Integrating Sentinel-2 and Planet Optical Imagery to Monitor Land Cover and Land Use Dynamics in Cote D’ivoire
摘要
Land-use and land-cover (LULC) mapping derived from satellite imagery plays a critical role in landscape monitoring in Côte d’Ivoire. It provides key spatial information to support the development of regional and local planning frameworks. This study investigates the current LULC patterns and their recent dynamics over six years, based on land-cover datasets produced for 2016, 2019, and 2022. The methodological approach relies on the national LULC classification scheme combined with the Random Forest algorithm. A major methodological contribution of this work is the integration of PlanetScope imagery to address gaps caused by the unavailability or cloud contamination of Sentinel-2 scenes, thus ensuring the use of a consistent reference month across all three benchmark years. Map calibration and validation were conducted through systematic field surveys, drone image acquisition, and independent very-high-resolution satellite imagery analyses. By incorporating three temporal snapshots, the study provides a fine-grained assessment of spatio-temporal dynamics across the intervals 2016–2019, 2019–2022, and 2016–2022. It highlights contrasting trajectories, including acceleration and deceleration in specific land-cover transitions. Results indicate a substantial proportion of secondary forests (31%) and a notable share of dense forests (11%), driven by the presence of several classified forests, a natural reserve, and extensive old fallows that are considered secondary forest formations. Forest ecosystems showed a continuous decline over the study period, decreasing from 407,308 ha in 2016 to 362,697 ha in 2019 and 335,837 ha in 2022. The annual deforestation rates were estimated at 3.87% between 2016 and 2019 and 2.56% between 2019 and 2022. This reduction of more than 30% in the annual deforestation rate reflects the effectiveness of ongoing conservation initiatives and forest-protection efforts within the region.