Determination of urban heat island by computing land use land cover and land surface temperature from satellite image using machine learning technique for chennai

Abstract

Formation of Urban Heat Island (UHI) is one among the major global newlineissues faced by both developed and developing nations. UHI formation and newlineexpansion have significant aftermath in environmental factors particularly to newlinethe climate change and imbalance, agriculture and crop cultivation, the quality newlineof water and air. Urbanization and industrialization are the two factors which newlinehave made fundamental changes in the structure of the land topology. It is newlinealso witnessed that the severity in UHI will be more in arid regions and area newlinewith rapid urbanization without standard and customized town planning. newlinePeriodic monitoring of earth surface is essential to identify and mitigate the newlineemergence of UHI. Periodic and regular examination and monitoring the newlineformation of UHI is an arduous process which requires more time, man power newlineand cost expensive strategy. Emergence and expansion of UHI can be sensed newlineby different numerical and computational methods. Prominent and familiar newlineUHI analysis methods include urban heat island intensity (UHII), Building newlineEnergy Model (BEM) and Urban Canopy Model (UCM). newlineThis study demonstrates the advantage of Remote Sensing imagery newlinetechnique for evaluating and forecasting UHI by relating LST and vegetation newlinehealth amount in a fast exploring metropolitan city Chennai, India. Vegetation newlinehealth is studied with LULC changes analysis. LULC classification is done newlinewith Support Vector Machine (SVM), Maximum likelihood classification newline(MLC), and Random Forest (RF). The LULC classifiers visualize the changes newlinethat take place in the natural structure of land cover and classify into eight newlinedifferent classes. newlineThe classifiers group each class by computing the similarity of newlinethe pixel in the image. newline