Accuracy assessment of land use land cover LULC 2020 (ESRI) data in Con Dao island, Ba Ria - Vung Tau province, Vietnam

Abstract

Managing land use in an efficient way results in food security and economic development. Recently, the Environmental Systems Research Institute (ESRI) announced a new product, LULC 2020 (ESRI), that has been produced by the classification with Artificial Intelligence (AI) technique of satellite data at 10-meter resolution. This data has global coverage and is freely available on the internet, but the questions are how good the data is and how to apply it to local practice and land use system. This study aims to answer those questions by emphasizing the accuracy assessment of the LULC 2020 (ESRI) data. The systematic samples of 281 test sites and the Geographic Information Systems (GIS) analysis have been conducted by using the Confusion matrix method in comparison with the visualization of high-resolution data from Google Earth images and field survey data. Topographic information data such as Digital Elevation Model (DEM) and slope data have been integrated for achieving better data accuracy, with overall accuracy increasing from 79.72% (Overall Kappa Statistic 0.421) to 84.34% (Overall Kappa Statistic 0.589). The improved land use types are hilly non-forest, other agricultural, and flat unused types that are typical for the local land use classification system. The study provides a new data processing method to apply LULC 2020 (ESRI) for land use mapping in Con Dao island, Ba Ria - Vung Tau province in the East Sea of Vietnam.