The Spatial Distribution of Plant Biodiversity in the Greater Cape Floristic Region of South Africa: Comparing Distribution Models with Expert Range maps

Abstract:

This paper compares the results of spatial patterns of biodiversity in the Greater Cape Floristic Region (GCFR) derived from species distribution models (SDM) and expert range maps, and evaluates their performances in predicting the species occurrence. Our study area is GCFR in South Africa, well known for its plant biodiversity – it contains nearly 20% of Africa’s plant species in less than 0.5% of Africa’s total area. Focusing on plant species from the restio family, we map spatial patterns of species biodiversity from expert range maps and SDMs predictions in 1km res and set another independent presence/absence plot-level species data as the standard to evaluate performances of expert range maps and SDM predictions in 10th, 50th, 75th percentile threshold in several confusion matrix indices. As a result, we obtain spatial patterns of species richness, beta diversity, turnover and nestedness from expert range maps and SDMs prediction in the 10th percentile threshold. In species richness, we find that the values in patterns derived from the expert range maps are changed smoothly, in contrast, the patterns derived from SDM predictions clarifies the biodiversity hotspots with distinct contours. In beta diversity, lines are highlighted at range boundaries of species ranges because the resolution is too fine, and areas with high beta value are distributed along the mountain ridges in spatial patterns derived from SDM predictions. In balanced accuracy, and sensitivity, the predicted results in the 10th percentile threshold have much higher value than the 75th percentile threshold, however, we find contrary results in total accuracy and specificity.