Satellite-based machine learning analysis reveals marked vegetation loss and expanding mining zones in southwestern Nigeria, highlighting the need for sustainable land-use planning.
Key Points
To assess historical mining-driven land use and land cover dynamics from 2002 to 2024 and project future landscape changes through 2034.
Classified multi-temporal Landsat 7, 8, and 9 satellite imagery across 2002, 2016, and 2024 using Smile Random Forest (SRF) and Smile Gradient Tree Boost (SGTB) algorithms.
Integrated spatial drivers including digital elevation models, precipitation data, geology, and proximity to roads and streams to analyze transition pathways and simulate 2034 projections.
The SRF classification model achieved overall accuracies of 82% (2002), 81% (2016), and 85% (2024).
Vegetation cover declined from 55.11% in 2002 to 37.46% in 2024, whereas cultivated land grew from 42.47% to 55.51% and mining areas expanded from 0% to 4.59%.
Predictive modeling forecasts mining footprints to reach 5.15% by 2034 alongside ongoing urbanization, with geological formations and road access acting as primary driving factors.