Spatiotemporal land use and land cover dynamics across Sierra Leone's districts (2017–2024) using a global deep-learning Sentinel-2 product

Abstract

Land use/land cover (LULC) dynamics are at the heart of debates about sustainable development, especially across tropical Africa, where change tends to be both rapid and poorly documented. This study tracked LULC transitions across 16 administrative districts in Sierra Leone between 2017 and 2024. We used the openly available ESRI/Impact Observatory global land-cover product, pre-classified from Sentinel-2 imagery with a deep-learning U-Net model. Tree cover shrank by roughly 3,076 km² (equivalent to 4.3% of the country’s land area), with northern districts bearing the brunt of the loss (Falaba, −746 km²; Koinadugu, −419 km²; Bombali, −414 km²). Over the same period, the rangeland area expanded by 2,758 km². The built area increased by 150 km² at an annual rate of 23.75 km²/year (p = 0.001). Growth was concentrated in the Western Area Rural (+64 km²) and regional centers such as Bo (+25 km²) and Kenema (+12 km²). Flooded vegetation expanded by 107 km² mostly by coastal districts, while cropland posted a more modest gain of 56 km². The study aims to assess the spatiotemporal dynamics of LULC across 16 administrative districts in Sierra Leone from 2017 to 2024, quantify the magnitude and spatial patterns of change across districts, and evaluate national temporal trends. This study advances SDG 15 (Life on Land) by quantifying forest loss and biodiversity habitat implications, SDG 13 (Climate Action) by linking deforestation to carbon storage reduction, and SDG 11 (Sustainable Cities and Communities) by documenting urbanization patterns and land-use competition pressures.