Preface : The present era is increasingly being modernization and urbanization with rapid industrialization along with which there is a decrease in the forest cover, with the root cause of population explosion, which has led to a huge increase in urban cover, which in turn has made significant changes in the land surface. Thereby it has become important for us to increase our awareness about such conditions present. I hope this project will bring a better idea of the phenomena of the increasing land surface temperature and its variation across different places on Earth and the causes behind it. This project tries to answer some questions of the remote sensing with special reference to Guwahati. Guwahati, being one of the fastest growing commercial, educational and industrial hubs of the entire north eastern region experiences a subtropical and humid climate with less extremes in the temperature variation.

GIS and remote sensing techniques such as mapping, monitoring, evaluation and modeling have been done which gives an insight into the local climate in that area concerned. To establish the relationship between land use and land cover different some biophysical indicators like that of NDVI, NDWI and NDBI have been calculated. Landscape metrics have been used which further facilitated the cluster analysis which gives an insight into the patch density.

1.1  INTRODUCTION:

Urbanization changes the natural land surface into sophisticated structures such as settlements and transportation networks which leads to change in temperatures of land surface. It is common these days that there is a higher temperature in urban areas. The developing urban infrastructure and increasing urban population is putting more pressure on natural land since it is being converted into man-made grounds (Gill, Forest, & Ennos, 2007). This change increases the land surface temperatures (LST) because of the artificial structures which keeps and emits heat, which in turn causes the urban heat Islands (UHI) (Gill et al., 2007). Land Surface Temperature is the normal temperature we feel on the Earth surface which is calculated through the remote sensing (thermal bands). The estimation of this depends on albedo, the cover of vegetation and the soil moisture present. (https://land.copernicus.eu/global/products/lst).

These days, automobiles and industries have increasingly contributed to increase in the atmospheric Greenhouse Gases (GHGs) and Chlorofluorocarbons (CFCs), which has put a lot of pressure on the vegetation cover and belts, which has turned to increase in temperatures between the urban and rural areas.

Due to this the urban areas are constantly having increasing temperatures and can be studied with temperature parameters (Isotherms), that indicate the rapid changes compared to the rural areas. This modern phenomenon is also known as Urban Heat Island (USI). The land use and land cover (supervised classification), LST and NDVI techniques are used to calculate the UHI (Borthakur & Nath, 2012). In the comparative study between the Indian cities of Mumbai and Delhi, it was revealed that Mumbai has a higher intensity of UHI than Delhi because of the higher rate of urban development in Delhi which has led to decrease in the vegetation cover. LST is a dependent variable and the NDVI is an independent variable had been used to make the regression analysis to show the correlation between these components and shows how the NDVI directly affects the LST (Grover and Singh, 2015).

The occurrence of the LST has an increasing trend in the core of the urban area of the city Kunming in China. This had a very rapid urbanization along with the increase in built up area and had a drastic change in a short period of time. The regression is calculated by the LULC, LST, NDVI and NDWI (Zhou and Wang, 2010). In Lagos (Nigeria), vegetation cover has significantly reduced from 70.043% to 10.127% over 3 decades and this huge difference has increased the land surface temperature. LST has increased through the process of urban sprawl, growing urban population, degradation of agricultural land and vegetation cover. Regression analysis was taken between LST and LULC to know about the relationship between LST and LULC over the year (Babalo and Akinsanola, 2016).

Guangzhou of southern China experienced a rapid urbanization process due to economic growth and this rapid process has resulted in formation of the Urban Heat Island (UHI) in the area. Barren land, built-up

area and grass land were found in the inner part of the city which justifies the fact that the city has experienced a high LST (high surface temperature) zone because of low thermal inertia. Forest and water bodies were found in the outer part of the city, so the outer part of the city has experienced a low LST (low surface temperature) zone because of high thermal inertia. High NDVI (Normalized Difference Vegetation Index) zone area has experienced low LST and low NDVI zone area has experienced high LST in the different parts of the city (Sun, Wu, & Tan, 2012). In India, large numbers of metropolitan cities have experienced the phenomena of Urban Heat Island (UHI) in recent decades. Recently, Guwahati has experienced UHI because of rapid urbanization and population growth. Guwahati metropolitan area became UHI in the year of 2009. The reason being the construction of new expressway from Jalukbari to Khanapara which has changed the land use pattern along with the settling up of industries along the National Highway 37 (Brothakur and Nath, 2012).

Guwahati (Assam) has experienced rapid urban growth because of rapidly growing economic activities. There is a great pressure of increasing settlement putting a gigantic pressure on forest areas, agricultural land and wetland areas in and around the premises of the city, so that can be seen a  change in LULC affects the change of LST. The aim of the study is to find out the temporal relationship between LULC and LST over 27 years (1991-2018).

GIS and Remote Sensing techniques have been used to find the relationship between LULC and LST. NDVI and NDWI is utilized as indicators of LULC and LST change using satellite images (Landsat dataset).The relationship between LULC and LST might help to provide some developmental plans in the city. If the developmental plans can be successfully articulated, the city development authority can implement some counter measures to minimize the local (micro) climatic change. This study may give significant inputs to the land use policy and planning in the upcoming satellite towns.

SIGNIFICANCE OF THE STUDY

The increasing development processes in the city of Guwahati have increased in the recent period and this has converted the natural land surface into man-made land surface in majority of the premises of the city and also led to urban sprawl. Deforestation and the subsequent changing Land use Land cover have been taken place because of the high rate of urbanization and modernization. As the vegetation is converted into built-up areas and since the built-up areas are made up with concrete and cement, land surface temperature is higher in the built-up areas and deforested areas. Due to this reason the regions of built-up areas have become the zones of urban heat Island. This phenomena is a major climatic condition specially in the urban areas since it has brought changes in the local climatic conditions which in turn has the capacity to hamper air quality and contribute to the problems in human health.

STUDY AREA (GUWAHATI) The study area, Guwahati, means “areca nut marketplace” in Assamese, (the gateway of North-East), is the capital city of Assam which is located in the north-eastern part of India (Desai, Mahadevia, & Mishra, 2014).The city is situated on the banks of Brahmaputra River and is 55 meters above the sea level. The total area of the study area is 375.781234 km2. The city is situated in the subtropical climate zone and the temperature ranges from 19°C to 26°C https://en.wikivoyage.org/wiki/Guwahati. In recent decades Guwahati has experienced some problems related to LULC change and the major problem being the increase of LST. Since Guwahati is gateway of north-east India, so the population growth has increased with the dramatic growth of urbanization. The population of Guwahati has increased from 809,895 in 2001 to 963,429 in 2011 and population density has also increased from 3736 persons per km2 in 2001 to 4445 persons per km2 in 2011. (Census of India, 2011).

RESEARCH QUESTIONS:

1. To show the changes in Land Surface Temperature (LST) changed over the period of time from 1990 to 2018?
2. Depict the relationship between changing nature of LST and LULC over the period of time?
3. What is the relationship among Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), Normalized Difference Built-up Index (NDBI) and Land Surface Temperature (LST) over the period of time?
4. What are the factors responsible for LST change and UHI?

OBJECTIVES:

The main objectives of the project are given below:

1. To examine the changes in Land Surface Temperature (LST) and Land use/Land cover (LULC) from the year 1990 to 2018
2. To examine the relationship of Land Surface Temperature (LST) with Land use/Land cover (LULC) types, Normalized Difference Vegetation Index (NDVI), Normalized Difference Built- up Index (NDBI) and Normalized Difference Water Index (NDWI).

To identify factors importance LST change and

DATA OBTAINED FROM:

Landsat 5 Thematic Mapper (TM), Landsat 7 Enhanced Thematic Mapper Plus (ETM+) and Landsat 8 Operational Land Imager (OLI) acquired from 1991, 2001, 2011 and 2018 are the main source of data for this project. The satellite images are accessible from USGS Earth Explorer which is developed by United States Geological Survey https://earthexplorer.usgs.gov/. The Landsat images have been taken for LULC classification, LST (Land Surface Temperature), NDVI (Normalized Difference Vegetation Index), NDWI (Normalized Difference Water Index) and Normalized Difference Built-up Index (NDBI) calculation. Thermal infrared band of Landsat 5, 7 and 8 has special resolution of 120m, 60*(30) m and 100m respectively https://eos.com/landsat-7/. Thermal bands of Landsat satellite images have been resampled with a pixel size of 30 meter using nearest neighbour algorithm to match with the optical bands and to estimate the change in Land use/Land cover types and Land surface temperature (Guha, Govil, Dey, & Gill, 2018).

Table 4.1. Details of the satellite image used for Guwahati
YEAR	SATELLITE	SENSOR	ACQUISITION DATE	PATH & ROW	CLOUD COVER
1991	Landsat-5	TM	1990-12-25	137/42	1.00
2001	Landsat-7	ETM+	2001-02-07	136/42	0.00
	Landsat-7	ETM+	2001-01-29	137/42	0.00
2011	Landsat-5	TM	2009-01-11	137/42	0.00
2018	Landsat-8	OLI- TIRS	2018-02-21	137/42	0.50

Note: *Resampled to 30 meter

LULC CLASSIFICATION:

To interpret and understand the Land use classes in the satellite images, it is first converted into false colour composite bands that have been used in different satellite images. False colour composite bands (4, 3, and 2) are used in different satellite images to compare between the different land cover classes. Five Land use and Land cover have been used for Guwahati LULC classification are Urban, Water Bodies, Wetland, Open Land, Forest, Plantation and Agriculture. Supervised Classification has been done to obtain the Land use/Land cover classification in Erdas Imagine (2014) and Saga GIS. Spectral signature is used for this supervised classification which is used to define the training sets and these will apply to the entire image. In this we use the Maximum Likelihood Algorithm because that can help in calculation for each class’ statistics in the bands which are distributed and it also helps in calculation of a pixel for each class. Every pixel obtained is a part of some class and this helps in highest probability. This algorithm is default mode in this supervised classification. After this classification is done, many mixed classified pixels will be added. So to minimize and remove them we will ‘recode’ the tool in Erdas Imagine software (2014). ArcMap 10.3.1 software has been used for different purposes to find out the significant outputs in LST, NDVI, NDWI, and NDBI. The data of these four aspects have been given for the time period of 1991, 2001, 2011 and 2018.

TESTING ACCURACY FOR LULC CLASSIFICATION:

‘Google Earth Pro’ software has been used to find out the accuracy for the Land use/Land Cover classification by using GCP points, 150 and 210 GCP’s points have been obtained from the Google Earth Pro to do the classification for the testing of accuracy in Guwahati. Jacob Cohen’s Kappa accuracy test has been used to estimate for every year.

• K=Po-Pe/1-Pe…………………………………………… (i)(K represents Kappa test; Po represents observed Agreement; Pe represents Expected Agreement).

1.00-0.81 = Almost perfect agreement	0.40-0.21 = Fair agreement
0.80-0.61 = Substantial agreement	0.20-0.00 = Slight agreement
0.60-0.41 = Moderate agreement	<0 = Poor agreement or no agreement

The Kappa statistics helps in giving the result to make significant interpretations. The range of this statistics ranges from 0 to 1 and this is subdivided further into 5 different classes for better understanding. (R Nichols, M Wisner, Cripe, & Gulabchand, 2010)

1. User’s Accuracy = (N𝑜. 𝑐𝑜𝑟𝑟𝑒𝑐𝑡𝑙𝑦 𝑖𝑑𝑒𝑛𝑡𝑖𝑓𝑖𝑒𝑑 𝑖𝑛 𝑡ℎ𝑒 𝑚𝑎𝑝 / 𝑁𝑜.𝑐𝑙𝑎𝑖𝑚𝑒𝑑 𝑡𝑜 𝑏𝑒 𝑖𝑛 𝑡ℎ𝑒 𝑚𝑎𝑝)…(ii)
2. Producer’s Accuracy = (N𝑜.𝑐𝑜𝑟𝑟𝑒𝑐𝑡𝑙𝑦 𝑖𝑑𝑒𝑛𝑡𝑖𝑓𝑖𝑒𝑑 𝑖𝑛 𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑝𝑙𝑜𝑡𝑠 𝑜𝑓 𝑐𝑙𝑎𝑠𝑠/ 𝑁𝑜.𝑎𝑐𝑡𝑢𝑎𝑙𝑙𝑦 𝑖𝑛 𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒

𝑐𝑙𝑎𝑠𝑠)…(iii)

1. Total Accuracy = (No.𝑜𝑓 𝑐𝑜𝑟𝑟𝑒𝑐𝑡 𝑝𝑙𝑜𝑡𝑠/ 𝑇𝑜𝑡𝑎𝑙 𝑛𝑜.𝑜𝑓 𝑝𝑙𝑜𝑡𝑠)…(iv)

LST EXTRACTION FROM THERMAL BAND:

LST is calculated with the Landsat thermal bands (Landsat 5- band 6, Landsat 7- band 6VCID (1) and Landsat 8- band 10) for 1991, 2001, 2011 and 2018. The thermal infrared band has resampled into a 30 meter spatial resolution to match the optical bands to find the relationship with LULC change. LST is calculated by the help of established formulas. (Dissanayake, Morimoto, Murayama, & Ranagalage, 2019; Grover & Singh, 2015; Kayet, Pathak, Chakrabarty, & Sahoo, 2016; Sannigrahi et al., 2018; Sannigrahi, Rahmat, Chakraborti, Bhatt, & Jha, 2017; Taylor, 2011; Yoo, 2018; Zhou & Wang, 2011).

1. Conversion of the Digital Numbers to spectral radiance:

All the objects on the Earth’s surface emits some electromagnetic thermal energy when its temperature is above zero (Kelvin). The satellite sensors receive the electromagnetic thermal energy (Kelvin) by the thermal sensors, with the principle and thermal energy can be converted to sensor radiance. The conversion of thermal energy to spectral radiance is calculated using the formula. (Ogunjobi, Adamu, Akinsanola, & Orimoloye, 2018; Pal & Ziaul, 2017; Taylor, 2011; Tran et al., 2017).

LMAX represent the spectral radiance that is calculated to QCALMAX; LMIN represent spectral radiance that is scaled to QCALMIN; QCALMAX represent maximum quantized calibrated pixel value; QCALMIN represents minimum quantized calibrated pixel value; QCAL represents quantized calibrated pixel value in DN. QCALMAX and QCALMIN are 255 and 1 for Landsat 5 and 7 and 65535 for Landsat 8 respectively. For, TM (Landsat 5) LMAX and LMIN are 15.303 and 1.238 respectively; for ETM+ (Landsat 7) LMAX and LMIN are 17.040 and 0.000 (band- 6.VDID (1)) respectively and for OLI-TIRS (Landsat 8) LMAX and LMIN are 22.00180 and 0.10033 respectively.   

	Conversion of spectral radiance to temperature in Kelvin:

 

Where, TB represents satellite brightness temperature in kelvin; K1 represents calibration constant 1; K2 represents calibration constant 2; L represents spectral radiance. For, TM (Landsat 5) K1 and K2 are 607.76 and 1260.56 respectively, for, ETM+ (Landsat 7) K1 and K2 are 666.09 and 1282.71 respectively and for OLI-TIRS (Landsat 8) K1 and K2 are 774.8853 and 1321.0789 respectively.

Correction of emissivity of surface temperature: (3)

Where, Ts represents LST in kelvin; 𝜆 represents the emitted radiance wavelength which is 11.5 µm; TB represents the satellite brightness temperature in kelvin;𝘢 represents 1.438*10-2 mk

(𝘢 calculated as 𝘢=hc/𝜎, where h represents planck’s constant 6.626*10-34 Js, represents the velocity of light 3*108 and 𝜎 represents Stefan-Boltzmann constant 1.38*10-23 J/K); Ɛ represents the surface emissivity (Yoo, 2018; Zhou & Wang, 2011).

1.  

	Conversion of surface temperature in Kelvin to Celsius:

 NDVI ESTIMATION:

NDVI is commonly and widely used for vegetation extraction which is applied in Guwahati. NDVI is used to map for estimation of vegetation coverage and health of vegetation in both the study areas from 1991 to 2018 to find out temporal change. The NDVI is generated from near-infrared and red band (Landsat 5 & 6 – near-infrared band (4) and red band (3), Landsat 8 – near-infrared band (5) and red band (4)). NDVI value ranges from +1 to -1. The positive value represents healthy vegetation cover and high density green vegetation cover, while negative value represents non-vegetative cover.

Where, NDVI represents normalized difference vegetation index; NIR represents near- infrared band of electromagnetic spectrum; RED represents red band of electromagnetic spectrum.

NDWI ESTIMATION:

Where, NDWI represents normalized difference water index; NIR represents near-infrared band of electromagnetic spectrum; GREEN represents green band of electromagnetic spectrum. NDWI index is commonly used for extraction of water body. NDWI index is used to map the extent of water body areas in Guwahati and temporal changes of water body in both the study areas from 1991 to 2018. The NDWI is generated from near-infrared and green band (Landsat 5 & 6 – near-infrared band (4) and red band (2), Landsat 8 – near-infrared band (5) and green band (3)). NDWI is used to map the extent of water body coverage area. NDWI value ranges from +1 to -1.

NDBI ESTIMATION:

NDBI Index is used for the extraction of built-up areas. In the study areas NDBI Index has been applied to detect the built-up areas. NDBI is used to map the extent of built-up area in both the study areas. The NDBI is generated from the near-infrared and green band (Landsat 5 & 6- near infrared band (4) and SWIR band (5), Landsat 8- near infrared band (4) and red band (6). NDBI values range from +1 to -1.

Where, NDBI represents normalized difference built-up index; NIR represents near infrared band of electromagnetic spectrum; SWIR represents short wave infrared band of electromagnetic spectrum.

LANDSCAPE METRICS:

Landscape Matrices had been developed for categorical map patterns and sometimes it is also used for topographic measures. These have different algorithms which quantifies the special pattern and characteristics of classes, patches and the entire landscape. (Evelin, Marc, Juri, Riho, & M, 2009; McGarigal, Cushman, Neel, & Ene, 2012; Navarro-Cerrillo, Guzmán-Álvarez, Clavero-Rumbao, & Ceaceros, 2012). The Landscape metrics is mostly used by the Fragstats software (Evelin et al., 2009). The use of these Landscape metrics have been increasing since 1999 and this is a part of broad metrics, ND which can be grouped into seven categories. These are namely;

1. Biodiversity and Habitat Analysis,
2. Use and Misuse/Selection of metrics,
3. Evaluation of landscape pattern and changes,
4. Estimation water quality,
5. Aesthetics of landscape,
6. Urban landscape pattern, road network and,
7. Management, Planning and (Evelin et al., 2009).

The landscape metrics is a combination of eight different metrics are: Area and edge metrics, Shape metrics, Core area metrics, Contrast metrics, Aggregation metrics, Subdivision metrics, Isolation metrics and Diversity metrics (Liu et al., 2017; McGarigal et al., 2012).

As urbanization is increasing over the years, urban sprawl has influenced urban growth and fragmentation. To show the spatial pattern of urban growth and fragmentation. The Landscape metrics have been used. In case of evolution of landscape pattern, the changing pattern of landscape is also used. (Evelin et al., 2009).

SPATIAL CLUSTERING:

Spatial clustering shows the different variables in a significant manner and is an important method of spatial data analysis. The data of proximal spatial data somehow contribute a role in urbanization. The scenario is simplistic to describe the pattern of urban growth, hence it misleads to policy perspective at local level. The data provides a spatio-temporal variability of driving forces by using these statistical models. A selection of an appropriate model and technique for the spatial data analysis fundamentally depends on the objectives of study. The spatial autocorrelation technique and spatial regression model have been selected to understand mutual agreements of local spatial variables, and urban pattern of growth and LULC Change.

To show the mutual understanding of the spatio-temporal urban growth or the LULC Change, spatially clustered analysis will be carried out. The Moran’s Index is an indicator of the spatial autocorrelation is performed in the study. All the analyses were executed in the GeoDa software to depict the LULC changes for each study area.

Using the formula of the Moran’s Index, we can classify those under four classes, namely:

1. High-High (HH) is the quadrant exemplifies with the positive value of all the observations in this zone and positive average of the contiguous neighbor of that location.
2. Low-Low (LL) is that quadrant exemplifies with a negative value of all the observations in that zone and negative average value of adjacent neighbors of that location.
3. Low-High (LH) is that quadrant exemplifies with a negative value of all the observations in that zone and positive average value of a contiguous neighbor of that location.
4. High-Low (HL) is that quadrant exemplifies with the positive value of all the observations in that zone and negative average value of the contiguous neighbor of that location.

Moran’s and Quadrant analysis gives an individual measure of the whole spatial pattern over the map, they are reported as Global statistics because single value describes the overall spatial pattern of the entire map. Other hand, “Local statistics” is also very much important which is used to find out the high or low locational cluster map. Local statistics gives the opportunity to test the significant location of the new cluster. LISA (Local indicator of spatial association) and Global Moran is different in their capacity to decomposition of the global indicator. LISA allows users to map out green shaded “Significance map” and blue-red shade “Clustered map.”