The classification and mapping of Lapalala Wilderness Reserve, Limpopo

Abstract

Vegetation classifications and maps form the foundation of understanding spatial variation in vegetation and the environmental conditions driving the occurrence of plant assemblages and form a baseline for detecting changes in vegetation. They are considered an important tool for land-use planning and conservation of natural ecosystems, allowing managers to make informed decisions. The aim of this dissertation was to (1) create a vegetation classification for Lapalala Wilderness Nature Reserve, distinguishing major plant communities and correlated environmental factors, and (2) map the distribution of these plant communities across the study area.

Lapalala Wilderness Nature Reserve spans 48 000 ha and is part of the Waterberg Bioregion in Limpopo, South Africa. The reserve plays an important part in conservation of both flora and fauna, and to support management and develop conservation strategies, the need for an updated vegetation map was recognised. One hundred and eighty 20 x 20 m relevés (comprised of 355 species) were sampled in January-March 2019 for this study. Canopy cover was estimated for all vascular plant species and environmental variables collected in the field include bare ground, rock cover, geographic location and elevation. Slope, aspect, curvature, topographic wetness index, topographic position index, distance to water, number of years since the last fire, and the number of fires in the last 10 years were determined for each relevé. Soil samples were analysed for phosphorus, sodium, calcium, potassium, magnesium, organic carbon and pH, and their particle size distribution was determined.

The OptimClass method identified that the best data-analytical combination for this dataset was Relativized Manhattan dissimilarity index and group average clustering with 10 clusters and no data transformations.The identified communities were Community 1: Combretum molle-Schmidtia pappophoroides woodland, Community 2: Senegalia nigrescens-Heteropogon contortus woodland, Community 3: Terminalia sericea-Aristida diffusa woodland, Community 4: Burkea africana-Eragrostis gummiflua woodland, Community 5: Cynodon dactylon-Eragrostis patentipilosa grassland, Community 6: Grewia monticola-Vachellia nilotica woodland, Community 7: Euclea linearis shrubland, Community 8: Cymbopogon pospischilii grassland, Community 9: Vitex obovata-Phyllanthus parvulus shrubland, Community 10: Andropogon eucomus-Eragrostis heteromera grassland. Out of the 37 environmentals variables, 21 had a significant effect on the composition of communities, with many of these variables being related to soil texture (n = 10) and soil nutrient content (n = 7).

CART was used to map the communities. However, mapping the study area was not very accurate due to weak relationships between satellite-derived variables and the occurrence of the communities, but estimates a heterogeneous mosaic of communities. Two communities were widely distributed across the study area, Community 1, comprising 66% of the mapped area, and Community 2 (26%), with small patches of Community 3 (3 %) and Community 5 (5 %). An accuracy assessment of the map showed an overall accuracy of 70 % and kappa index of 40%.

In summary, there was no strong differentiation between the communities in terms of species composition or environmental variables, and, as a result, the plant communities do not represent clear management units. Due to a paucity of vegetation studies and landscape-scale vegetation maps in the Waterberg, this study provides an important step in developing a deeper understanding of the vegetation in this ecologically-important region.