National-scale land use and land cover mapping using multi-temporal Sentinel-2 satellite imagery and machine learning classification.
A national-scale land use and land cover (LULC) mapping project using multi-temporal Sentinel-2 satellite imagery. The project produced a classified map at 10m resolution covering 5 primary and 18 secondary land cover classes, updated annually to track change over a 5-year baseline period.
Classification was performed using a Random Forest machine learning algorithm trained on field-verified ground truth samples. Multi-temporal composites were generated to capture seasonal vegetation dynamics and improve class separability.
The classification includes: broadleaf forest, coniferous forest, mixed forest, scrubland, natural grassland, agricultural land, vineyards, orchards, urban fabric, industrial areas, transport infrastructure, wetlands, rivers, lakes and bare rock.
GeoTIFF classified maps, accuracy assessment report, change detection analysis, area statistics by administrative unit, and a QGIS-ready map package — all delivered in both national coordinate system and WGS84.