Urban land cover mapping and change detection analysis using high resolution sentinel-2A data

Abstract

Land cover information is essential data required by urban planners and policy makers to understand existing development and to protect natural resources in a city or town. With the availability of high resolution satellite images from Sentinel-2A, it is now possible to prepare accurate land cover maps and the present study is an attempt in this direction. An approach based on unsupervised classification plus a post-classification editing (recoding) by referring to Google satellite images is proposed in this study and has been tested for the city of Chennai, India. An unsupervised classification using ISODATA technique with 150 clusters and 36 iterations was carried out first and then Google satellite images were used on the background to assign each cluster to a particular land cover type. The proposed approach is very promising as the overall accuracy was found to be 96% with Kappa coefficient of 0.94. It was found that the proposed approach performs well when compared to the supervised and object based classification. The land cover map from Sentinel-2A was compared with the topographical map of 1971 and it was found that there was a fourfold increase in built-up area over the years. Built-up area was induced to develop in proximity to important highways in Chennai as ribbon type of sprawl is noticed. The results showed that the availability of green space is only 7.626 m2 per person in Chennai against the recommended value of 9 m2. It was also found that almost 6 km2 of water bodies have disappeared in Chennai and buildings were constructed over them illegally. The government should ensure proper land use planning and control built-up area development in order to protect the natural resources in the city.