Coupling Diversity Indices with Boosted Regression Trees Predicts Species Abundance and Spatial Community Patterns in Dry Afromontane Forests

Abstract Understanding how species diversity interacts with spatial environmental gradients to shape landscape-level community patterns is a central question in contemporary landscape ecology. Although predictive modelling of biodiversity is advancing globally, dry Afromontane forests remain poorly represented in analyses that link spatial heterogeneity with community structure. This study addresses this gap by integrating diversity indices with a spatially explicit Boosted Regression Tree modelling framework to predict species abundance and community patterns across the Desa’a Forest, northern Ethiopia. Data from 301 plots were used to quantify woody species richness, abundance, and community composition, and to identify the environmental controls governing the distribution of dominant species. The forest contained ninety-four woody species and 7,207 individuals, with diversity values and plot-level abundances revealing a highly heterogeneous landscape composed of species-poor and species-rich patches. Cluster analysis delineated seven ecological communities, and dominant species such as Cadia purpurea , Tarchonanthus camphoratus , and Juniperus procera contributed disproportionately to spatial variation in forest structure. Boosted Regression Tree models showed that temperature, elevation, precipitation, and slope were the strongest predictors of Juniperus procera abundance, explaining sixty-three percent of deviance and revealing a narrow elevational and climatic niche. These findings clarify how landscape-scale environmental gradients shape species distributions and community mosaics in dry Afromontane forests. By demonstrating a data-driven approach for predicting spatial community structure, this study provides a modelling framework that can enhance conservation planning and landscape-level management in biodiversity-rich mountain ecosystems worldwide.