ANNOUNCEMENTS
Gurugram is one of the most briskly expanding cities of Northern India, growing both horizontally and vertically in terms of built up. Mostly when urban expansion is studied, only horizontal expansion is considered. Quantifying vertical expansion in terms of building heights can help in gaining a holistic understanding about urban growth patterns and their associated impacts. The aim of this study was to quantify the urban sprawl in Gurugram district, in both horizontal and vertical direction. To estimate the vertical builtup, a building heights dataset at a grid level (area of 0.1 km square) was generated using Machine Learning algorithms for the study area. Regression analysis was done taking the average building height at a grid-level as the dependent variable. Variables like NDBI, GAIA (Global Artificial Impervious Area) index, Nightlights data from VIIRS, Digital Surface Model, points of interest, building footprint count and Population were taken as independent variables. These variables were aggregated at hexagonal grid level with an area of 0.1 square kilometers. Machine Learning algorithms like RF, SVR and Linear Regression were used for the analysis. The results were validated using metrics of R squared and mean squared errors. The validation results indicated that the accuracy achieved using Linear Regression is R2=0.47, Random Forest algorithm is R2 =0.58, and SVR gave an R2 of 0.55. Random Forest gives better results out of the three algorithms both in terms of accuracy and correctness. For estimating urban sprawl in the horizontal direction a time frame from 2013 to 2022 was considered. The urban sprawl, was quantified for Gurgaon between this period using UEII index, Shannon’s Absolute and Relative Entropy values and Landscape Metrics. UEII value of 0.021 indicated a slow growth of the study area from 2013 to 2022. The values of Relative Entropy were 0.76 and 0.64 for 2013 and 2022 respectively. This indicates a move towards compact distribution of builtup. Finally the changes in the edge metrics and patch metrics corroborate the results of the Shannon’s Relative entropy, implying a compressed urban expansion in the past decade.
Keywords: Vertical Urban Expansion, Machine Learning, Building Heights.
