Assessing the importance of topographic variables for the spatial distribution of tree species in a tropical mountain forest

Abstract

Availability and improved access to high-resolution digital terrain models (DTM) enables new approaches for the analysis of spatially explicit biological data. In this study, the spatial distribution of 16 tree species in a tropical mountain rain forest in South Ecuador and its relationship with topographic variables was evaluated at a fine-scale ecological level using two presence-only species distribution modelling techniques: The maximum entropy model (Maxent) and the ecological niche factor analysis (ENFA). Spatially explicit tree data stem from long-term forest monitoring plots in three microcatchments with a total area of 11.1 ha. Topographic variables were derived from a high-resolution DTM. Model performance was assessed by the true skill statistic (TSS) and area under curve (AUC) of the receiver operator characteristic (ROC), using both a k-fold approach and null-models. Performance varied among species and techniques, but generally Maxent models showed better performance than ENFA models. Furthermore, the ecological plausibility of the models was confirmed by comparing them with a previously established forest type classification. Among the explanatory topographic variables, elevation and a Topographic Position Index (TPI) appear as the main determinants for the distribution of most of the tree species. This study demonstrates that even on a small scale, the use of presence-only species distribution modelling techniques is a viable option for modelling suitable habitat for tree species in tropical mountain rain forests, indicating suitability for supporting stand-level planning and site-species matching techniques for natural forest management.