2.1. Study Area The city of Rajshahi is the social, economic, and administrative hub of northern Bangladesh situated on the north bank of the Ganges River having geographical coordinates 24° 12′ to 24° 42′ N latitude and 88° 15′ to 88° 50′ E longitude in the northwestern region of Bangladesh. The study area is nearly flat, with a surface elevation between 15 and 19 m.
The dry-wet tropical monsoonal prevails in RCC with a maximum temperature varies from 30-35 0 C and having an annual average rainfall of 1448 mm (Ferdous and Baten, 2011; Kafy et al., 2020a). Since the last few decades, the city is facing rapid urbanization. Along with rapid urbanization and regional climate change phenomenon has drastically altered the duration and behavior of winter and summer seasons. This has detrimental effects on the cities climate, livelihood and green cover development (Kafy et al., 2020b). Land-use history in this area shows that over 19% of the green cover area has been lost. An increment in maximum LST was about 90 C in the last 20 years due to rapid urbanization (RDA, 2003, 2008; Kafy et al., 2019c; Kafy et al., 2020b).
2.2. Data For the present research two decades viz., years 2000, 2020, and 2020 have been selected. Three Multi-spectral Landsat Satellite data were collected from the United States Geological Survey (USGS) domain for estimating the changes in VC and LST dynamics in the study area. All of these images are downloaded (Table1) for the month of April to prevent the influences of seasonal variations(Kafy et al., 2017; Kafy & Ferdous, 2018; Kafy et al., 2019a; Kafy et al., 2020a). In the image downloading process, maximum cloud coverage was set to less than 10% for ensuring a realistic estimation of LULC and LST. However, across the study region, it was near to zero percent. No additional geo-correction or image processing required for the preprocessing of images, since the Landsat Satellite data is free of radiometric and geometric distortions. Images details were gathered from the USGS repository.
2.3. Landcover Classification The satellite images obtained from Landsat sensors were enhanced in Erdas Imagine V.15 software by 3*3 majority filtering technique for better visibility. True Color Composite (TCC.) was generated using the correct band combinations for all images to choose training samples of various LULC classes (Trolle et al., 2019; Kafy et al., 2020a). The collected Landsat images were classified into four LULC categories Urban area, Vegetation cover, Water bodies, and Bare land for the years of 2000, 2010, and 2020. Maximum Likelihood Supervised Classification (MLSC) technique used to estimate the LULC classification. In the process of creating LULC maps, about 25 training samples were taken for each LULC class. The accuracy of land cover maps is measured from available field data and Google Earth images through 150 ground truth points. Kappa statistics and Confusion Matrix are considered one of the best indicators for image classification accuracy were used in the study for accuracy assessment of the classified LULC maps (Story and Congalton, 1986; Foody, 2002; Congalton and Green, 2008; Pontius Jr and Millones, 2011).
2.4. Land Use/Land Cover Transformation The transformation of one LULC to another LULC is essential to identify the most dominated LULC class in the study area. As the study aims to identify the changes of VC influence by urban development, the "combined" technique under "spatial analyst toolset" in Arc GIS 10.6 software used to estimate the transformation rate of VC pixel to the UA from 2000-2010, 2010-2020 and 2000-2020 respectively. The combined toolchains multiple rasters so that a unique output value is assigned to each unique combination of input values.
2.5. Estimation of Land Surface Temperature Using the digital numbers (DN) of the thermal bands (Band i6 in Landsat 5TM and Bands10 in Landsat 8TIRS),the LST was estimated. The spectral radiances(λ) of the Landsat 5TM and Landsat 8TIRS bands were computed at the preliminary phase, by using the equation(1) and equation(2), respectively. Lλ was used to derive the LST in Degree Celsius using the equation (3).
2.6. Temperature Variations in the Urban Area and Vegetation Cover To establish the relationship between LULC and LST, the temperature variation in different land use is important. The "Tabulate area" technique under zonal toolset in Arc GIS 10.6 was used to estimate the LST variation over different LULC classes. The tool calculates cross-tabulated areas between two datasets and outputs a table. Ward wise zone data is defined as all areas in the input that have the same value. The areas do not have to be contiguous. Both raster and feature can be used for the zone input. The vegetation and UA data was used as class raster in the tabulate area process.