Forest and Non-Forest Maps for Muchinga Province, Zambia (2014, 2017, 2022)

Overview This dataset provides forest and non-forest classification maps for Muchinga Province, Zambia, produced using Landsat-8 OLI, ALOS/PALSAR, and SRTM data processed in Google Earth Engine (GEE). Optical, radar, and terrain features were combined to improve classification accuracy. Two indices were derived: NDVI (from Landsat-8) and VV/VH cross-polarization ratio (from ALOS/PALSAR). Texture features were computed using the GLCM method applied to seasonal NDVI composites. A Random Forest classifier (500 trees) was trained with 2,272 samples (1,696 forest, 576 non-forest) collected from high-resolution imagery. Maps Accuracy For validation, 300 random samples were used, achieving overall accuracies between 0.90 and 0.94 for 2014, 2017, and 2022. Table 1: Forest and Non-Forest maps validation, including overall, producer (PA), and user (UA) accuracy values. class 2014 2017 2022 UA PA UA PA UA PA Forest 0.93 0.91 0.93 0.92 0.91 0.9 Non-Forest 0.9 0.92 0.92 0.92 0.9 0.91 Overall Accuracy 0.92 0.92 0.91 Spatial coverage: Muchinga Province, Zambia Temporal coverage: 2014, 2017, 2022 Spatial resolution: 30 meters Projection: EPSG:4326 - WGS 84 Data format: GeoTIFF File description Muchinga_2014: Forest and Non-Forest map 2014 Muchinga_2017: Forest and Non-Forest map 2017 Muchinga_2022: Forest and Non-Forest map 2022 Difference_maps_Muchinga_2017_2014: Difference between Muchinga_2017 map and Muchinga_2014 map Difference_maps_Muchinga_2022_2017: Difference between Muchinga_2022 map and Muchinga_2017 map Difference_maps_Muchinga_2022_2014: Difference between Muchinga_2022 map and Muchinga_2014 map Qgis_style_forest_nonforest_maps: QGIS style file for the Forest and Non-Forest maps Qgis_style_difference_maps: QGIS style file for the difference maps Muchinga_training_samples: samples used for training the models