0.4639 of the variation in LST being explained by NDBI, slightly stronger than in previous years. Overall, these R ² values suggest that the relationship between LST and NDBI is consistently moderate, with NDBI explaining a substantial portion of the variation in LST. The values indicate that impervious surface, as measured by NDBI, have a significant and relatively stable influence on LST over the years, with the strongest relationship observed in 2022.

Overall, the R ² values suggest that the relationship between LST and NDWI is generally weak, with NDWI explaining only a small portion of the variation in LST. The strongest relationship was observed in 2005 (R ² 0.2589), indicating that water presence, as measured by NDWI, had a more substantial influence on LST in that year compared to other years. The regression results revealed that R ² values indicate the relationship between NDVI and NDBI has strengthened over time, with the correlation becoming very strong by 2022 with R ² 0.7519. This trend suggests that as impervious surface areas have increased, green surface has correspondingly decreased, reflecting the significant impact of urban expansion on green surface. The correlation matrix between LST with NDVI, NDBI, NDWI over the four-time period are graphically represented in Figure 7 below.

Derived 28 URGZ, showing the distribution of impervious surface and green surface in the Galle MC area within a 210-meter radius from the city center (Figure 8). This URGZ distribution map clearly illustrates that FIS decreases as one moves away from the city core, while FGS increases with distance from the city center. The map shows that impervious surface density decreased from the urban core toward the rural periphery from 1996 to 2022, while green surface density increased, particularly from 2005 onwards. Figure 9 shows the spatial pattern of mean LST and the FIS and FGS along 28 URGZs. The highest mean LST value was recorded in URZ₁ near the city center. The highest mean LST, ranging from 28.3°C to 27.1°C, was concentrated within URGZs 1 to 7. In 1996, the lowest mean LST was recorded in URGZs 24, 25, and 26, with temperatures of 25.4°C, 25.5°C, and 25.6°C. The mean LST in URGZ₁ was 28.3°C in 1996 and increased to 31.7°C, 30.3°C, and 30.7°C by 2005, 2014, and 2022, respectively. Additionally, the mean LST of all other URGZs increased from 1996 to 2022. The lowest mean LST in 2005 was reported in URGZ₂₅ at 27.7°C. In 2014, the lowest mean LST was reported in URGZ ₂₆ at 26.1°C. In 2022, the lowest mean LST was concentrated in URGZ₂₄ at 27.3°C.

The FIS within all URGZs has been rapidly increasing from 1996 to 2022. The highest FIS reported in 1996 years was 10.2% in URGZ₃. In 2005, the highest FIS was reported in URGZ₉ at 8.2%. In 2014, the highest FIS was reported in URGZ₈, as at 7.8%. By 2022, URGZ ₉ reported the highest FIS at 7.5%. The lowest FIS in 1996 was located in URGZ₂₁ at 0.2%. In 2005, the lowest FIS (0.27%) was reported in URGZ₂₃. In 2014, the lowest FIS was located in URGZ₂₄ at 0.33%. In 2022, the lowest FIS was reported in URGZ₂₆ at 0.36%. In 1996, the highest FGS was located in URGZ₁₂ at.

11.2%. By 2005, URGZ₁₃ reported the highest FGS at 13.4%. In 2014, the highest FGS was located in URGZ₁₅, recorded at 13.5%. By 2022, the highest FGS was reported in URGZ₁₄ at 11.9%. The lowest FGS in 1996 and 2022 was recorded in URGZ_28,27, respectively. In 2022, the FGS was reported at 0.05% for URGZ₂₈ and 0.06% for URGZ₂₇. Overall, the R ² values show that while there is a strong relationship between mean LST and the fraction of impervious surfaces along URGZ_1-28, this relationship has weakened little, particularly in the year 2022 (R ² 0.2631). Conversely, the mean LST showed a moderate negative relationship with the FGS across three time periods except in 2014.

Figure 10 shows multi-directional and multi-temporal profiles of the mean LST and FIS for the years 1996, 2005, 2014, and 2022. The mean LST shows a decreasing pattern from the city center to URGZ₁₇ in the north-south direction for all years from around 32°C–26°C. But, for the northeast-southwest and northwest-southeast directions from the city center, the mean LST shows a decreasing trend with some fluctuations. In all three directions and across temporal dimensions, FIS is showing a rapidly increasing trend, particularly noticeable in URGZ_1-4.

The spatial distribution and concentration patterns of mean LST, FIS (%), and FGS (%) in the urban core across the 210 × 210 m grid zone from 1996 to 2022 are shown in Figure 11. The results showed that the highest mean LST, exceeding 33°C, was concentrated on the left side of the city core in both 1996 and 2005. However, in 2014 and 2022, the pattern changed to concentration and no longer showed dispersion. Additionally, in 2014 and 2022, the LST within the urban core (210 m × 210 m) grid areas showed a considerable decrease compared to the situation in 1996 and 2005. The beachside consistently exhibited lower mean LST, ranging from 26°C to 30°C, during all four time points. During all four time points, a high fraction of IS was distributed along the northern side of the city core. Comparatively low FIS was observed in the southwestern part of the urban core until 2014. After 2014, a few vegetation patches occupied that area. Due to Galle City inherently evolving as a fortress city and being a UNESCO World Heritage site, the urban core area, extending 210 m × 210 m, is primarily occupied by the fort and other historical buildings, rather than new constructions and green surface. Statistical analysis revealed a low positive but increasing trend in the relationship between mean LST and FIS within Galle MC, with an R ² value of 0.0784 in 1996 and 0.2919 in 2022 (Supplementary Figure A6). Overall, the relationship between mean LST and the fraction of green surface showed fluctuations, with the strongest correlation occurring in 2014 (R ² 0.1927).

The growth patterns of the urban landscape of the Galle MC were derived for three distinct periods: 1996–2005; 2005–2014; 2014–2022 (Supplementary Figure A7; Supplementary Table A6). The expansion of Galle City is mostly compacted, making the IS more continuous, as illustrated in fragmentation metrics, when one considers the compact growth as a mix of infilling and edge–expansion. It is also a unique feature that no outlying patches can be identified in the IS within all three-time period. The LEI pattern was determined based on edge expansion and infilling. The % of edge-expansion patches were 99.9%, 95.46%, and 99.46% in the three periods, indicating a rising tendency. An overview of the three periods reveals that edge expansion is higher than infilling. The proportion of infilling patches was 0.0054%, 4.53%, and 0.55% during the 3 decades. From 2005 to 2014, there was notable growth, and from 2014 to 2022, there was an overall pattern of first increasing and then decreasing. Throughout all periods, the quantity and size of edge-expansion patches were continuously greater than those of infilling patches.

Environmental criticality maps for the Galle MC area, derived by combining LST and NDVI images in the Galle MC area following the steps described in the methodology section, are illustrated in Figure 12. When deriving the ECI maps, all areas covering green surface and water bodies were excluded, as these areas have no criticality. According to the derived maps, the EC has increased from 1996 to 2022, with very high and high critical areas concentrated and expanding outward from the city core. In the year low EC areas were reported in URGZ₂₄ as 0.27% while zone 8 reported high EC areas as 7.35%. In the year 2005 low EC area coverage was occupied in GRGZ₂₃ at 0.33% and high EC area coverage was reported by zone 9 (8.44%). In 2014 low spatial coverage of EC was occupied in URGZ₂₄ (0.30%) while high coverage could be seen in zone 10 (7.76%). Overall, low or very low EC areas could be observed on the northwestern and northeastern periphery of MC in all four time points. The ECI typically follows the distribution pattern of LST (Figure 4) and NDBI values (Supplementary Figure A4) in the study area.

The results found that the growth of the impervious surface came at the cost of reduced green surface from 1996 to 2022. The extent of green surface decreased from 5.6 km² (32.7%) in 1996 to 1.7 km² (10.9%) in 2022. A 3.57 km² area covered by green surface and a 3.8 km² area under another category was solely transformed into impervious surface within the concerned period. This trend is comparable to the previous research findings observed in the Galle MC (Dissanayake, 2020) as well as other cities like Colombo (Fonseka et al., 2019; Senanayake et al., 2013; Ranagalage et al., 2017), Kandy (Ranagalage et al., 2018a), Kurunegala (Ranagalage et al., 2020), and Nuwara Eliya (Ranagalage et al., 2019). Similar landscape changing patterns were also observed by Dissanayake (2020) which demonstrated the rapid transformation of green surface and other areas into impervious surface during the last 2 decades (Dissanayake, 2020). 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 green surface and other LULC types by the impervious surface. The urban growth in the Galle MC was driven by the swift transformation of green surface into impervious surface between 1996 and 2022. 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 evidence of increasing LST and UHI intensity. The mean LST in the Galle MC has experienced a 2.7 C increase between 1996–2022, as shown in Figure 4. Also, the UHI intensity index has shown a considerable increase from 1°C to 10°C during the concerned period (Figure 5). The URGZ analysis results also demonstrated that UHI intensity is concentrated within a 1.5 km² radius from the city core. The spatial and temporal distribution patterns of mean LST and FIS provide a clear image of the relationship between UHI intensity expansion and impervious surface expansion between 1996 and 2022. On the other hand, the mean NDVI has decreased by 0.15 within 26 years. Especially, afterward 2005 increasing LST zones has expanded parallel to the impervious surface expansion direction. The cause of this phenomenon might be attributed to the recently developed urban areas, as indicated by Dissanayake (2020). Urban expansion in the Galle MC may lead to a subsequent rise in the overall LST and UHI intensity shortly. The declining NDVI and inclining NDBI have further supported this trending pattern. Though, LST and NDVI have low negative relationship and values indicate that NDVI can partially explain changes in LST, the inverse correlation between LST and NDBI indicates that impervious surface have a direct effect on raising the overall LST and UHI intensity in the study area. The distribution maps of NDBI and ECI also indicate that impervious surface areas with higher LST are more susceptible to environmental criticality. Newly added impervious surfaces in the study area may lead to an increase in the vulnerability of LST (Dissanayake, 2020) and ECI.

The findings indicated that the Galle MC underwent considerable landscape transformations during 1996–2022, as demonstrated by the rapid expansion of impervious surface. Throughout the 26 years, the impervious surface expanded by 7.34 km² (Table 3). The rapid expansion of impervious surface occurred more during 1996–2005. Dissanayake (2020) also found that the Galle MC displayed a comparable trend of urban expansion, which aligns with the findings presented in this study (Dissanayake et al., 2019). The URGZ analysis results also proved that FIS decreases as one moves away from the city core, while FGS increases with distance from the city center. High fraction of impervious surface could be observed within URGZ_1-9 and it gradually decreased toward rural periphery while green surface density increasing. However, that study discovered that the increase of the built-up area primarily occurred in a north-south direction. The expansion trajectory of Galle City exhibits similarities to other growing cities like Kandy and Kurunegala, where evidence of the UHI effect has been documented (Fonseka et al., 2019; Senanayake et al., 2013; Ranagalage et al., 2017; Ranagalage et al., 2019; Dissanayake et al., 2019; Ranagalage et al., 2020). Though, Ranagalage et al. (2020) discovered that urban expansion in Kurunegala primarily took place in the southern direction, following a linear pattern, in Galle, it is apparent that urbanization occurred mainly along the north-south direction showing a compacted growth as Figure 10 (Ziaul and Pal, 2016). This growth pattern is related to the location of MC as a coastal city, which limits its expansion towards the south which collaborated with the findings of Dissanayake (2020). However, due to the rich urban heritage dating back to the colonial era, the city is linked to other cities such as Matara and Kalutara via the A2 road. Indeed, the urban expansion along the A2 road is also significant. The improved physical and social infrastructure facilities have had a substantial influence on the increasing pattern of urbanization within the concerned period. This is supported by the rising demographic data observed in the Galle MC area between 1996 and 2022. This is further supported by the growth in the population of the Galle urban area, which has risen from 86,568 in 1996 to a subsequent increase of 109,695 in 2022 (Department of Census and Statistics, 2023).

In light of the impervious surface expansion with decreasing green surface pattern and tendency, city planning agencies such as the Urban Development Authority (UDA) and MC should adopt sustainable city planning strategies, such as the implementation of green concepts in urban planning as suggested by other scholars (Senanayake et al., 2013; Ranagalage et al., 2017; Dissanayake, 2020). These strategies may include the establishment of green belts and the promotion of urban agriculture since the green surface has a cooling effect on LST due to its ability to generate cool island effects through evapotranspiration (Ranagalage et al., 2017; Dissanayake, 2020). These measures would be essentially in the high LST and UHI intensity hot spot areas including along busy roads and other dense settlement zones in and around the city core within URGZ_1–9 identified as in (Figure 8). Since transportation networks have greatly affected compacted urban growth, the liner green belts beside the roads will be more effective in reducing the effect of heat emitted from automobiles.

Though this research focused on quantifying the linear relationship between LST and impervious surface expansion, further investigation is necessary to explore the nonlinear relationship between LST and impervious surface, as well as the inverse correlation between NDVI, NDBI, ECI, and urban growth. This analysis should be incorporated with additional data on precipitation and temperature, similar to the suggestions made by Fonseka et al. (2019). Due to time and resource constraints, we did not perform short-term (monthly/annual) trend analysis of UHI. However, to enhance the robustness and reliability of similar studies in the future, we recommend including short-term trend analysis. In future research, high-resolution thermal remote sensing images can be utilized to derive more precise LULC maps and other indices such as LST, NDVI, and NDBI, ensuring improved reliability by mitigating the impact of cloud cover, as observed in Landsat images. Additionally, those indices are not only reliant on the LULC classes as well other variables such as wind speed, solar radiation, and surface moisture, have an impact on the situation (Dissanayake et al., 2019). However, it is not feasible to measure such variables using Landsat data. Furthermore, the measurement of nighttime temperature using Landsat data was unattainable. Similar to the findings of (Dissanayake, 2020) it is crucial to do further research to determine if the Galle MC has reached its maximum capacity in terms of available land for new IS (Dissanayake, 2020).