Assessment of critical thermal characteristics and land surface dynamics of an Indian metropolitan city Open Access

The ceaseless sprawl of urban built-up areas has led to remarkable changes in the regional environment (Han 2020; Cheng & Hu 2023). The effects of urbanization on local climate have been diverse. Intensification of surface and ambient temperature (Nimish et al. 2020; Dutta et al. 2024), increase in extreme precipitation events (Chang et al. 2018; Wang et al. 2021), fluctuations in evapotranspiration (Rim 2009; Gan et al. 2023), altered properties of surface albedo (Tang et al. 2018; Ouyang et al. 2022), etc., have been caused by the urbanization process. Among such changes, the surge in heat waves due to unplanned development directly affects the health of the vast urban population. With significantly rising cases of heat-related mortality (Taylor et al. 2015), urban climate researchers rigorously study the anomalies in local temperature patterns. Even in temperate cities like Paris or London, the risk of heat-related mortality has increased by 70 and 20%, respectively (Mitchell et al. 2016). A study by De Troeyer et al. (2020) showed that mortality rose by 4.9% per 1 °C over Antwerp and by 3.1% over Brussels in Belgium. Another worldwide analysis linked rising heat waves and increased mortality in cities like Delhi, Hong Kong, London, Bangkok, Kyoto, Bangladesh, Osaka, and Berlin (Wong et al. 2013). India, with the largest population in the world, is also a tropical country with a long coastline that witnesses extremely hot and humid weather conditions. Aided by the current rate of urbanization, the heat wave intensity for Indian cities is rising sharply. A few case studies in India have attempted to highlight this. Heat wave-related excess mortality was noted over Ahmedabad in 2010 (Azhar et al. 2014), very high heat stress recurrence over Delhi (Kumar et al. 2022), and a strong relation between severe heat waves and mortality over east coast states was present (Singh et al. 2021). Due to such occurrences, research on the intensification of temperature in city regions is becoming important.

The above-outlined urban climate issue is generally documented by the urban heat island (UHI) effect (Deilami et al. 2018). UHI raises the temperature of cities much higher than the neighboring rural or natural locations and enhances the risk of concentrated heat waves. This magnitude of relative temperature difference from urban center to non-urban fringe can depend on city sizes (Santamouris 2015; Dutta et al. 2023), building arrangement (Boccalatte et al. 2020; Li et al. 2020), rural temperature patterns (Zhou et al. 2010), etc. Vinayak et al. (2022) performed numerical experiments to depict how urbanization can raise thermal discomfort by changing ambient temperature and wind flow characteristics. Numerous studies across the globe have investigated the UHI phenomenon in Tehran (Iran) (Moghbel & Shamsipour 2019), Shanghai (China) (Li et al. 2011), Munich (Germany) (Alavipanah et al. 2015), Abuja (Nigeria) (Koko et al. 2021), Delhi (India) (Kumar et al. 2023), Singapore (Jung 2024), etc. The quantitative analysis of UHI in terms of its areal growth or magnitude is often carried out using land surface temperature (LST) patterns (Rasul et al. 2017; Saha et al. 2021). The LST pattern represents the radiative energy balance of the global surface system. Hence, it is a crucial climatological variable. LST is closely associated with the heat flux between the surface and the atmosphere (Ayanlade & Howard 2019). The changes in such processes indicate climate alterations and can be identified using the LST profile. It is a vital tool for Sustainable Development Goal 13 (Climate Action) as it also reflects the near-surface ambient temperature patterns. Recent research works recurrently use satellite images to produce LST patterns due to their better spatial and temporal resolution than in-situ data collection (Dar et al. 2019; Li et al. 2023).

The UHI studies are often associated with land use land cover (LULC) patterns (Liu et al. 2020; Derdouri et al. 2021). A study in Delhi metropolitan city (India) statistically linked LULC changes and surface UHI increase using geographically weighted regression (Shahfahad et al. 2022). The land cover characteristics can again be procured from satellite image processing. LULC changes can indicate many sensitive phenomena like population growth and decline, migration, rural–urban population patterns, economic policies such as land subsidization, price changes, taxes, and perceptions like history (Acheampong et al. 2018; Genet 2020; Dong et al. 2021). Monitoring the LULC trends can reflect human decision-making processes that drive land use patterns. LULC dynamics is used as the key to explaining UHI's growth patterns. Mohammad et al. (2022) estimated that due to LULC changes in Ahmedabad (India), more than 70% of the city will experience temperatures above 45 °C in the near future. Another approach to link surface characteristics with LST is using local climate zones (LCZs). The spatial distribution of LCZs in a city provides a detailed morphological setting of any built environment. Linking LST patterns to LCZ highlights how changes in building density, building heights, vegetation type, and surface bareness can alter local thermal attributes (Dutta et al. 2022; Hou et al. 2023).

Some studies also generate profiles of surface characteristics such as building indices (Varshney 2013) or vegetation indices (Moussa Kourouma et al. 2021) or water indices (Chen et al. 2020) to evaluate the cause and effect relation between land cover changes and ambient climate characteristics. These indices are also generated from multispectral satellite data and can precisely identify a particular surface attribute. A study in an Indian metropolitan city linked diurnal and nocturnal thermal hotspot values with the presence of higher impervious surface index and lower vegetation index values compared with the average condition of the region (Dutta et al. 2019). Another study carried out in Prayagraj City, India, explained the intensification of summer UHI by nearly 4 °C using dynamics of six different land indices (Sarif et al. 2022). All these parameters provide insights into the causes and possible future trends of UHI growth. Thus, these are crucial in urban climate analysis.

Many more empirical UHI studies are needed to monitor their characteristics constantly and to enrich the urban climate database. Such studies are needed more in tropical cities, which already experience very high baseline temperatures in summer. Also, the comparative assessment of seasonal UHI or day–night variations needs to be elaborated, as such characteristics are significant to understanding the local UHIs. Due to excessive heat-trapping, studies have observed the diurnal temperature range to decrease over time (Shahfahad et al. 2023). Thus, studying the day–night temperature variations becomes significant to assess the presence of thermal discomfort. Further, studies generally need more application of different satellite sensors that can enhance and validate the obtained results. A single Landsat layer-based LST profile does not reveal the dynamic characteristics of the thermal environment, which changes within seasons and months. Considering this, the current work attempted to quantitatively inspect the UHI problem in an Indian metropolitan area by intensively using both Landsat and Terra satellite data. Temporal UHI growth was assessed for the last two decades. The land cover changes that led to UHI development were also explained using spatiotemporal dynamics of LULC, surface indices, and LCZs. Multiple statistical evaluation was used to link LST variation with urbanization trends. An array of comparative assessments was produced between satellite sensors, seasons, diurnal–nocturnal changes, and decadal scenarios. The results from this study are likely to enhance climate-responsive city planning.

The region has hot and humid climatic conditions, with an average yearly temperature of approximately 27 °C and an average annual rainfall of nearly 165 cm, mainly occurring from June to September. The summer season is critical, with daytime temperatures soaring above 40 °C. Such conditions are worsened by the high relative humidity fluctuating between 72 and 90%. This local climatic condition is further degraded by ongoing urbanization and population rise. According to the latest Indian Population Census of 2011, the overall population of the KMA was 14.11 million, with 83% of the population residing in the core city districts, 11% in suburban areas, and the remaining 6% in the periphery areas. The continuing trend anticipates that the population will reach 21.1 million in 2025. Environmental degradation in this region is likely to affect multiple districts with vast populations. This calls for urgent attention to the urban climate of KMA.

This work uses Landsat satellite data covering the study area from the United States Geological Survey (USGS) Earth Explorer (collection 2, level 1) for 2003, 2014, and 2023. The dataset comprises four generations of Landsat satellites: Landsat 5, Landsat 7, Landsat 8, and Landsat 9. Multiple Landsat bands were used to derivate LST estimation, UHI demarcation and intensity computation, LULC change analysis, calculation of surface indices, and LCZ classification in the study area. Only cloud-free scenes are taken with the scanning time around 10:30 a.m. to 11:30 a.m. local time. The Landsat images used for the work are listed in Table 1.

The Terra/MODIS LST dataset (MOD11A2, version 6.1) for the years 2003, 2014, and 2023 were collected from NASA Earthdata. MOD11A2 is a mean composite of 8-day clear-sky LST conditions obtained from the MOD11A1 daily data, which can effectively overcome the data gaps caused by cloud coverage (Wei et al. 2021). The spatial resolution of MOD11A2 is 1 km in a 1,200 × 1,200 km grid. Both daytime and nighttime LST patterns were extracted for both summer and winter. This will bring out the diurnal and seasonal LST variations. The MODIS imageries used in the work are listed in Table 2.

Overall, the Landsat bands have a finer spatial resolution and are more effective in detailed spatial evaluation of LST patterns. Still, the temporal resolution of Landsat is relatively low (16 days). Comparatively, MODIS has a much better temporal frequency of twice-a-day coverage. MODIS is thus more suitable for a continuous study covering days in a season (Yang et al. 2020). The spatial resolution of MODIS scenes is coarser and can be applied to larger areas.

The average LST conditions of both summer and winter seasons were obtained using two sets of images from April and May (summer) and December and January (winter) for 2003, 2014, and 2023. This process was followed in the case of both Landsat and MODIS data. Further, Landsat-derived LST is compared/validated by MODIS-derived LST. The detailed procedure to retrieve LST information is discussed in Supplementary Appendix 1.

Before linking the LST patterns with land cover changes, the Landsat LST patterns were verified using MODIS-derived LST data. Since the MODIS LST product is provided at 1 km, the Landsat LST images are resampled to 1 km resolution (Li et al. 2021). The nearest neighbor assignment of pixel values was used while resampling (Lall & Sharma 1996). The process would alter the 30 m Landsat images to the same spatial resolution of MODIS for suitable pixel-based comparison. The images are also re-projected to the same coordinate system (World Geodetic System 1984). These images are hereafter vectorized to point files, and then, the point-to-point comparison is carried out between the LST values from two different sources. The overall spatial profile is compared with average, maximum, and minimum LST values, along with the correlation computation for the entire scene.

Landsat multispectral data was used to monitor LULC changes. Images were obtained from the same month to remove any effect of seasonal changes. Analyzing the temporal dynamics of LULC classes is crucial. Particularly, estimating the changes for cooler land cover types like vegetation and water bodies will help to strategize UHI mitigation plans. In the current work, decadal changes in LULC pattern from 2003 to 2014 and from 2014 to 2023 were mapped. Thus, the changing pixels from natural land cover classes to built-up land use can be identified, and critical areas can be demarcated. Five LULC classes were extracted from the study area using the maximum-likelihood supervised classification technique, which has desirable mathematical and optimality properties (Dewan & Yamaguchi 2009a). Spectral signatures were collected for each LULC class from the decadal stacked images. These samples were then used to train the maximum-likelihood supervised classification technique.

NDWI is calculated using green (Band 3) and near-infrared (Band 5) bands for Landsat 8/9 and green (Band 2) and near-infrared (Band 4) bands for Landsat 5 and 7. The datasets of September, October, and November have been used, as July to October is monsoon season in India, so the water spread area is expected to be at maximum for these months. For 2003, no satellite imagery of less than 50% cloud coverage was available during these monsoon months, so post-monsoon has been considered for a better comparison.

Normalized Difference Vegetation Index (NDVI) is globally applied in agricultural, ecological, or forestry studies to record changes in vegetation cover and detect any stress or damage to vegetation health. The index values are also often used to map and categorize different types of vegetation, along with seasonal or yearly change mapping (Huang et al. 2021). The computation of NDVI is shown in Supplementary Appendix 1. The datasets of April and May have been used during the last two decades as vegetation thrives during these months in India, so the vegetation cover is expected to be maximum for these months.

LCZ classifies a city region using rigorous morphological characters. The categorization may use the patterns of built-up density, building heights, open spaces, vegetation types, etc. Thus, within the urban boundary, LCZ can demarcate various types of built-up categories and add more details to intracity environment data than LULC patterns. According to this scheme, 10 classes are there for urban built-up areas (LCZ 1 to LCZ 10), and natural land cover is classified into 7 types (LCZ A to LCZ G).

In the current study, the World Urban Database Access Portal Tools (WUDAPT) instructions were then followed to classify LCZs. The Google Earth Engine was used to generate the training areas and identify the classes precisely based on their building arrangement or vegetation type. The ‘ruler’ tool was used for all the necessary measurements following the description of LCZ classes. At least 15 polygons were created for each class that can be used as the training signatures. A band composite was then generated using the Landsat dataset (2023). The training signatures from Google Earth were loaded into GIS software, and the composite Landsat image was classified. The maximum-likelihood algorithm of supervised classification was used for its higher accuracy (Dewan & Yamaguchi 2009b).

The UHI maps were derived from the Landsat LST maps using cluster-outlier analysis. The seasonal UHI pattern is shown in Supplementary Appendix Figure S3. HH clusters were identified as hotspots with similar temperatures, and LL clusters were identified as cold spots. The red patches show the HH cluster, and within that area, some data points deviate from behavior from the surrounding areas, referred to as LH outliers. Similarly, the blue color patches show the LL cluster, and within that area, some data points deviate from behavior from the surrounding areas, referred to as HL outliers. The core area of KMA shows the HH cluster, and water bodies or vegetation cover in this core area is depicted as an LH outlier.

The SUHI intensities using the urban–rural dichotomy for 2003, 2014, and 2023 were 1.71, 3.30, and 3.98 °C, respectively. The temperature values in detail are shown in Table 3. The SUHI intensity has been significantly soaring during the last two decades. Also, the results suggest that SUHI intensity is getting increasingly critical in the summer.

Vegetation loss was most noticeable along with water bodies as these classes shrank by −8.50 and −5.17%, with a yearly decrease rate of −0.425 and −0.258%, respectively, from 2003 to 2023. A remarkable change was observed in that the bulk of dense vegetation was converted to light vegetation and built-up areas. The total vegetation cover drastically decreased over time at the rate of 4.02% from 2003 to 2014 and 4.47% in the year 2014–2023. Simultaneously, the highest built-up growth occurred between 2014 and 2023 at a rate of 8.35%. The overall change from 2003 to 2023 shows that among all the classified categories of LULC in the study area, light vegetation (+52.16%) has the maximum share with respect to the total land coverage in the study area, followed by built-up areas (+39.90%). These spatiotemporal distributions are shown in Figure 6.

The OA and class-wise accuracies generated with Google Earth Pro are shown in Supplementary Appendix Tables S1–S3. For all 3 years, the percentage of OA was computed to be higher than 87%. Additionally, the kappa coefficient produced a satisfactory result as the values were higher than 0.8 (B R and S V 2018).

Different LULC indices like NDWI and NDVI were analyzed to detect the changes in the study area.

The value of NDWI lies between −1 and +1. The higher the NDWI value (value closer to +1), the more water content will be present. From the NDWI images of KMA of 2003, it has been found that values for NDWI range from −0.49 to 0.43. The percentage of water content in 2014 decreased comparatively compared with 2003. In 2014, values of NDWI in KMA ranged from −0.5 to 0.18. In 2023, the values of water indices range from −0.53 to 0.15, showing that the presence of water content has decreased during the last two decades. The variation is shown in Supplementary Appendix Figure S4.

The NDVI index was mapped (Supplementary Appendix Figure S5) to understand the spatial characteristics of 2003, 2014, and 2023 vegetation patterns. The value of NDVI lies between −1 and +1. The higher the NDVI value, the denser and healthier the vegetation is. Values greater than 0.6 are considered dense vegetation. Sparse vegetation has values between 0.2 and 0.5, whereas other LULC classes, like water bodies, built-up, and barren areas, have values of 0.1 or less. The percentage of vegetation cover in 2003 is higher in the southern KMA compared with 2014 and 2023. In the year 2003, the NDVI values for KMA ranged between −0.38 and 0.59. Over the next two decades, green cover has substantially lowered in the Central Business District (CBD) region. This region is nearly vegetation-less. In 2014, NDVI in KMA ranged from −0.15 to 0.57. The percentage of areas covered by moderate vegetation decreased notably in 2014 compared with 2003. By the year 2023, NDVI range for the city further dropped to values between −0.10 and 0.56.

The newly formed heat islands between each decade were exhaustively identified and mapped. These UHI locations were verified using past and current images of Google Earth for a clear verification of the results. The locations for the two decades are listed separately and are represented in Tables 5 and 6. The continuously developing residential zone like Jadavpur is marked with high UHI (increase by 3.42 °C in 2003–2014 and again increase by 2.17 °C in 2014–2023). On the other hand, industrial sections like Dhulagori and Dankuni areas reflected comparatively very high UHI values in 2003–2014 and 2014–2023. The very recently built residential sectors of Garulia and Gayeshpur also emerged as newly growing UHIs in 2014–2023.

Table 5

Newly developed thermal hotspots between 2003 and 2014 in KMA

View Large

Table 6

Newly developed thermal hotspots between 2014 and 2023 in KMA

View Large

The areas with high UHI intensity are termed urban hotspots (UHSs), and those with low intensity are considered urban coldspots (UCSs). UHSs were mainly located in the areas around the center of the city along the Hooghly River, the southern part of KMA due to densely built-up areas, and also a few small pockets located in the coldspots of the city. UCSs were mostly found in the periphery of the city, with some pockets also located in hotspot zones. The high UHI intensity zone covered almost the urban area of KMA, including the localities of Jadavpur, Dankuni, Ballygunge, Agarpara, Bara Bazaar, Garulia, Gayeshpur, Jagaddal, etc., and small hotspot pockets like Dhulagori, which is an industrial area lying between the cold area (due to the presence of agricultural land and wetlands). The localities of KMA with low UHI intensity included Andul, Alipore, Behala, Howrah, and others.

For this study, six areas have been identified. Three from UHS zones (Dhulagori, Bara Bazar, and Jadavpur) and three from UCS (Andul, Alipore, Behala) in order to analyze the land use characteristics and recommend planning strategies for those UHS zones and the rest of other heat stress areas to mitigate UHI by studying UCSs. They are shown in Supplementary Appendix Figure S6, along with the respective LST changes at these locations.

The error matrix for the accuracy assessment of the derived LCZ pattern is documented in Supplementary Appendix Table S4. Similar to the accuracy assessment of LULC, all three parameters were calculated for the LCZ map of KMA, i.e., OA, producer's accuracy, and user's accuracy. The outcome shows the OA value reaching 81.33% and the kappa coefficient value at 0.8, indicating good and reliable results.

The relative LST values compared with the mean scenario are also shown in Figure 9. Each bar depicts the minimum, mean (blue points), and maximum observed LST values from each class. In the case of the relative SUHI intensity, the LST difference was negative only for LCZ G (water) and LCZ A (dense vegetation) compared with LCZ D (low plant), which indicated that these LCZ classes have a lower LST than low plant. Relative SUHI intensity was positively high for the LST differences between LCZ 1–8, LCZ 10, LCZ B, LCZ C, LCZ E, and LCZ F.

The changes in the absolute values of summer daytime LST were +4.16 °C between the years 2003–2014 and +0.29 °C between the years 2014–2023. Similarly, for the winter daytime, the rise in LST was higher in the first decade, with a change of +4.36 °C from the year 2003 to 2014 and +0.44 °C between the years 2014–2023. The LST changes can be attributed to the built-up growth around Kolkata, as higher LST patterns followed the same spatial trend of urbanization (Figure 6). Simultaneously, the loss of vegetation, as noted from decadal NDVI and LULC maps, would change evapotranspiration rates, albedo pattern, and shading effect to reduce cooling. The decadal LST change reflected that environmental changes at KMA were more prominent from 2003 to 2014. In the last decade, urbanization has reached saturation in the central city regions. On the other hand, suburbs may face drastic alterations in local climate in the coming years with a sharp rise in LST intensity similar to what KMA faced between 2003 and 2014. The current scenario of LST changes demonstrates a notable increase in the area covered with higher surface temperature values. The high-temperature zone that covered 1.35 and 2.85% of the area in 2014 and 2023, respectively, is likely to cover a larger extent in the near future with the current rate of temperature rise (see Supplementary Appendix Figure S1). This remarkable growth of UHI may enhance hazardous extreme events and even adversely affect water accessibility (Shakhawat Hossain et al. 2019). Temporal UHI growth analysis showed that even in locations near the rural areas, SUHI magnitude increased by +1.43 and +1.54 °C between 2014 and 2023. This reflects the immediate effects on regional climate with transformation in LULC at any location. The seasonal and day–night analysis showed summer days to be more intense than summer nights or winter conditions (Supplementary Appendix Figures S1 and S2). A contemporary study in a similar city of Dhaka witnessed average ambient UHI intensity to be higher during winter nights (Tabassum et al. 2024). This contrast highlights the nature of congestion and prolonged heat-trapping in Dhaka compared with Kolkata.

The validation of Landsat LST patterns with MODIS added better clarification of the remote sensing procedure adopted in the study. The RMSE of LST values between two satellite data products on average for the summer season is 3.77 °C, while in winter, it was 2.41 °C, indicating a strong positive association between Landsat LST and MODIS LST values. The cross-comparison highlighted that while Landsat is suitable for mapping the minute spatial variation in LST patterns and UHI boundary, MODIS can also be used in urban studies for day-to-day fluctuations in weather. Since the LST patterns obtained from these two satellites are very similar, diurnal–nocturnal UHI studies can also be carried out using Landsat and MODIS to see the changes in heat-trapping in a city.

The dynamics of LULC showed remarkable similarities to the UHI growth patterns. It was observed that the natural covers like green spaces, water, and barren surfaces were mostly replaced by large-scale infrastructural development and haphazard urban sprawl. The reduction in the presence of water cover is depicted by a drop in the maximum NDWI values from 0.425 to 0.152. The dense green cover was reduced by a massive 652.69 km² during the last two decades, and degradation in the quality of dense vegetation was shown by a drop in maximum NDVI values from 0.597 to 0.561, reflecting the major cause behind hotspot development.

The inter-LCZ temperature analysis showed that the compact building arrangement of LCZ 1–LCZ 3 (all high-, mid-, and low-rise) resulted in much higher summer LST values compared with the open arrangement of built-up (LCZ 4–LCZ 6). Comparable findings were reported for another Indian city, Nagpur, where heat index and humidex stress were highest for LCZ 3 (Kotharkar et al. 2021). This established that LST has a strong positive correlation with building density. The densely arranged urban sections also reflect a larger population and concentration of more anthropogenic activities, leading to the rise of LST. Thus, building density regulations must be prioritized to curtail the impact of UHI. The dense trees (LCZ A) had the second lowest LST, which indicates that increasing vegetation density would help alleviate the SUHI effect effectively. Like LCZ A, water (LCZ G) had the lowest LST, indicating that protecting the water area would help alleviate the SUHI effect. SUHI magnitude was maximum for LCZ 1–5, LCZ 7–8, LCZ 10, and LCZ E classes. This was mainly because these LCZ classes had poor heat capacity. As the temperature increased on summer days, these particular classes warmed up at a much quicker rate. Contrastingly, classes like LCZ A–C and LCZ G have much higher heat capacity than the built-up zones. Hence, these natural LCZs readily prevented quick heating up. The SUHI intensity computed using LST variation between the city and neighboring rural areas could only represent a coarser spatial profile of UHI and may greatly vary from an empirical case study point of view.

LCZ classification-based relative LST analysis in the study provided an objective-based SUHI assessment approach. The stability of this procedure makes it suitable for inter-case comparison and produces a more thorough method and urban–rural difference-based SUHI evaluation. This also provides a planning scope to enhance urban climate using blue-green spaces optimally (Pritipadmaja et al. 2023; Gupta & De 2024). For the compact settings, the green cover should be increased on rooftops, building facades, and pedestrian levels. Rooftop reflectance can be increased by altering material albedo. Traffic could be diverted to reduce anthropogenic heat from the identified compact commercial zones. Canopy and shading facilities over the summer UHI can also bring down the heat intensity. Future development of LCZ 1–LCZ3 should be regulated. The demarcated green cover and water bodies must be preserved. Green belts should be planned around critical SUHI pockets. All the mapped open spaces have to be protected or, if possible, converted to green parks. The necessary location-based information regarding all these land cover or land use characteristics was generated in this study.

Several cities and districts in India have adopted heatwave action plans. Examples include the Ahmedabad Heat Action Plan released in 2013. In 2019, the Nagpur Municipal Corporation and Maharashtra State Public Health Department collaborated to implement a regional heat preparedness plan, marking the adoption of India's first collaborative approach to heat wave planning and management across Nagpur and four neighboring cities. As per the Intergovernmental Panel on Climate Change (IPCC) report, the West Bengal state should prioritize the development of a comprehensive HAP to prevent fatalities and address vulnerabilities due to extreme heat risk. The built environment's impact on urban ecosystems (Das & Das 2019) underscores its vital role in structured HAPs, aiding policy formulation and decision-making. Knowledge of urban hotspots aids spatial heat reduction frameworks and prioritizes risk mitigation measures, while heat vulnerability maps identify spatial heat risks (Fragomeni et al. 2020; Gupta et al. 2024). Remote sensing inputs derived from multiple approaches were provided in this study to meet the sustainable goals.

The severity of the UHI problem was rigorously analyzed in a hot and humid urban climate. The critical rise in local summer LST over two decades was shown by all parameters of average (4.62 °C), maximum (4.45 °C), and minimum temperature (4.77 °C) values. The LST pattern was confirmed using separate satellite sensor datasets with an average RMSE of 3 °C. Such verification produced an exhaustive thermal climate assessment method. Further, the application of the clustering technique on LST values provided a methodology of UHI demarcation that can be applied in temporal case studies and inter-case comparisons. Decadal analysis of UHI showed that not only has the areal extent increased, but also the intensity has sharply risen. In 2023, the overall UHI intensity was nearly 4 °C. Given the absolute temperature trends of Kolkata, such intensity will notably harm the local population. A seasonal analysis revealed that UHI magnitude and growth have been higher for summer months compared with winter. The maximum LST of the UHI zone soared by nearly 4.5 °C in summer over two decades, while it dropped slightly in winter. Even near the rural areas, new UHI spots showed a magnitude around 1.5 °C, while at the city center, the intensity was as high as 5 °C. Additionally, the use of both LULC and LCZ gave a small-scale to large-scale understanding of the causes behind UHI development. LULC dynamics depicted rapid vegetation loss of around 151 km² in 20 years and identified remaining patches of dense vegetation that must be protected. Since both dense trees (LCZ A) and water (LCZ G) are found to have the minimum LST values, increasing the vegetation density and protecting water bodies would help reduce the SUHI effect. Dense trees can bring down average LST from built-up areas by 3.57 °C. The cooling effects were also verified with decadal NDVI and NDWI patterns for the study region. The LCZ map for the recent time provided intricate information on KMA building arrangements. This could be used for future smart urban policymaking. The study highlighted compact mid-rise building arrangements as the most crucial zones of hotspot creation. These locations need urgent attention and landscape planning. Along city fringes, 10.31% of the area is covered by bush or scrubs. This land cover should be converted to dense trees for regional cooling purposes. Location-based solutions must be provided to enhance urban health conditions. Urban authorities should integrate local climate knowledge, monitor meteorological data, and use evidence-based mapping to plan for extreme heat events effectively.

The authors are also thankful to the anonymous reviewers for their constructive suggestions to improve the quality of the present work.

A.G. and A.S.: Writing – editing original draft preparation; B.D.: proofreading and supervision. All authors have read and agreed to the published version of the manuscript.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.