The influence of satellite image spatial resolution on mapping land use/land cover: a case study of Ho Chi Minh City, Vietnam

Different spatial resolutions of optical satellite imagery provide different details in land use/land cover (LULC) mapping. Therefore, it is essential to decide the appropriate classification system for specific spatial resolution. This study used the same classification system to extract land use/land cover for District 2 and District7 of Ho Chi Minh City, Vietnam in 2018 from two different spatial resolutions, Sentinel-2 (10 m) and VNREDSat-1 (2.5 m), images. Using the same set of ground truth data, the accuracy of the classified maps was assessed to select the right image for the study area corresponding to the given classification system. The outcomes revealed that both optical satellite images produced high overall accuracy (83% for Sentinel-2 images and 86% for VNREDSat-1 images) on the same classification system. It is easier to classify the cropland, grassland, roads, and industrial zones with VNREDSat-1 image whereas perennials, bare land, and aquaculture were easily classified with Sentinel-2 due to their homogeneity at a lower spatial resolution. This shows the importance of choosing the right spatial resolution following the requirements of the classification system.