Modelagem da distribuição espacial da Brachystegias spp. em Moçambique

Abstract

This study aimed to evaluate the spatial distribution of Brachystegia species in the Miombo woodlands of Mozambique using Species Distribution Modeling (SDM) techniques. The analysis was based on data from the National Forest Inventory (NFI), selecting individual occurrences of Brachystegia spp. and their coordinates across all recorded plots. Environmental variables such as precipitation, temperature, hillshade, and topographic characteristics were used as independent predictors. Several modeling approaches were tested—including Random Forest, Generalized Linear Models (GLM), Generalized Additive Models (GAM), and Maxent—with Random Forest achieving the best performance, as measured by AUC, COR, TSS, and Deviance metrics. The final model enabled the creation of potential distribution maps, identifying environmentally suitable areas for Brachystegia spp. occurrence. The most influential variables were hillshade, quarterly precipitation, and the minimum and maximum temperatures. These findings provide valuable insights for forest planning and biodiversity conservation, reinforcing the importance of predictive modeling in the sustainable management of Miombo ecosystems, especially under increasing anthropogenic and climate pressures.