Modelling the potential distribution of bramble (Rubus Cuneifolius) in the KwaZulu-Natal, Drakensberg, South Africa.
Ndlovu, Protasia Phindile Penelope.
MetadataShow full item record
Invasive Alien Plant (IAP) invasions have been attracting increasing attention as a result of their substantial effects on native ecosystems. Hence, tools for explaining and predicting IAP distributions have been increasingly promoted for proactive ecological management, and Spatial Distribution Models (SDMs) are one such tool. The main aim of this study was to explore the application of SDMs in modelling the potential distribution of invasive American bramble (Rubus cuneifolius) in the Ukhahlamba Drakensberg Park, South Africa. The rapid proliferation of this alien plant has had significant adverse impacts on native plants and the stability of grassland ecosystems. However, there is lack of adequate data on its distribution and factors potentially influencing its present-day habitat range expansions. In that regard, the first objective provides a review of the application of SDMs in modelling the distribution of IAPs and associated challenges and opportunities. As a result of the limitations in traditional methods such as ground surveys, SDMs have demonstrated potential in providing relatively quick and feasible means of predicting IAP distributions, ecological niches and suitability of areas not yet invaded. Literature has shown growth in the use of SDMs for predicting biological invasions with presence-only methods gaining popularity than traditional analyses requiring both presence and absence data. Comparative analyses of model performance found contemporary methods such as Maximum Entropy (Maxent) to have better statistical performance compared to well established modelling approaches. Recent studies also demonstrated that remotely sensed data offers opportunities to explore underlying ecological relationships of species beyond climatic factors and improve the performance of SDMs. The second objective was to model the potential distribution of American bramble using topographic, bioclimatic and remotely sensed data using the Maxent modelling approach. Specifically, this study tested whether variable selection affected model accuracy and the spatial distribution of the species. Model performance was evaluated using the Area Under the curve (AUC), True Skill Statistic (TSS) and Kappa statistic. A quantitative comparison of all models showed that the model built with a composite of all variables yielded the highest AUC score of 0.957. The inclusion of spectral reflectance values improved model accuracy from 0.896 to 0.949. Elevation and rainfall of driest quarter were the most influential variables in modelling bramble distribution. Results of this study showed that bramble are species characteristic of warmer areas with sufficient rainfall and low elevation ranges. In addition, this study demonstrated that the Maxent approach based on topographic, bioclimatic and spectral reflectance values effectively predicted areas susceptible to bramble invasion. Overall, identification of these areas would assist to guide appropriate management measures and control further incursions.