1 Introduction

Cities play an important role as service centers for their own and their hinterland population. It attracts rural people for employment, quality of life, and better infrastructural facilities. Some cities are popular for their geographic location, financial capital, religious and cultural centre. They are topophilia for people due to employment generation, availability of basic amenities, infrastructural development, and better connectivity in comparison to rural areas [32]. Hence, people migrate from rural to urban areas [45] on large scale. That is why, the rapid urbanization [14] and faster pace of infrastructural development are converting agricultural land and vegetation cover into the Urban Concrete Island (UCI). The high concentration of population and anthropogenic activities [20] generate various urban problems [13] and increase the environmental threats such as air, water, sound pollution, urban heat islands, habitat losses, shrinking of water bodies and social pollution [36]. At the global level, cities are casing 3 per cent of the earth’s land area but they are consuming 60 to 80 per cent of energy and 75 per cent of the world’s carbon emission. In Asia and Africa, the level of urbanisation will be 90 per cent within the next 30 years [44]. Similarly, India is moving towards the high rate of urbanisation. The urban area is predominating in the form of horizontal and vertical sprawl [31] along with the increasing population in Indian cities. The land use planning and norms of governing bodies/authorities fail to control [16] such problems which are interrupting urban morphology and environment [33].

According to Sustainable Development Goal (SDG) number-11 (make cities inclusive, safe, resilient, and sustainable), the cities should be sustainable and safe for their inhabitants in few decades, the cities are becoming an Urban Heat Island (UHI) due to concretisation of land [6], high emission of greenhouse gases (GHG), high consumption of energy, lack of open and green spaces etcetera [21]. The cities throughout the world are heating up twice as much as the global mean (1.5 °C) and their temperatures will rise 4 °C by 2100 [44]. These facts are insufficient for explaining each and every city region because of the different geographical conditions of each city. Indian cities are crossing the hierarchy level and converting towns/cities into megacities [12, 38] or metropolis swiftly. Consequently, UHI and urban heatwaves are a serious problem for Indian cities and almost all million cities are facing this problem. There is the differentiation between different cities about the formation of UHI [48] because every city has its own specific geographical location [18], issues and morphology [43].

The examination Land Use/Land Cover (LULC) classification are essential for knowing the proportion of different types of land use. The examination of geospatial indices like Normalised Difference Vegetation Index (NDVI), Normalised Difference Built-up Index (NDBI), Normalised Difference Water Index (NDWI), Normalised Difference Bareness Index (NDBaI), Normalised Difference Latent-heat Index (NDLI) and other land features is the most important practice to understand the environmental condition [3]. Land Surface Temperature (LST) is a satellite-based calculation and representation of surface temperature, one of the easiest approach to study the thermal environment [23]. There are various types of land cover that has strong control over the LST ranging from green space, crop lands [5], rivers and palaeochannels [41] etcetera.

In the context of Indian cities, various works have been done regarding the dynamics, pattern and relationship among LULC, LST and geospatial indices. For instance, various scholars conducted studies on the evaluation of urban environment using LST as one of the essential environmental indicators. Pal and Ziaul (2017) studied in English Bazar town to show that LST was highly dependent on LULC in urban areas, and sustainable land use planning can control the heat environment [12, 34]. Jain et al. [21] studied UHI and geospatial indices (NDVI, NDBI, and NDBaI) in Nagpur city and suggested street-level planning to control the UHI problem. The high LST or UHI problem primarily affects summer time [9], and it varies across seasons and months in Guwahati city. A study in Pune city focused on the role of local climate and Diurnal Temperature Range (DTR) in impacting the thermal environment and found that the high proportion of built-up area is the primary cause of UHI creation in Pune city [38]. Similarly, a study on Mumbai and Delhi metro cities describes the increase of LST and explains the worst condition of thermal environment and increasing the urban hotspots and urban thermal field variance index zones in both cities [42]. Other significant studies on Indian cities, such as Faridabad, Barrackpore, and Panchkula [27] reveal a trend of increasing LST [7, 32] and decreasing urban environmental health [37]. Several studies have been conducted in different Indian cities, but they were limited to the municipal boundaries [27]. However, for sustainable planning, it is necessary to take into account the development region of the city. Hence, rapid ranges in LULC has been identified in the study area from 1991 when the Varanasi city has became a million city. Such LULC changes are responsible for the various environmental problems and increasing LST is one of them [40]. That is why, VCDR was selected as study area. The objectives of this study are: 1) To assess the pattern and dynamics of LULC, geospatial indices (NDVI, NDWI, NDBI, NDBaI and NDLI) and LST in the context of space and time. 2) To investigate the relationship among LULC, selected indices and LST for the years 1991, 2001, 2011 and 2021.

2 Dataset, tools, and methodology

2.1 Study area

Varanasi City (Kashi/Benares/Banaras) is one of the oldest settlements in the world. Varanasi city (Capital of Knowledge) lies between rivers Varuna and Assi [22]. The Varanasi City Development Region (VCDR) is located between 25º 08′ 01" N to 25º 28′ 35" N and 83º 10′ 17" E to 82º 51′ 04"E. The total geographical area is 67,334.22 hectares (Master Plan-2031). The VCDR lies in the mid-Gangetic plain and on both banks of the river Ganga. The average elevation is 77 m but it varies in different parts like 83 m on the northern part, 70 m in CBD, 72 m in the southern part and 69 m along the river. The average annual rainfall and temperature are 1067 mm and 33.4ºC respectively. April is the driest and August is the wettest month of the region [30]. The total population of VCDR was around 2.5 million in 2011 and 3.36 million in 2021 which was estimated by the Varanasi Development Authority (VDA). The sex ratio was 887 females/1000 males and the literacy rate was around 80 per cent. Shri Kashi Vishwanath temple, Banaras Hindu University (BHU), Banaras Locomotive Works (BLW), Sarnath Archaeological sites, Cantonment Railway Station and Lal Bahadur Shastri International Airport are the major service centers of the city region. The demarcation and delimitation of VCDR have been prepared jointly by VDA, District Urban Development Authority (DUDA) and Varanasi Municipal Corporation (VMC) under the mentorship of the Town and Country Planning Organization (TCPO) in Master Plan-2031. The planning of the city region is the most ideal approach for futuristic and sustainable development and most of the scholars are agreed for making the master plan for the upcoming 20–50 years. The land-use and environmental planning should be done in the inner part of the city as well as its periphery region. That is the reason for choosing VCDR as the study area (Fig. 1).

Fig. 1
figure 1

Study Area

Full size image

2.2 Dataset and tools

The base map of VCDR has been digitised and prepared through the master plan-2031 of VDA using QGIS application. The Varanasi city’s map with ward boundary has been taken from the District Census Handbook (DCHB-2011) report. The Landsat-5(TM) and Landsat-8(OLI/TIRS) imageries of the United States Geological Survey (USGS) have been obtained and used for this study. The collected satellite data (level-1 and also corrected) was preprocessed on geometric and atmospheric standards. Furthermore, data was converted into Universal Transverse Mercator (UTM) zone 44N projection with WGS-84 datum. For the geometric correction of data, Google Earth Engine Pro was utilised and 50 Ground Control Points (GCPs) were taken by mobile GPS for ground truthing. The obtained Landsat-5 and 8 data of different time series have been mosaic using ERDAS Imagine (version-14) application and Arc GIS 10.8 software was used for the final layouts of maps. The Origin Pro application (version 9.6.5) was used for making graphs and diagrams. The statistical analysis has been done in Statistical Package for Social Sciences (SPSS-2013) application. Table 1 provides a broad description of satellite data i.e. satellites and their sensors, acquisition date, resolutions and path/row etc.

Table 1 Description of Satellite Data
Full size table

2.3 Classification of land use/land cover

After the band composition, area of interest has been extracted to enhance the clarity of land features. The training signatures was prepared creating polygons for different LULC classes (Table 2). Furthermore, supervised classification with maximum likelihood method [28] has been applied for the image classification through prepared signature files. Finally, pixel correction was done using Erdas Imagine application.

Table 2 Land Use/Land Cover classes and their features
Full size table

2.4 Accuracy assessment

Accuracy assessment is the most significant process for the reliability of LULC classification. It is an estimation of the precision of LULC maps [34]. In this study, 400 randomly selected reference points have been selected for preparation of error matrix. With the help of error matrices, 3 types of accuracy estimation were computed which are the Overall, User and Producer accuracy. Representatively,

$$\alpha \, = \,\frac{n}{N}\, \times \,100$$
(1)

where,\({\prime}\alpha\)’ is the classification accuracy, ‘n’ denotes the number of corrected points classified on the image, and ‘N’ is the number of points verified in the field.

$$\beta \, = \,\frac{\sum Dij}{{Rj}}$$
(2)

where, ‘β’ is the producer accuracy, ‘Dij’ represents the corrected classified pixels in row ‘i’ (diagonal cell), and ‘Rj’ is the total number of pixels in row ‘j’ [2, 4]

$$\gamma \, = \,\frac{\sum Dij}{{Cj}}$$
(3)

where, ‘γ’ denotes user accuracy, ‘Dij’ is the corrected & classified pixels in column j (diagonal cell), and ‘Cj’ is the total pixels in column ‘j’. Having significantly greater interclass discrimination than the overall accuracy. Kappa statistic is the most broadly used measurement of LULU classification accuracy [24, 47].

$$k\, = \,\frac{{N\mathop \sum \nolimits_{i - 1}^{r} X_{ii} - \mathop \sum \nolimits_{i = 1}^{r} \left( {X_{i + } \times X_{ + i} } \right)}}{{N^{2} \, - \,\mathop \sum \nolimits_{i = 1}^{r} \,\left( {X_{i + } \, \times X_{ + i} } \right)}}$$
(4)

where, ‘k’ is representing the kappa coefficient, ‘N’ is the total number of observations, ‘r’ denotes the total number of rows present in the error matrix, Xii is the number of observations present in row and column ‘i’ respectively, Xi+ is the total number observations in row ‘i’ and ‘X+i is the total number of observations in column ‘i’[8].

2.5 Computation of geospatial indices

NDVI is a satellite-derived vegetation index which used comprehensively in the assessment of green environment and also very useful to understand the urban environment as an indicator of green cover [19]. It is measured using red and near-infrared bands in multispectral images (Table 3) with the help of the following formula [26]:

$$NDVI = \left( {NIR - RED} \right)/\left( {NIR + RED} \right)$$
(5)
Table 3 Land Use/Land Cover Classes and their Proportion Area
Full size table

NDBI is helpful for mapping and measurement of urban concretised areas or impervious surfaces. The positive NDBI value indicates the high built-up area, the value around 0 indicates vegetation cover and the negative value represents water bodies [46]. In this study, NDBI has been measured by the following formula:

$$NDBI\, = \,\left( {MIR\, - \,NIR} \right)/\left( {MIR\, + \,NIR} \right)$$
(6)

NDWI is a spectral index that calculate the concentration of water in vegetation canopies, soil and other surface materials. NDWI values range from -1 to 1, with positive values indicating a higher concentration of water and negative values indicating a lower concentration of water [29]. In remote sensing applications, NDWI is often used to detect and monitor changes in surface water, such as in wetlands, rivers, lakes, and reservoirs. The formula for NDWI is as follows:

$$NDWI\, = \,\left( {GRE\, - \,NIR} \right)/\left( {GRE\, + \,NIR} \right)$$
(7)

NDLI is an index that is commonly used to estimate the water stress level of crops. It is calculated using the thermal infrared and near-infrared bands. The NDLI value ranges from -1 to + 1. A positive NDLI value indicates that the crop is in a relatively stress-free state, while a negative value indicates that the crop is under stress due to lack of water or other environmental factors. The formula for calculating NDLI (formula 8) is:

$$NDLI\, = \,\left( {GRE\, - \,RED} \right)/\,\left( {GRE\, + \,RED\, + \,SWIR} \right)$$
(8)

NDBaI is useful to identify and map the areas of bare soil or exposed ground. It is calculated using two bands of remote sensing data, which are sensitive to the amount of vegetation cover in a given area. In general, areas with low vegetation cover and high amounts of bare soil will have a high NDBaI value, while areas with high vegetation cover and little bare soil will have a low NDBaI value [49]. NDBaI values can range from -1 to 1, with values close to -1 indicating high levels of bare soil and values close to 1 indicating high levels of vegetation cover.

$$NDBaI\, = \,\left( {SWIR\, - \,TIR} \right)/\left( {SWIR\, + \,TIR} \right)$$
(9)

2.6 Computation of land surface temperature

LST is an important aspect of the monitoring the urban thermal environment. It is calculated through obtained satellite data such as Landsat-5 TM (band-6) and Landsat-8 TIRS (band-10 & 11) using their thermal bands. The month of May has been chosen on the basis of problem and avoiding the cloud cover problem. The conversion of digital number (Dn) values into spectral radiance for Band-6 of TM, the following formula (Eq. 10) was used [25]:

$$L\lambda \, = \,\frac{{L_{max} \lambda - L_{min} \lambda }}{{QCAL_{max} \, - \,QCAL_{min} }}\, \times \,\left( {QCAL\, - \,QCAL_{min} } \right)\, + \,L_{min} \lambda$$
(10)

where, \(^{\prime}L\lambda ^{\prime}\) presents the radiance value of the atmospherically corrected cell, \({^{\prime}L}_{max}\lambda ^{\prime}\) corresponds the spectral radiance scaled to \(^{\prime}Q{CAL}_{max}^{\prime}\), \({^{\prime}L}_{min}\lambda ^{\prime}\) shows the spectral radiance scaled to \(^{\prime}Q{CAL}_{min}^{\prime}\), \(^{\prime}Q{CAL}_{max}^{\prime}\) and \(^{\prime}Q{CAL}_{min}^{\prime}\) present the maximum (255) and minimum(1) values respectively and \(^{\prime}QCAL^{\prime}\) express the digital image value. The spectral radiance of TIRS bands (Band-10 & 11) of Landsat-8 has been converted into digital number. The formula for TIRS image (Eq. 11):

$$L\lambda \, = \,ML\, \times \,QCAL\, + \,AL$$
(11)

where, \(^{\prime}L\lambda ^{\prime}\) denotes spectral radiance on the top of the atmosphere, \(^{\prime}ML^{\prime}\) and \(^{\prime}AL^{\prime}\) uses for radiance multi-band ‘X’ (Band number) and radiance add band ‘X’ respectively, \(^{\prime}QCAL^{\prime}\) represents of the quantified and calibrated standard product pixel value. \(^{\prime}ML^{\prime}\) and \(^{\prime}AL^{\prime}\) both are rescaling factors which were collected from Metadata (.mtl) files. For obtaining ‘at-satellite brightness temperature’ (Eq. 12) of spectral radiance (Kelvin ‘K’), emissivity modification has been added with radiant temperature on the basis of land cover nature [17].

$$T_{b} \, = \,\frac{{K_{2} }}{{L_{n} \left( {\frac{{k_{1} }}{L\lambda }\, + \,1} \right)}}\, - \,273.15$$
(12)

where, \(^{\prime}{T}_{b}^{\prime}\) is representing ‘at-satellite brightness temperature in K’; \(^{\prime}L\lambda ^{\prime}\) is the radiance of at-satellite in W/(m2sr.μm); K1 and K2 denotes thermal calibration constant which are, for Landsat-5 TM (band-6) K1 = 607.76, K2 = 1260.56, and for Landsat-8 TIRS band-10 & 11, K1 = 774.8853 and K1 = 480.8883 respectively; for band-10 & 11, K2 = 1321.0789 and K2 = 1201.1442 in that order. The modification of Spectral emissivity (E) is essential because the temperature has been taken with the comparison of black entities (Eq. 14). The proportion of vegetation (Pv) is a rational method to calculate the emissivity values for every pixel (Eq. 13).

$$P_{v} \, = \,\left\{ {\frac{{\left( {NDVI_{max} \, - \,NDVI_{min} } \right)}}{{\left( {NDVI_{max} \, - \,NDVI_{min} } \right)}}} \right\}^{2}$$
(13)
$$E\, = \,0.004\, \times \,P_{v} \, + \,0.986$$
(14)

After that, the computation of LST (Eq. 15) using the following formula:

$$LST\, = \,T_{b} \, + \,W\, \times \,\frac{{T_{b} }}{p}\, \times \,L_{n} \,\left( E \right)$$
(15)

where, ‘\({T}_{b}\)’ is brightness temperature (eq. xii), ‘W’ is the wavelength, ‘p’ is \(h\times c/\sigma\) (1.4388 × 10–2 mK = 14,388 μmK), ‘h’ is Planck’s constant (6.626 × 1034j/sec), ‘σ’ is denoting the Boltzmann’s constant which is 1.38 × 1023 Joule/K, ‘c’ denotes candela (light velocity 2.99 × 108 m/sec) [10].

2.7 Correlation analysis

Correlation analysis is useful to know the degree of relationship between two or more variables. In this study, Pearson’s product-moment correlation coefficient (PPMCC) has been used for analysing the correlation between LULC, spectral indices, and LST [11]. The PPMCC is calculated using following formula (Eq. 16):

$$r_{xy} \, = \,\frac{{n\sum x_{i} y_{i} \, - \,\left( {\sum x_{i} } \right)\left( {\sum y_{i} } \right)}}{{\sqrt {\left[ {n\sum x_{i}^{2} \, - \,\left( {\sum x_{i} } \right)^{2} } \right]} \left[ {n\sum y_{i}^{2} \, - \,\left( {\sum y_{i} } \right)^{2} } \right]}}$$
(16)

where, \({^{\prime}r}_{xy}^{\prime}\) is the PPMCC, \(^{\prime}\sum {x}_{i}^{\prime}\) is the summation of x value, \({^{\prime}\sum y}_{i}^{\prime}\) denotes the summation of y values and \(^{\prime}n^{\prime}\) is the number of observations.

3 Result

3.1 Pattern and dynamics of land use/land cover

Table 3 presents the spatial and temporal distribution of the LULC classes for the VCDR which has been classified through the supervised classification for year 1991, 2001, 2011 and 2021. The images clearly revealed that there was a significant land alteration in the central part (municipal area) of the study area (Fig. 2).

Fig. 2
figure 2

Land Use/Land Cover (1991–2021)

Full size image

Fragmented vegetation cover and the high proportion of built-up area also explicit the effect of rapid urbanization in the Varanasi city. Between 1991 to 2001 water bodies (1.56 in 1991, 1.49 in 2001), agricultural land (72.94 in 1991, 71.36 in 2001), barren land (7.41 in 1991, 6.24 in 2001), fallow land (5.19 in 1991, 4.30 in 2001), river sand (1.91 in 1991, 1.69 in 2001), vegetation cover (6.53 in 1991, 4.90 in 2001) exhibited declining trends and only built-up area (4.45 in 1991, 10.01 in 2001) has been increased (Fig. 3). Similarly, during 2011–2021, agricultural land (69.14 in 2011, 59.31 in 2021), fallow land (3.91 in 2011, 3.83 in 2021), river sand (1.79 in 2011, 1.78 in 2021) and barren land (2.83 in 2011, 1.45 in 2021) showed the decline trends, and vegetation cover (4.51 in 2011, 5.08 in 2021) and built-up land (16.44 in 2011, 27.05 in 2021) had inclining trend (the areas in the above paragraph are expressed in per cent). The decrease of vegetation cover in VCDR over the period (1991–2001) reached to 10.93 km2 and with a 2.61 km2 loss in 2011 while in 2021, the vegetation cover made a positive aptitude of increase by 3.8 km2 area (Table 4). In the context of spatial distribution, the concentration of built-up area was noted at the left bank of river ganga which come under the municipal boundary. The most of the transformed LULC class has been observed within the city boundary and eastern part of the VCDR (near the Mughal Sarai Town). The fallow and barren land of peri-urban area have been altered into either built-up or agricultural land during the 1991–2001 time period. While, vegetation cover of city has been damaged and it was converted into impervious surface. During 2001–2011, built-up area was majorly grown towards the north direction and following the major highways. While, barren land almost changed into agricultural land and built-up areas. The humongous LULC change was noted during 2011–2021 period (Fig. 4). The built-up area increased towards all direction and it was crossed the municipal boundary. An axial growth of urban area has been observed as well as high density of impervious surface.

Table 4 Temporal Changes in Land Use/Land Cover
Full size table
Fig. 3
figure 3

Spatial Distribution of Normalised Different Vegetation Index

Full size image

3.2 Accuracy assessment

The accuracy assessment has been done with randomly selected 400 reference points and prima facie, all reference points have been verified with the help of already prepared LULC maps by National Remote Sensing Centre (NRSC), India and Google Earth Imageries. A GPS survey was organised for collecting the ground data and again verified the prepared LULC maps. During the ground truthing process, an oral information was collected regarding previous time period LULC patterns. The user and producer accuracy were estimated with the help of error metrics for individual time period. The user accuracy was quite acceptable for the built-up (93.33, 100, 100 and 91.34 per cent in 1991, 2001, 2011 and 2021 respectively), water bodies (86.67, 86.67, 80 and 92.33 per cent in 1991, 2001, 2011 and 2021 respectively), river sand (above 86.67 over the period), agricultural land (above 85.67 per cent over the period), barren land and fallow land over the period. The result of producer accuracy was above 79.30 per cent over the period for all classes which was quite good. The result of overall accuracy was 95.23 (1991), 94.28 (2001), 91.42 (2011) and 88.34 per cent (2021) which were found reliable on the standard of USGS (USA). The result of Kappa coefficient were 0.94 (1991), 0.93 (2001), 0.89 (2011) and 0.87 (2021) which mean the excellent agreement has been computed between user and producer accuracy variables (Table 5).

Table 5 Summary Statistics of Accuracy Assessment
Full size table

3.3 Pattern of geospatial indices

Table 6 presents the summary of NDBI in the VCDR which ranged from -1 to 0.40 (mean 0.23) in 1991 and -0.43 to 1 (mean 0.06) in 2001. In 2011, it was between -1 to 0.40 (mean 0.22) and -0.44 to 0.29 (mean -0.02) in year 2021. The high NDBI values were found in the central parts of the region and along the axis (highways and railway lines) of the city (Fig. 4). However, low range of NDBI values were seen in a scattered pattern in the outskirts areas but positively related to the distance from core of the city. The mean NDVI values were 0.04, 0.16, 0.05 and 0.19 in 1991, 2001, 2011 and 2021 respectively (Table 7). While, maximum NDVI values have been found 0.4615, 0.5882, 0.5209 and 0.5478 in 1991, 2001, 2011 and 2021 respectively. The concentration of high NDVI values was identified in the outer and eastern parts of the region and the proportion of vegetation was very low in the urban areas (Fig. 3).

Table 6 Statistics of NDBI
Full size table
Table 7 Statistics of NDVI & NDWI
Full size table
Fig. 4
figure 4

Spatial Distribution of Normalised Different Built-up Index

Full size image

Table 7 presents the summary of NDWI which is highly intense along the central parts, on the River Ganges and in the north-east parts of the region. The mean NDWI values were -0.11, -0.16, -0.15 and -0.18 (in 1991, 2001, 2011 and 2021 respectively) which indicates the degradation of water content in vegetation cover and as well as low water content richness (Fig. 5). The maximum NDWI was 0.2605, 1, 0.3125, and 0.2226 in 1991, 2001, 2011 and 2021 respectively. Which is not an appropriate indicator of water content richness in any area because it represents the highest value of a pixel/point only. Overall, declining trend has been observed in NDWI over the period in the VCDR.

Fig. 5
figure 5

Spatial Distribution of Normalised Different Water Index

Full size image

The result of NDLI presents the ups and downs trends. For example, the mean NDLI was 0.0097 in 1991 and it has found 0.0088 in 2001 which shows the declining trend. Whereas, it has hugely decreased and noted -0.1011 in 2011. Again, NDLI was 0.0032 in 2021 which presents the incline trend (Fig. 6). At the end game, the mean NDBaI was noted -0.0895, -0.3522, -0.0555 and -0.3156 in 1991, 2001, 2011, and 2021 respectively (Table 8). This spatial and temporal trend of result presents the gradual decrease of bareness in soil in the study area (Fig. 7).

Fig. 6
figure 6

Spatial Distribution of Normalised Different Latent-heat Index

Full size image
Table 8 Summary Statistics of NDLI and NDBaI
Full size table
Fig. 7
figure 7

Spatial Distribution of Normalised Different Bareness Index

Full size image

3.4 Pattern and trends of land surface temperature

Table 9 corresponds the summary statistics of LST and Table 10 represents the change in LST between 1991 and 2021. In 1991, the maximum LST was noted 36.84 °C which has increased 2.54 °C in 2001. The slight increase was noted in 2011 year and it has reached 40.32 °C. In 2021, 2.7 °C surface temperature increased. Overall, the maximum LST increased by 6.18 °C during 1991–2021. The minimum temperature was 24.98 °C in 1991 for may month (peak summertime) which was increased by 2.28 °C in 2021. The mean LST was the most important indicator of the thermal environment of any region. In VCDR, the mean LST was found 32.56, 32.76, 33.62 and 34.67 °C in 1991, 2001, 2011 and 2021 respectively. Overall, 2.11 °C LST was inclined over the study area which presents the high health risk in future (Figs. 8, 9, 10 & 11). LST of the built-up area was 35 °C in 1991 and it was increased by 6.7 °C to reach 41.7 °C in 2021 (Figs. 12 & 13). Surface temperature variations of water bodies was just 1.1 °C between 1991 and 2021. Between 1991 and 2021, the LST of the vegetation class increased by 0.8 °C, indicating that the scattering of sunlight is high from the vegetation cover only. Overall, it was revealed that LST trends were rising across each class, generally concrete regions depicted greater growth in LST than other LULC classifications.

Table 9 Summary of the distribution and changes in land surface
Full size table
Table 10 LULC Classes and their Mean Land Surface Temperature (°C)
Full size table
Fig. 8
figure 8

Land Surface Temperature (1991)

Full size image
Fig. 9
figure 9

Land Surface Temperature (2001)

Full size image
Fig. 10
figure 10

Land Surface Temperature (2011)

Full size image
Fig. 11
figure 11

Land Surface Temperature (2021)

Full size image
Fig. 12
figure 12

Pattern of Land Surface Temperature

Full size image
Fig. 13
figure 13

Different Land Classes and their Mean LST

Full size image

3.5 Relationship between land surface temperature and geospatial indices

Tables 11, 12, 13, 14 present the results of correlation among LST and geospatial indices. Over the whole-time period, there was a negative relationship between LST and NDVI (r-values were −0.178, −0.319, −0.263 and −0.159 in 1991, 2001, 2011 and 2021 respectively) was noted because high vegetation cover control the surface temperature. Meanwhile, the relationship among LST, NDWI, and NDLI have been observed negative (‘r’ coefficients value) as water bodies maintain the surrounding temperature and latent heat directly related to evapotranspiration process. While, there was a positive relationship between LST and NDBI (0.701, 0.554, 0.586, and 0.284) due to impervious surface and dense concretisation of land. Similarly, a positive relationship has been observed between LST and NDBaI (0.456, 0.290, 0.274, 0.285) over the period (Figs. 14, 15, 16 & 17).

Table 11 Pearson's correlations between LST and geospatial indices (1991)
Full size table
Table 12 Pearson's correlations between LST and geospatial indices (2001)
Full size table
Table 13 Pearson's Correlations between LST and Geospatial Indices (2011)
Full size table
Table 14 Pearson's correlations between LST and geospatial indices (1991)
Full size table
Fig. 14
figure 14

Correlation between LST and Geospatial Indices (1991)

Full size image
Fig. 15
figure 15

Correlation between LST and Geospatial Indices (2001)

Full size image
Fig. 16
figure 16

Correlation between LST and Geospatial Indices (2011)

Full size image
Fig. 17
figure 17

Correlation between LST and Geospatial Indices (2021)

Full size image

4 Discussion

This study reveals the dynamics and pattern of LULC and distribution of LST during 1991–2021 and also analysed the relationship between LULC geospatial indices and LST. In the current scenario, study of LST and LULC changes is a crucial topic when geographers, environmentalists and planners are thinking about urban climate change on micro-level. The study area is not limited to city boundary because of the horizontal sprawl and for a better understanding of LST and LULC change, the boundary of VDA has been considered. That is the reason this study also focuses on the peri-urban areas of Varanasi City, which are better to understand than urban areas, including Ramnagar Town, Pt. Deendayal Upadhyay Nagar and other nearby areas. The results demonstrate that between 1991 and 2021, the built-up area in the VCDR increased significantly, because of a rapid urban growth caused by emigration from surrounding districts of Varanasi and Bihar, Jharkhand and Madhya Pradesh states. This emigration phenomena are happening due to increase in road network and connectivity, holy nature of the city, increasing tourism activities, and enhance basic amenities like health, education and markets. This has led in the increase of service sector with more employment opportunity.

Varanasi city expanded rapidly in axial form and it is expanding along the highways and railways lines. In the last 30 years the 91.79 km2 of old and fertile alluvial soil was lost. Similarly, the barren land decreased by 40.14 km2 because of urban expansion and few patches has been transformed into agricultural land over the period. Vegetation cover have been converted into built-up land, replaced by roads, residential houses, commercial and administrative buildings etc. Overall, 22.16 per cent of vegetation cover decreased over the period because, several commercial buildings, residential houses and major roads were constructed during this period. There was a change in the vegetation cover of VCDR, which were 43.94 km2, 33.01 km2 and 30.40 km2 in 1991, 2001 and 2011 respectively but in 2021, 3.8 km2 increased between 2011 and 2021 (34.20 km2). This was largely because of social forestry, people’s awareness and government initiatives. Plantation in the areas along the River Ganges and Varuna start escalating. In the last decade, VMC, VDA and Forest Department of Varanasi district have worked on the plantation drive along the Varuna river and also on the right bank of Ganga river. This area was declared a sensitive zone for Turtle sanctuary (Kachchua Sanctuary) and Ganga Dolphine and that is the reason, the development activities have been prohibited. The plantation activities under the Ganga Vriksharopan Abhiyan (Ganga Plantation Campaign), Namami Gange Project and Varuna River Corridor Project have been promoted. The 13.24 km2 area of river sand could be transformed into green belt, which will help to promote healthy urban environment and reduce urban heat problem of the city region. Other LULC classes like fallow land and water bodies depicted low change but the fallow land which was 25.80 km2 in 2021 could be utilized as green spaces and gardens in city areas, and as agricultural land in suburban areas.

Similar to the results of research on some major cities i.e. Prayagraj [39], Agra [35], Pune [15], eight sub-tropical semi-arid cities of Western India (Nagpur, Pune, Ahmedabad, Rajkot, Vadodara, Kota, Jaipur and Udaipur) and Delhi-Mumbai metropolitans [42], the mean temperature was increased by 2.24 °C from 1991 to 2021 (more than the global average 1.5 °C) due to changing the land use and morphology of the city region, rapid concretization and addition of large number of vehicles. The maximum temperature was increased by 6.18 °C which was high during the daytime because heat waves generate on the concreted areas due to higher heat absorption and retention properties of concrete materials. Whereas, the minimum surface temperature was found in vegetation cover, crop lands and water bodies due to high evapotranspiration and heat absorption. In the context of favourable localities, the minimum temperature was noted in the Banaras Hindu University, Banaras Locomotive Works, Cantonment and left bank areas of the Ganges river due to high concentration of vegetation and presence of water bodies whereas maximum LST was observed in the sand along the river due to high reflectance and built-up areas (Mughal Sarai town, Ramnagar town and Varanasi city Municipality areas). The noticeable increase of LST has been found at 6.7 °C in built-up area due to the high concentration of concreted surface, on the other hand, the lowest increase has been noticed on vegetation cover (0.8 °C) and water bodies (1.1 °C) over the period in the city region. Which meant, the LULC change is responsible for increasing of LST and heat waves in the city region. Likewise, the results of various studies on the different Indian cities explain the similar causes and consequences for increasing LST in the urban areas [1].

The relationship among NDVI, NDWI, NDLI and LST was reported as a negative r-value which expresses the high cover of vegetation and water entity which are necessary for maintaining the surface temperature in the city region. Contrary relationship was found amongst NDBI, NDBaI and LST because of the high influence of anthropogenic activities like the heavy traffic, congested buildings, large number of air conditioners, metal roads and high energy consumptions. The spatial distribution of NDBI was found low with negative r-values in suburban areas due to low concretisation and high proportion of agricultural land which reflects the necessity of sustainable infrastructural futuristic planning for sub-urban areas. The NDBaI was used for evaluating bareness in soil as well as bifurcation of other LULC classes. The mean NDBaI throughout the period was negative which means most of the area was cultivated land. The seasonal variation and atmospheric phenomena are responsible for the fluctuation in NDBaI over the region but it triggers for heat waves. The relationship between NDBaI and LST has moderate positive correlation which means the high bareness of soil slightly influences LST. This study may be useful to understand the thermal environment of the study area on micro scale and it can be used for making sustainable plan to maintain the healthy urban life and environment within the city boundary and in peri-urban areas.

5 Suggestions

Based on the findings of this study, five points action plan SPURS (Save, Plant, Undo, Reduce and Social) has been proposed. It lists the five procedures that should be implemented to control all aspects of the thermal environment. It is urged that trees, water bodies, agricultural land, and open spaces be maintained in the VCDR while their sustainable proportions can regulate the LST in urban and peri-urban areas. Likewise, planting trees on unused land and in river sand contributes to reducing the city's average temperature, and installing solar panels on roofs is one of the more subtle steps taken to address several environmental issues. The lost water bodies should be replenished, and the creation of satellite towns is likely to reduce the city's high pressure. These strategies can be used to correct earlier planning and policy mistakes. To decrease air pollution and surface-temperature generation, it is urged that heavy vehicles be reduced and the consumption of energy be rationalized. However, public transportation and bicycles are also encouraged. Last but not least, social awareness about green and blue environment is one of the most efficient approaches for promoting the local environment. Hard laws and restrictions should be encouraged, as well as the conservation of natural features of the land. According to Fig. 18, it has been suggested that a green belt be established around the major roads, railway tracks, and river banks. The blue patches (excluding the river) were demarcated for making artificial water bodies in the study region. Hence, the thermal and overall environment of VCDR can be maintained in a sustainable manner through the SPURS plan and proposed insights.

Fig. 18
figure 18

SPURS Plan for Controlling Land Surface Temperature

Full size image

6 Conclusion

The most significant changes were detected in built-up land which has increased four times. It means the total 131.27 km2 has been converted into the built-up area from barren, agricultural land and vegetation cover due to faster pace of urbanisation. These facts are significant to understand the futuristic land-use planning and quality of environmental conditions in VCDR. In the last 30 years, the mean LST has been amplified to 2.24 °C and the maximum temperature has been 6.18 °C due to rapid concretization and anthropogenic activity in the study area. The spatial distribution of high LST was found in built-up areas which are denoting that the concrete areas are converting into urban heat islands. While, the water bodies and copious vegetation can reduce the LST of any city region. However, the vegetation cover has increased between 2011 and 2021, which is a positive indication but more concerns are needful in this development. It is essential that the policymakers and planners continue to ensure the maintenance of green spaces and water bodies within the city and peri-urban areas. Administrative authorities should make plan for problem specific and suggested SPURS plan can be helpful to better urban thermal environment in the VCDR. This work focuses on the changing land-use pattern and its impact on environmental issues like urban heat problem which will help to promote the academicians, planners and policymakers to work on other city regions of the state, country and the world to ensure sustainable urban planning and environmental management.