Question: How is pattern detected and spatial scale defined in a manner that is meaningful to management?
Location: Protected areas to the west of the Kruger National Park, South Africa.
Methods: A Two-Way Indicator Species Analysis (TWINSPAN) based on composition and structure was applied to woody survey data. The resulting TWINSPAN classes for individual monitoring sites were used in a supervised classification of Landsat ETM imagery across the study area. The training sample's co-ordinates were fed into a GIS and the resulting TWINSPAN point-feature shape file was processed using a 1 km theme-buffer function.
Results: The supervised classification using the theme-buffer signatures yielded a satisfactory overall accuracy (κ = 0.75; r2 = 0.80; p = 0.05) using a test sample compiled by reserve wardens throughout the study area. The derived vegetation map was smoothed using a majority filter and after on-screen digitizing a small gabbro intrusion, it was accepted as the best representation of the woody vegetation of the study area at a scale of 1:250 000. Seven plant communities were identified in the current study and satisfactorily accommodated within various topographical units of four extrapolated Landscapes of the Kruger National Park.
Conclusions: Vegetation patterns are described of the areas under conservation management to the west of the KNP at a spatial scale that allows for the meaningful examination and comparison of the structure, functioning, and ultimately effective management, of these savannas. This contribution thus links to the co-ordinated effort extending into the Trans-Frontier National Park in Mocambique. A key objective is to better understand the functioning of these savanna systems for effective management and we discuss some of the key ecological issues within the plant communities of each landscape. The latter illustrate the usefulness of the technique in practice.