Landsat-Based Random Forest Land Cover Classification and Validation Dataset for Deforestation Monitoring in Rangamati, Bangladesh (1993–2023)
Year: 2026
Type: dataset
Abstract: This dataset supports a multi-decadal deforestation monitoring study of Rangamati district in the Chittagong Hill Tracts, Bangladesh, covering the period 1993–2023 using Landsat 5, 7, and 8 imagery. The dataset includes:- Random Forest land cover classification maps (GeoTIFF, 30m resolution) for six epochs: 1993, 1998, 2003, 2008, 2018, and 2023- Forest binary masks per epoch (GeoTIFF)- 200-point stratified random validation sample with manual interpretation labels (CSV)- Validation point locations (KML) for Google Earth verification- NDVI change raster (1993–2023) Five land cover classes were mapped: Dense Forest, Degraded Forest/Jhum, Water, Agriculture/Settlement, and Bare Land. Independent visual validation yielded an Overall Accuracy of 77.10% (Kappa = 0.712). Area-adjusted accuracy following Olofsson et al. (2014) was 73.17%. All processing was performed in Google Earth Engine using open-source tools. Analysis scripts are available at the linked GitHub repository.
Authors: Shohan, Shohinur Pervez, Sarma, Dhiman
Institution: Rangamati Science and Technology University
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
SDGs: Life on land, Climate action
Open Access status: green
OpenAlex ID: W7204468128