Physio-climatic classification of South Africa's woodland biome

Authors
Citation
Dhk. Fairbanks, Physio-climatic classification of South Africa's woodland biome, PLANT ECOL, 149(1), 2000, pp. 71-89
Citations number
61
Categorie Soggetti
Environment/Ecology
Journal title
PLANT ECOLOGY
ISSN journal
13850237 → ACNP
Volume
149
Issue
1
Year of publication
2000
Pages
71 - 89
Database
ISI
SICI code
1385-0237(200007)149:1<71:PCOSAW>2.0.ZU;2-N
Abstract
In an effort to develop more holistic ecosystem approaches to resource asse ssment and management, landscapes need to be stratified into homogeneous ge ographic regions. These regions can then be used in a monitoring framework to develop reliable estimates of ecosystem productivity. A regional charact erization of the woodland biome has been developed for South Africa, deline ated by satellite imagery and using environmental data and a rigorous stati stical methodology. Distribution maps of key environmental variables are an alyzed by factor analysis, an iterative clustering technique and maximum li kelihood classification to quantify and identify homogeneous physio-climati c units. A spatial clustering technique was used to identify regions, which are stat istically different with regard to five physiographic, climatic and edaphic variables deemed important within southern African savanna woodlands. The woodland biome of South Africa at 1km resolution was successively divided. Thirty year mean monthly temperature, total plant-available water balance o f soil, elevation, landscape topographic position, and landscape soil ferti lity were used as input classification variables. The map data were submitted to a factor analysis and varimax axis rotation. The factor analysis removes correlations from the input variables, reduces the dimensionality, and normalizes the axis measurements. A cluster analys is was performed on the three principal factor scores using a modified iter ative optimization clustering procedure to determine the finest level of cl asses statistically permitable. Twenty-seven identified unimodal cluster si gnatures were then submitted to a maximum likelihood classification where t he statistical probability of the GIS cell assignment is carried out to det ermine class membership. The final map of custom physio-climatic regions is described, and these custom regions are compared with a vegetation potenti al map of the woodland types identified in the South African summer rainfal l zone.