Estimation of Urban Sprawl Using Machine Learning Methods for the city Hyderabad, India

Abstract:

Urban sprawl is defined as an increase in the impervious surface over a region, and quantification of urban sprawl aids in urban flooding modeling. The study area chosen is the Greater Hyderabad Municipal Corporation (GHMC) in Telangana, India with a geographical area of 620 sq. kms. The city has been experiencing floods and one of the major reasons is rapid urbanization of the city. The main objective of the study is to quantify spatial- and temporal- variations of urban sprawl in the GHMC region. Two different machine learning algorithms, i.e., Random Forest- and Support Vector Machine (SVM)- are applied on the landsat data to identify urbanization patterns of the region. The data set correspond to the years 1989, 2000, 2005, 2010, 2015 and 2020. The Landsat 5 Multi Spectral Scanner (MSS) mission has provided data for the year 1989, Landsat 7 Enhanced Thematic Mapper Plus (ETM+) for 2000, Landsat 5 Thematic Mapper (TM) for 2005, 2010 and Landsat 8 for the years 2015, 2020 is downloaded from USGS website. The data handled in this study is cloud free and corresponds to months of March, April and May. The accuracy of the current study is analyzed by calculating verification metrics considering the Google Earth data as truth. A comparative analysis between the two machine learning algorithms provides insights on urbanization patterns in the GHMC region. Major increase in the impervious surface is observed from the year 2000 to 2005 and it is followed by the year 2005 to 2010.