ASSESSMENT OF IMPERVIOUS SURFACE CHANGES FROM MULTI-TEMPORAL LANDSAT DATA AND MACHINE LEARNING TECHNIQUES: A CASE STUDY IN HO CHI MINH CITY

Abstract

Impervious surface is an artificial surface that prevents water from seeping into the ground. Impervious surface is not only an indicator of the level of urbanization but also a key indicator of the quality of the urban environment. This article presents the results of an impervious surface classification in the Ho Chi Minh City area from multi-temporal Landsat image data. Three Landsat image scenes from 2010 - 2020, including Landsat 5 images taken on February 11, 2010, Landsat 8 images taken on February 9, 2015 and February 23, 2020 are used to classify land cover/land use, including impervious surface. Three machine learning algorithms (Random Forest, Support Vector Machine, Classification and Regression Tree) and maximum likelihood method are tested to select the algorithm with the highest accuracy. The results indicate that the Random Forest algorithm achieves the highest accuracy in classifying impervious surfaces, with an overall accuracy exceeding 93% and a Kappa coefficient of 0.915. The results received in the study also show an increase in impervious surface area in Ho Chi Minh City in the period 2010 - 2020. This is important information, helping managers in monitoring and planning urban areas.