Conservation planning in the face of global change is still in its infancy. A suggested approach is to
incorporate environmental gradients into conservation planning as they reflect the ecological and evolutionary
processes generating and maintaining diversity. Our study provides a framework to identify the dominant environmental
gradients determining floristic composition and pattern. Nonmetric multidimensional scaling was used on
2155 sampling plots in savanna and grassland habitat located across the province of KwaZulu-Natal, South Africa
(94 697 km2
), a floristically rich region having steep environmental gradients, to determine the dominant gradients.
Hierarchical cluster analysis was used to group similar plots which were then used in a Classification and Regression
Tree analysis to determine the environmental delimiters of the identified vegetation clusters. Temperature-related
variables were the strongest delimiters of floristic composition across the province, in particular mean annual
temperature. Frost duration was the primary variable in the Classification and Regression Tree analysis with
important implications for savanna/grassland dynamics. Soil properties (base, pH status) and moisture variables
accounted for most of the variation for the second and third axes of floristic variation. Given that climatic and
edaphic variables were well correlated with floristic composition, it is anticipated that a changing climate will have
a marked influence on floristic composition. We predict warmer temperatures may facilitate the spread of frost
sensitive savanna species into previously cooler, grassland areas. Species associated with specific soil types will not
easily be able to move up the altitudinal gradient to cooler climes because geology is aligned in an approximately
north-south direction compared with increasing altitude from east-west. Future conservation planning should
take cognisance of these gradients which are surrogates for ecological and evolutionary processes promoting
persistence.