Comparative Analyses of Land Use and Land Cover Dynamics From Landsat-8 OLI, ASTER LIT, and Sentinel-2A MSI Imageries over Selected LGAs in Osun State.

Abstract

Land use and land cover (LULC) mapping in tropical humid environments is frequently constrained by persistent cloud cover and complex landscape heterogeneity. This study comparatively evaluates the performance of Landsat-8 OLI (resampled to 15 m), ASTER L1T (15 m), and Sentinel-2A MSI (10/20 m) imageries for multi-temporal LULC mapping (2015, 2018, and 2024) across selected Local Government Areas in Osun State, Nigeria. Supervised image classification was executed using the Maximum Likelihood Classifier across six discrete land cover classes within ENVI, ERDAS Imagine, and ArcGIS Pro environments. The results demonstrate that Sentinel-2 achieved the highest overall classification accuracy (95.40%; Kappa = 0.89), followed by ASTER (92.50%; Kappa = 0.86) and Landsat-8 (91.70%; Kappa = 0.82). The superior discrimination of Sentinel-2 was supported by a higher Jeffries Matusita distance (1.96) between undisturbed and disturbed forest classes, driven by its 10 m spatial resolution and strategic red-edge spectral bands. However, McNemar’s test revealed no statistically significant accuracy differences among sensor pairs (p > 0.05: Sentinel-2 vs. ASTER, p = 0.0663; Sentinel-2 vs. Landsat-8, p = 0.0673; ASTER vs. Landsat-8, p = 0.0653). While Sentinel-2 offered enhanced delineation of fragmented peri-urban areas, both ASTER and pan-sharpened Landsat-8 maintained high, comparable accuracies (> 90%), confirming their continued reliability for long-term historical analyses in data-scarce tropical regions. These findings provide vital empirical benchmarks for sensor selection and sustainable land management in heterogeneous humid mosaics.