Integrative Ecological Niche Modelling of Baobab in the Mid-Zambezi Biosphere Reserve, Zimbabwe

Understanding the spatial distribution of keystone species, such as baobab (Adansonia digitata), is essential for effective conservation planning in the face of climate change. The current study used ensemble species distribution modelling to predict the current potential distribution of Adansonia digitata within the Mid-Zambezi Biosphere Reserve (MZBR). Using the BIOMOD2 platform, Random Forest (RF) and Extreme Gradient Boosting (XGBoost) algorithms were applied to presence-only data, complemented by pseudo-absence points generated via a disk strategy. Environmental predictors included selected bioclimatic variables, land cover, slope and soil type, after filtering for multicollinearity. All models were calibrated using 5-fold cross-validation and evaluated based on TSS, ROC-AUC, and Kappa metrics, which indicated excellent predictive performance. Model outputs indicated that baobab distribution in the MZBR is primarily concentrated in the low-lying, fertile zones of the central and northern reserves. In contrast, the southern regions demonstrated lower suitability due to higher elevation and greater temperature seasonality. These patterns reflect ecological thresholds influenced mainly by temperature extremes and seasonal variability. The findings offer a baseline for identifying conservation priority areas and potential climate refugia for baobabs. Future research should incorporate biotic interactions and anthropogenic influences to enhance model accuracy, utilise fine-scale habitat data, and inform adaptive management strategies in response to projected environmental changes.