Land-Use and Land-Cover Change Detection and Classification to Analyze Dynamics of Dragon Fruit Farming in Sand Dunes Area of Binh Thuan Province of Vietnam

Mapping of land use and land covers (LULCs) and their temporal changes are important subjects in remote sensing field. This study applies change detection and classification (CDC) in mapping and analyzing changes of dragon fruits area in sand dunes area of Binh Thuan province using MODIS NDVI time-series data from 2000 to 2020. Unlike the single-phase classification which assumes unchanged LULC in analyzed periods, the CDC divided time series into segments defined by change points for segment classification. The per-pixel change detection using Pelt method identified 18 wide-scaled change points in study area. Afterward, two-phased multi-class classification was conducted to delineate dragon fruit area using the Random Forest algorithm with statistical features of four time-series segments representing four periods 2000–2005, 2005–2010, 2010–2015, and 2015–2020. A robust paradigm combining validation of classification results with area estimation was applied to analyze changing of dragon fruit areas. The result suggests a steady increase of dragon fruit areas in sand dune areas of Binh Thuan province from 2000 to 2020 especially in two districts Ham Thuan Bac and Ham Thuan Nam and recent conversions to dragon fruits in Bac Binh, Tuy Phong, and La Gi districts.