Abstract
The research examines the application of Spectral Variation Hypothesis (SVA) in an Afromontane forest ecosystem using features derived from high and medium resolution images combined with macro-ecological data to predict tree species distribution. Alpha diversity (α) of tree species were calculated from in situ data obtained from survey of two study sites. The Object Based Image Analysis (OBIA) was adopted for the tree species distribution modelling. Spectral and textural metrics from the both QuickBird and Landsat images were computed with the segmentation algorithm. While the macro ecological parameters (temperature, humidity, elevation and slope) were derived from 30 m ASTER DEM and CHELSA high resolution climatic data. The relationships between diversity and spectral, textural features derived from the two images and the macro-ecological parameters were assessed with random forest algorithm. Elevation (r=0.55), and slope (r=0.46) were the determinant of tree species distribution in the study area. While spectral and textural features significantly contributed to the enhancement of the alpha diversity model in both QuickBird and Landsat images. QuickBird and Landsat ETM-8 spectral and textural heterogeneity showed a significant correlation with species richness (r=0.78) and (r=0.47) respectively. The empirical models developed can be used to predict landscape-level species density in the Afromontane forests of Nigeria and the adjourning Cameron highlands.

Key word: Afromontane, Ecosystem, Tree species, Modelling, Satellite images, Macro-ecological