1 Introduction

At present, over half of the global population resides in urban areas, and this percentage is anticipated to rise further in developing nations, while urbanization is occurring at a remarkable pace on a global scale (Chen et al., 2017). The transformation of natural landscapes and climate by humans is prominently reflected in the process of urbanization. The urban heat island (UHI) phenomenon serves as a prime illustration of climate alterations induced by human activities through urbanization, leading to elevated temperatures in urban areas compared to their rural environs (Grimm et al., 2008). The primary factors contributing to the UHI effect are the heightened proportions of impermeable surfaces in cities, encompassing structures and infrastructure (Deilami et al., 2018; Morabito et al., 2021; Vujovic et al., 2021). This phenomenon has impacts on energy consumption, water usage, air quality, and human health (Fan et al., 2017). Also, urbanization disrupts climatic systems, threatens biodiversity, and influences ecosystem productivity by altering energy balances and habitats, as well as diminishing carbon storage and biomass (Cui et al., 2012). The rising urbanization rate in the upcoming decades poses significant stress, as projections indicate that over 66% of the global population will likely be living in cities by 2050 (Schwarz et al., 2012; Yuan & Bauer, 2007). By 2030, the overall urban land area worldwide is anticipated to expand by over 1.5 million square kilometers (Seto et al., 2011). Climate predictions also suggest a heightened occurrence and intensity of extreme events like heavy rainfall, storms, and heat waves, posing a potential threat to the vulnerability of urban areas (IPCC, 2007).

Urban development, driven by economic and urban growth, relies significantly on the expansion of industry and transportation. Nonetheless, this progress alters the urban microclimate by increasing the Earth’s average Land Surface Temperature (LST) (Song et al., 2020). There is a positive correlation between LST and urban areas, and a negative correlation between LST and green areas, indicating that urban regions generally experience higher temperatures compared to rural areas. However, it’s worth noting that urban green spaces tend to have slightly lower temperatures (Hua et al., 2020; Yang et al., 2020). The significant temperature rises primarily result from the UHI phenomenon, a modification that is intensified by human activities (Santamouris, 2020). Research conducted on the industrial development authority of New Okhla of India during the period from 2011 to 2019 reveals a rise of 6.4 °C in LST and a 33.56% increase in Urban Hotspots (UHS) (Sharma et al., 2021); a study on 397 cities in China revealed a substantial yearly rise in both daytime and nighttime Surface Urban Heat Island (SUHI) effects from 2001 to 2017, while research conducted on four cities in China spanning the years 1961 to 2014 identified a decline in environmental humidity associated with the growth of SUHI (Li et al., 2011; Yao et al., 2019). In exploring such phenomena, remote sensing emerges as a standout methodology among various options, enabling comprehensive urban investigations into LST, UHI, Urban Thermal Field Variance Index (UTFVI), and Land Use and Land Cover (LULC) through the utilization of satellite images equipped with infrared sensors (Song et al., 2018). Analyzing LULC is crucial for comprehending the impacts that changes in urban area coverage can exert on LST and the SUHI (Tepanosyan et al., 2021).

Urbanization has experienced a swift surge in developing nations, with South Asia, Latin America, and sub-Saharan Africa standing out as key hotspots. Approximately 15% of the global urban population is concentrated in South Asia, contributing to this trend, and it is attributed to changes in land use (Kotharkar & Surawar, 2016). There has been a notable increase in publications related to the UHI phenomenon in South Asia, reflecting significant concerns about its implications in major cities such as Delhi, Chennai, Dhaka, Chittagong, Khulna, and Kolkata (Hidalgo-García & Arco-Díaz, 2022; Kotharkar & Surawar, 2016). The rising population density in Bangladesh’s urban areas is leading to a decline in surface water bodies and vegetation, accompanied by an expansion of built-up areas. This trend is a key factor contributing to the accelerated increase in LST (Khan et al., 2019). In addition, Kalnay and Cai (2003) emphasized the significant impact of land use change, estimating a surface warming of 0.27 °C per century in continental America (Kalnay & Cai, 2003). In this respect, according to Ullah et al. (2019), potential mitigative measures for addressing urban environmental issues include urban plantation and decentralization. These measures can be evaluated using parameters such as NDVI (Normalized Difference Vegetation Index) and NDBI (Normalized Difference Built-Up Index) (Ullah et al., 2019). NDVI values serve as indicators reflecting the condition or status of vegetation (Sultana & Satyanarayana, 2018). The NDVI process generates a single-band dataset primarily depicting healthy biomass. Studying the progression of these phenomena in the urban environment is essential for mitigation efforts, aiming to establish a balance within a city. Insights from research studies can equip urban areas to enhance sustainability and resilience, particularly in addressing the escalating temperatures associated with climate change (Hidalgo-García & Arco-Díaz, 2022).

The UHI effects can be reduced by proper landscape design in urban planning (Ünal et al., 2020). As the UHI is more evident at night, a nighttime cooling effect should be the norm, rather than an exception, for any potential UHI mitigation strategy (Deilami et al., 2018). Various approaches can be used to mitigate UHI, such as installing cool or vegetated green roofs, planting trees and vegetation, and replacing typical paving surfaces with cool pavements (Kyriakodis & Santamouris, 2018; Zhou & Shepherd, 2010). Cool pavements represent various materials and technologies used in pavement modification to lower their surface temperature and the quantity of heat released from their surface compared to traditional pavements (Vujovic et al., 2021). The most efficient urban heat-island mitigation strategies can be implemented to safeguard the vulnerable urban environment and enhance the quality of life of the population through sustainable urbanization and the implementation of nature-based solutions as suggested by the European Commission (EC, 2015). Among the most effective mitigation actions to counteract the SUHI phenomenon, tree-cover-based greening represents one of the more efficient nature-based heat-related mitigation strategies, providing cooling effects and several other environmental and air-quality advantages. For this reason, planting more trees should receive attention in cities with the UHI phenomenon (Amani-Beni et al., 2019; Ziter et al., 2019).

With an estimated population of 162 million, Bangladesh is experiencing rapid changes in its land use and cover. These changes are characterized by the substantial conversion of arable land and forest cover into urban surfaces. Bangladesh is recognized as one of the most vulnerable countries to the impacts of climate change (Dewan et al., 2021; Rahman & Shozib, 2021; Xu et al., 2020). There is an observed increase in the built-up area at a rate of 3.55 km2 per year in the industrial megacity of Chattogram. Concurrently, the LST has risen from 20.17 °C in 1990 to 25.83 °C in 2018 (Roy et al., 2020). Here, NDBI exhibited a positive correlation with LST. In contrast, vegetated hilly areas, forest covers, and floodplain regions showed a negative correlation with LST. Similar studies conducted on other major divisional cities in Bangladesh have also shown these trends (A. M. Dewan & Corner, 2012). Also, a study conducted on Narayanganj city in Bangladesh reveals a significant impact of built-up areas, particularly those housing heavy industries, on LST. The mean LST between 2011 and 2019 increased by 1.86 °C. The study identifies a negative correlation between LST and NDVI, indicating that green coverage mitigates the urban temperature effect. Conversely, a positive correlation between LST and NDBI suggests that built-up or paved areas intensify LST, contributing to the development of the UHI phenomenon (Rashid et al., 2022).

Furthermore, a study conducted on Khulna City indicated that the rapid reduction of green spaces and the expansion of built-up areas are significant contributors to the escalating UHI effect. The city has experienced an overall increase in LST of nearly 2 °C in the past 5 years. This underscores the need for more urgent climate adaptive planning and actions to address the rising temperatures (Chowdhury et al., 2021; Leya et al., 2022). These findings collectively highlight the necessity to examine cities and towns experiencing swift urbanization. Gazipur, situated near the Dhaka Metropolitan area, stands out as one such city undergoing rapid industrialization, mirroring trends observed in Chattogram and Khulna. Examining alterations in space and time holds significance across various domains such as agriculture and industrial sectors (Belenok et al., 2021). It is crucial to comprehend the evolving patterns of LST and associated factors to efficiently tackle challenges arising from UHI effects. Simultaneously, the risks associated with changes in land use can be mitigated by implementing appropriate measures, such as effective land use management (Noszczyk et al., 2020).

Cities around the world are undergoing rapid urbanization as a result of uncontrolled physical infrastructure growth and poorly managed urbanization (Chuai et al., 2015). Due to rising urbanization, land cover (LC) is undergoing a drastic transformation. Rapid replacement of vegetated land with impermeable land (built-up areas) is a major factor in the increase in Land Surface LST (Faisal et al., 2021). Increased LST worsens the temperature in cities and creates the Surface Urban Heat Island (SUHI) effect. Urban Thermal Field Variance Index (UTFVI) phenomena can be used to quantitatively and qualitatively analyze the SUHI effect’s vulnerability (Naim & Kafy, 2021). Urbanization is expanding and to ensure sustainable management of land and vegetation, it is crucial to understand the changing pattern of vegetation and LST of growing city areas in Bangladesh. While numerous studies have explored LST and LULC in major cities of Bangladesh (Bari et al., 2021; Imran et al., 2021; Leya et al., 2022; Roy et al., 2021), there is a notable gap in research concerning the Gazipur district. Furthermore, several studies have been conducted on Gazipur district regarding temporal changes in urban settlements and land surface temperature (LST) (Roy et al., 2021). Similarly, studies focusing on the Dhaka district using GIS and remote sensing approaches to measure summer-winter variations in LULC indices and surface temperature (Sresto et al., 2022) have been conducted. Moreover, studies focusing on the Greater Dhaka region that explore the relationship between climate change and land-use dynamics with satellite-derived surface indices and temperature (Kulsum & Moniruzzaman, 2022) have been conducted. However, studies examining the relationship among NDVI, LST and UTFVI to measure vegetation loss and LST specifically in Gazipur district, Bangladesh, are lacking at a significant level. Given its current significance in terms of industrialization and urbanization, conducting a study in this area becomes highly crucial. Following the background, this study is conducted to investigate and quantify the reciprocal link between land surface temperature and vegetation loss to understand how it affects the SUHI phenomenon in the Gazipur District. Specifically, this study is conducted to measure vegetation loss and LST during the period between 2000 and 2020 and explore the relationship between NDVI and LST and between LST and UTFVI. The study’s findings will help urban climatologists, urban planners, architects, and politicians develop climate-resilient policies, frameworks, and decisions to balance urban green spaces for a healthy urban environment.

2 Methodology

2.1 Study area

This study was conducted in the Gazipur district of Bangladesh, which is situated in the north of Dhaka, the capital of Bangladesh. It is located between 23°53’ and 24°21’ N latitudes and between 90°09’ and 92°39’ E longitudes (Fig. 1). The total area of the district is 1806.36 Sq. km consisting of 5 Upazilas, 4 Municipalities, and a City Corporation. The district’s annual average temperature ranges from a maximum of 36 °C to a minimum of 12.7 °C (Simu et al., 2016).

Fig. 1
figure 1

Location of Gazipur district in Bangladesh (Location of study area)

Full size image

2.2 Data

The satellite imagery data used in this study were collected along a specific geographic path and row specified as 137/43 using two different Landsat sensors: Landsat 5 TM (Thematic Mapper) and Landsat 8 OLI (Operational Land Imager). The cloud coverage percentage at the time of each image acquisition is shown in Table 1. Two separate times, on February 20, 2000, and January 30, 2010, Landsat 5 TM collected data with no cloud cover. On the other hand, on February 11, 2020, Landsat 8 OLI took images that showed slightly more cloud coverage of 0.08%.

Table 1 Data description of this study
Full size table

2.3 Normalized difference Vegetation Index (NDVI) calculation

One popular vegetation index that is derived from satellite or aerial imagery is the Normalized Difference Vegetation Index (NDVI). It offers details on the quantity and condition of vegetation in a certain area. The reflectance of two bands typically the red and near-infrared (NIR) bands from remote sensing data is used to calculate the NDVI. The following is the NDVI formula (Eq. 1):

$$\:NDVI=\frac{\text{N}\text{I}\text{R}-\text{R}\text{E}\text{D}}{\text{N}\text{I}\text{R}+\text{R}\text{E}\text{D}}$$
(1)

Here, NIR refers to the surface-reflectance values of Band 4 for Landsat 5 TM and Band 5 for Landsat 8, and Red refers to the surface-reflectance values of Band 3 for Landsat 5 TM and Band 4 for Landsat 8 OLI (Yue et al., 2007). The resulting NDVI values are between − 1 and 1. Surfaces with no vegetation, including aquatic bodies or bare ground, are represented by values close to or below zero. There is vegetation when the value is between 0 and 1, and higher values indicate denser and healthier vegetation.

2.4 Land Surface temperature (LST) calculation

The study used Landsat TM 5 (Band 6) data prior to 2000 to 2010 and Landsat OLI 8 (Band 11) in 2020. Landsat imagery was used in this study to obtain Land Surface Temperature (LST) data for the years 2000, 2010, and 2020, with geometric and radiography corrections to the thermal bands. Landsat sensors record temperature data as digital numbers (DNs). These DNs were converted into LST for Landsat 5 data and Landsat 8 OLI data. The Stefan-Boltzmann Law was used to compute the Land Surface Temperature (LST).

2.4.1 Digital Numbers to Radiance Conversion

If the thermal band digital numbers (DN) are available, use the following formula to convert them to at-sensor radiance (Eq. 2):

$$\:Radience={\text{K}}_{1}+\:{\text{K}}_{2}\times\:DN$$
(2)

Where, K1 (constant) = 1260.56 K and K2 = 607.66 (mW × cm−2 × sr−1 μm−1).

2.4.2 Apply the Law of Stefan-Boltzmann

Apply the Stefan-Boltzmann Law, which connects a black body’s radiance to its temperature, using the at-sensor radiance (Eq. 3).

$$\:LST=(\frac{{\text{K}}_2}{\text{l}\text{n}(\frac{{\text{K}}_1}{Radience}+1)})-273.15$$
(3)

Where LST indicates the temperature on the land surface in degrees Celsius, the satellite metadata includes the thermal band calibration constants K1 and K2. radiance is the at-sensor radiance in the thermal band. Here’s a more comprehensive description of each part.

2.4.3 Conversion to celsius

The result is initially in Kelvin. Converted it to degrees Celsius by subtracting 273.15.

2.5 Urban Thermal Field Variance Index (UTFVI) and Surface Urban Heat Island (SUHI) calculation

The phenomenon called the Surface Urban Heat Island (SUHI) occurs when temperatures in urban areas are greater than in adjacent rural areas. Here’s a simplified formula for calculating SUHI (Zhang et al., 2006) (Eq. 4).

$$\:SUHI=(\frac{{\text{T}}_{S}-{\text{T}}_{m}}{STD})$$
(4)

Where Ts represents the surface temperature, Tm represents the area’s mean surface temperature, and STD stands for the standard deviation. In this study, the equation was used to apply the Urban Thermal Field Variance Index (UTFVI), which provides a quantitative understanding of the SUHI vulnerability (Eq. 5) followed by the methodology described by Naim & Kafy (2021), Rashid et al. (2022) and Tepanosyan et al. (2021).

$$\:UTFVI=(\frac{{\text{T}}_{S}-{\text{T}}_{m}}{{\text{T}}_{m}})$$
(5)

Where Ts represents the surface temperature, Tm represents the area’s mean surface temperature. The Urban Thermal Field Variance Index (UTFVI) classifies urban thermal field variance into distinct phenomenon types based on specific score ranges (Singh et al., 2017). When the UTFVI score is less than 0, it falls into the “None” category, indicating an absence of significant urban thermal field variance. A score between 0 and 0.005 is labeled as “Weak,” suggesting a minimal level of variance. In the “Middle” category (score range: 0.005–0.01), there is a moderate level of urban thermal field variance. Moving up the scale, “Strong” (0.01–0.015), “Stronger” (0.015–0.02), and “Strongest” (> 0.02) categories represent increasing levels of intensity in the observed phenomenon, with a UTFVI score greater than 0.02 signifying the highest level of urban thermal field variance.

2.6 Statistical analysis (correlation calculation)

In Excel, a correlation graph was generated by organizing NDVI and LST, UTFVI, and LST data into two columns and creating a scatter plot. A trendline was added with equation and R-squared values displayed, visually representing the correlation between NDVI and LST for analysis and interpretation in this journal entry. The graphical representation aids in conveying the strength and direction of the relationship between vegetation index and land surface temperature in the dataset.

2.7 Landuse wise temperature change calculation

To analyze land use-wise temperature changes over different years using ArcGIS software, a systematic approach was used to calculate the land surface temperature and land class of the same pixel. This involved creating random points within each land use class and then calculating the temperature for the exact same pixel locations represented by these points.

To determine sampling points for correlation analysis using a regression equation between land cover, UTFVI, and LST, a stratified random sampling technique is highly effective. Begin by categorizing the study area into distinct strata based on different land cover types. This ensures that all land cover categories are adequately represented. Within each stratum, randomly select sampling points to avoid bias and capture the diversity within each land cover type. Using GIS and remote sensing tools, collect data on LST, land cover, and UTFVI at these sampling points. This data collection should include high-resolution imagery and temperature readings. Once the data is collected, perform a regression analysis to explore the relationship between the variables. The regression equation will help quantify the impact of land cover and UTFVI on LST, providing insights into how different land cover types and urban heat effects contribute to temperature variations. This method ensures a thorough and unbiased analysis, leading to more accurate and meaningful results.

The methodology used in this study is shown by a flowchart in Fig. 2.

Fig. 2
figure 2

Methodology Flowchart of the study

Full size image

3 Result and discussion

3.1 Changes of normalized differences Vegetation Index value

The NDVI values for the 20 years, from 2000 to 2020, are presented in Table 2; Fig. 3, and Fig. 4, which offer important insights into the dynamics of vegetative health. A thorough picture of the distribution and central trend of the NDVI values throughout this time is provided by these important summary statistics. The maximum NDVI values, which occurred in 2000 at 0.55 and were subsequently followed by 0.36 in 2010 and 0.45 in 2020, represent the highest degree of vegetation greenness seen in each year. The maximum levels of vegetative vigour are represented by these values.

Table 2 Statistical description of NDVI values from 2000 to 2020
Full size table
Fig. 3
figure 3

NDVI value change during the year of 2000–2020

Full size image
Fig. 4
figure 4

Normalized Difference Vegetation Index (NDVI) of 2000, 2010, and 2020

Full size image

In contrast, the lowest recorded levels of vegetation greenness appear in the limited NDVI readings. The presence of non-vegetated surfaces was indicated by the minimum NDVI of -0.28 in the year 2000. This minimum improved to -0.09 in 2020 after rising to -0.13 in 2010, indicating a gradual decline in the amount of unvegetated land.

The average greenness for each year over the whole dataset is given by the mean NDVI values. The increase in the trend over the two-decade period, from 0.21 in 2000 to 0.22 in 2020, points to a general improvement in the health of the vegetation. The middle points of the NDVI distributions are represented by the median NDVI values, which act as reliable indicators of central tendency. From 0.22 in 2000 to 0.26 in 2020, the median NDVI values increased, indicating a positive movement toward higher vegetation greenness in the center of the range. The degree of variability or spread in the dataset is quantified by the NDVI values’ standard deviation. The standard deviation, which ranges from 0.09 to 0.11, has not changed much over time.

3.2 Changes of Land Use and Land Cover (LULC) in different years

Furthermore, Figs. 5 and 6 show that land cover changes in Gazipur over the years 2000, 2010, and 2020 reveal significant trends across different categories. The area of Water bodies has decreased consistently, from 140.34 Sq. km in 2000 to 117.38 Sq. km in 2010, and further down to 111.43 Sq. km in 2020. In contrast, the Build Up area has shown a gradual increase, from 332.88 Sq.km in 2000 to 345.92 Sq.km in 2010, and reaching 373.13 Sq. km in 2020. Vegetation area experienced a decline from 990.81 Sq. km in 2000 to 853.05 Sq. km in 2010 but then saw a substantial recovery to 998.27 Sq. km in 2020. The category labelled as Others saw an initial increase from 354.67 Sq. km in 2000 to 502.36 Sq. km in 2010, followed by a notable decrease to 335.87 Sq. km in 2020. These shifts indicate a complex interplay of urban development, environmental conservation efforts, and possibly changes in land use policies in Gazipur over the two decades. This finding aligns with recent research on Chattogram City (Gazi et al., 2021; Roy et al., 2020), indicating a gradual increase in built-up area alongside a corresponding decrease in other land covers over time.

Fig. 5
figure 5

Land use and land cover maps in different years (2000, 2010, and 2020)

Full size image
Fig. 6
figure 6

Percentage of different land use and land cover statistics in different years (2000, 2010, and 2020) (Sq.km)

Full size image

3.3 Change of temperature in different years

Table 3 provides a comprehensive statistical summary of Land Surface Temperature (LST) data for the two decades between 2000 and 2020. The maximum LST values show the greatest temperatures ever recorded annually. They show a possible increasing trend, increasing from 25.83 in 2000 to 32.55 in 2020. On the other hand, the minimum LST values represent the lowest temperatures; they vary from 18.38 in 2000 to 17.76 in 2020, indicating differences in surface cooling. A warming tendency in the distribution of temperatures overall is suggested by the mean and median LST readings, which show a general upward trend over time. The degree of variability in LST is shown by the standard deviation values, which show a significant increase from 0.68 in 2010 to 1.15 in 2020, highlighting the increasing temperature heterogeneity.

Table 3 Statistical description of LST (Degree Celsius) from 2000 to 2020
Full size table

Table 4; Fig. 7 represent the change in temperature over the 20 years based on these four land-use types: bare land, built-up area, vegetation, and waterbody. Bare land has experienced a significant increase, especially from 2010 to 2020. The temperature was 21.19 0C, 21.62 0C, and 25.96 0C in 2000, 2010, and 2020 respectively. The built-up area which resembles the impervious surfaces, has also faced a gradual increase in temperature. In 2000, it was 21.92 0C, 21.18 0C in 2010, and 23.93 0C in 2020.

Table 4 Landuse wise Mean temperature (in 0C)
Full size table
Fig. 7
figure 7

Land Surface Temperature (LST) of 2000, 2010, and 2020

Full size image

The mean temperature changes over the studied years and land use categories reveal significant trends in the region. From 2000 to 2010, bare land temperatures increased by 0.43 °C, while built-up areas experienced a cooling trend of -0.74 °C. Vegetation and water bodies both showed slight decreases of -0.24 °C and subsequent increases of 0.3 °C, respectively, during this period. The subsequent decade (2010–2020) demonstrated substantial temperature increases for all land use types, with bare land, built-up areas, vegetation, and water bodies experiencing rises of 4.34 °C, 2.75 °C, 2.49 °C, and 1.93 °C, respectively.

Understandably, the number of roads, buildings, etc. is increasing day by day, while a similar kind of trend goes for vegetation and waterbody. In the case of vegetated areas, the temperature in 2000 was 21.14 °C, in 2010 it slightly decreased to 20.9 °C, and then in 2020, it jumped to 23.39 °C. At last, in the waterbody, there is a gradual increase in temperature i.e., 20.83 °C, 21.13 °C, and 23.06 °C in 2000, 2010, and 2020 respectively.

3.3.1 Statistical examination of the relationship between NDVI and LST

The simple linear regression analysis between NDVI and LST indicates a statistically significant relationship (Table 5). The Multiple R value of 0.157 shows a weak positive correlation between NDVI and LST. However, the R Square value of 0.025 reveals that only about 2.5% of the variability in LST is explained by NDVI. The Adjusted R Square of 0.024 adjusts slightly for the number of predictors in the model. The Standard Error of 1.142 indicates the typical distance that observed values fall from the regression line. The analysis includes 1406 observations.

Table 5 Correlation analysis between LST and NDVI
Full size table

The ANOVA results show a Significance F (p-value) of 3.466e-09, confirming the model’s statistical significance (p < 0.05). The intercept has a p-value of 0, confirming its significance, with an expected LST value of 23.990 when NDVI is zero (Table 6). The slope for NDVI is -1.734, with a p-value of 3.466e-09, indicating a significant decrease in LST for each unit increase in NDVI. Despite the significant p-values, the low R Square value suggests that other factors also influence LST beyond NDVI.

Table 6 ANOVA analysis involving Land Surface Temperature and Normalized Difference Vegetation Index
Full size table

3.3.2 Statistical examination of the relationship between LST and UTFVI

The simple linear regression analysis between LST and UTFVI reveals an almost perfect relationship (Table 7). The Multiple R-value of 1.000 indicates a perfect positive correlation between LST and UTFVI. The R Square value of 1.000 shows that 100% of the variability in LST can be explained by UTFVI. The Adjusted R Square is also 1.000, confirming the model’s robustness. The Standard Error of 0.000 suggests that the observed values perfectly fit the regression line. The analysis includes 1406 observations.

Table 7 Correlation analysis between LST and UTFVI
Full size table

The ANOVA results indicate a Significance F (p-value) of 0.00, highlighting the model’s statistical significance (p < 0.05) (Table 8). The intercept is -1.00, with a p-value of 0.00, confirming its significance. The slope for UTFVI is 0.0424, with a p-value of 0.00, indicating a significant positive relationship between LST and UTFVI. The high t-statistics for both the intercept (-1473452.1349) and the slope (1476015.6054) further emphasize their significance. Overall, the regression model demonstrates a perfect fit with significant p-values, indicating a strong and reliable relationship between LST and UTFVI.

Table 8 ANOVA analysis involving Land Surface Temperature and Urban Thermal Field Variance Index
Full size table

3.4 Estimation of SUHI and UTFVI

To assess the SUHI effect, the Urban Thermal Field Variance Index (UTFVI) has been quantified. Figure 8 is a visualization of SUHI-affected areas in Gazipur. The index is classified into six categories defining separate SUHI meanings.

Fig. 8
figure 8

UTFVI of Gazipur District in 2000, 2010, and 2020 respectively

Full size image

By calculating, the area of the strongest SUHI zone has increased by around 28% from 2000 to 2020. Besides, the highest index value in 2000 was 0.39, which increased to 0.43 in 2010. After 10 years in 2020, it became 0.49. Hence, it is evident that the intensity of the SUHI effect is rising.

3.5 Assessing the relationship between LST, UTFVI, & NDVI

3.5.1 Correlation between NDVI and LST

Figure 9 represents the correlation between NDVI and LST across three-time points (2000, 2010, and 2020) that reveal notable patterns. The linear regression equations for the relationships are: in 2000, y = -2.0472x + 21.869; in 2010, y = -2.93x + 21.503; and in 2020, y = -1.734x + 23.99. The correlation coefficients (R) for each year are − 0.2058, -0.3805, and − 0.1567, respectively, indicating a weak negative correlation between NDVI and LST. Furthermore, the corresponding R-squared values (R²) of 0.0423, 0.1448, and 0.0246 suggest that the linear regression models explain only a limited portion of the variability in the data, emphasizing the complexity of the relationship between NDVI and LST.

Fig. 9
figure 9

Correlation between NDVI and LST of Gazipur District in 2000, 2010,2020

Full size image

3.5.2 Correlation between UTFVI and LST

The correlation analysis of UTFVI and LST over three-time points (2000, 2010, and 2020) shows a strong and complete positive linear relationship. The relationships’ linear regression equations in 2000, 2010, and 2020 are strikingly similar: y = 0.0467x − 1, y = 0.0424x − 1, and y = 0.0476x – 1 (Fig. 10). Every year’s correlation coefficients (R) are 1, showing a perfect positive correlation, and the corresponding R-squared values (R²) are likewise 1, indicating how well the linear regression models account for the data’s variability. These results highlight a clear and constant correlation between UTFVI and LST throughout the study, indicating that alterations in the proportion of urban vegetation are a strong predictor of corresponding variations in LST.

Fig. 10
figure 10

Correlation between UTFVI and LST of Gazipur District in 2000, 2010, and 2020

Full size image

Urban areas are notable for their numerous heat sources, which significantly affect the urban thermal environment and contribute to the formation of UHIs (Scafetta & Ouyang, 2019). Over the past thirty years, rapid urbanization has transformed natural landscapes into built environments, while the replacement of vegetated land with impermeable surfaces or built areas has resulted in adverse socioeconomic and environmental impacts both globally and locally (Kikon et al., 2016; Ranagalage et al., 2018). It is vital to understand the impacts of vegetation loss and LST on SUHI, while vegetated land can regulate the SUHI effect, as discussed by a study conducted on Sri Lanka (Galagoda et al., 2018). Also, a study conducted by Akter et al. (2021) found that the lowest LST in the north western region of Bangladesh was observed in areas with vegetation cover, while the highest LST was found in built-up and riverside areas over 28 years.

Gazipur district has experienced exponential growth in both population and industrial development over time. This surge in industrial and business activity has led to a significant transformation in land use. This study seeks to assess the decline in vegetation and changes in LST in Gazipur district from 2000 to 2020 and investigate the connections between NDVI, LST, and UTFVI. The increase in mean and median NDVI values over the two decades indicates an overall improvement in vegetative health across the district. This trend suggests positive impacts from conservation initiatives or natural regeneration processes. Also, regarding land cover changes, the consistent decrease in water bodies alongside a gradual increase in built-up areas highlights ongoing urban expansion in Gazipur. These changes signify the conversion of natural landscapes to urban infrastructure, posing challenges for water resources and biodiversity conservation. Previous research indicates that water bodies in Dhaka city and its surrounding areas have decreased over time, with many of these areas being converted into built-up regions or other types of land cover (Imran et al., 2021; Moniruzzaman et al., 2020).

On the other hand, the overall trend in UHI effects is closely linked to LST values, which are influenced by different LULC types. Industrial areas exhibit the highest LST among all LULC types and are particularly affected by the NDVI compared to other areas. Here, a positive correlation was observed between NDVI and LST, and UTFVI and LST. The regression results were strong, supporting the observed correlations, and similar findings were reported by study conducted by Rashid et al. (2022). This suggests that increasing vegetation cover in the Gazipur district could be the most effective strategy for reducing temperatures. Enhancing vegetation, particularly in built-up and industrial areas, as well as surrounding residential zones, may significantly improve the urban thermal characteristics of the region. Based on the observations of this study, urban planners and policymakers can prioritize enhancing urban green spaces to mitigate the SUHI effect based on the study’s findings. Policymakers can also contribute by developing climate-resilient policies. Additionally, to promote a cooler urban microclimate, environmental engineers can advocate for the use of permeable materials in structures.

4 Conclusion

Over the past two decades (2000–2020), Gazipur has experienced significant impacts from vegetation loss and rising land surface temperatures (LST), notably contributing to the increase in its surface urban heat island (SUHI). The upper range of UTFVI’s strongest zone has consistently risen during this period, magnifying the SUHI’s influence. LST has shown a substantial increase of 6.72 °C from 2000 to 2020, marking a pronounced warming trend. Interestingly, each land cover type recorded its highest temperature in 2020, indicating a uniform warming trend across varied land uses. Surprisingly, there is insufficient evidence to support a direct link between LST increase and vegetation loss. Instead, a robust positive correlation between LST and UTFVI suggests that the growth of the UHI effect is driven by escalating LST. These findings highlight the intricate dynamics between changes in land cover, temperature trends, and urban heat island effects, offering crucial insights into Gazipur’s evolving thermal landscape. Further study can be conducted to investigate the spatial distribution of UHI, LST, and UTFVI within urban areas at various scales (local, regional, and global). Also, analysis can be done to understand how these variables vary across different types of land use, urban forms (e.g., dense urban cores vs. suburban areas), and geographic contexts.