In order to make the VEGMAP database as representative as possible, as many vegetation scientists as possible throughout South Africa have been contacted and copies of their data requested for the database. These data will be loaded into a secure database separate from the ACKDAT database but compatible with it. This will preserve the integrity of the separate databases.
Minimum data requirements
· The first requirement of the VEGMAP project is floristic data. Floristic data are lists of plant species taken at sample sites or relevés of variable size (determined by the type of plant community). Data of this kind are most often collected in different formats by different research workers and so the first important step is to standardize the data. When standardizing, taxonomic compatibility of data (i.e. the same name used for the same taxon) is the primary aim and must be achieved before analysis can start.
· Secondly, relative cover and abundance ratings should be made compatible across surveys. This, however, is less important since 'presence/absence data' is also acceptable. The analysis would hinge more on the floristic composition of stands of vegetation than on the relative abundance of the species concerned.
· The third requirement is an accurate record of the locality of a sample. The latitude and longitude of each sample location is necessary to locate and relocate sample positions. This information will be entered into a geographic information system (GIS) database which will ultimately facilitate mapping the vegetation.
A database package TURBOVEG has been designed specifically for entering vegetation data and is now available in a Windows® version TVWIN. The package is linked to the floristic lists of a country or geographic area as required. It has an automatic species-lookup procedure which enables relevé data (species lists with cover/abundance data) to be rapidly loaded into a database. In the southern African context TVWIN is linked to the Plants of South Africa database (PRECIS) consisting of about 24 000 species. Efficient species-data accession and checking and rapid loading of relevé data to a database of southern African vegetation samples is now in progress. The TVWIN computer package has been adopted as the standard database package for the VEGMAP project. In this way it will be the 'common denominator' for vegetation data accession to a single database. Data from sources such as databases in other packages can also be transferred to the TVWIN system and files in the Cornell Condensed Format can be readily imported into the database.
Classification of floristic data
The classification of large sets of vegetation data requires that the data are reduced in some meaningful way for efficient handling during analysis. With modern computing capacity it is conceivable to analyse huge data sets but the variation within such data sets may prove unmanageable. A proposal is therefore to make an decision to split the data into 'blocks' along some predetermined lines. The most obvious primary classification would be on the basis of biomes or natural regions. It is proposed that 'working groups' of vegetation scientists should be established in each biome to handle the data analysis and mapping.
With the establishment of a 'common' database, the companion program to TVWIN known as MEGATAB will be used to classify the data. This program uses TWINSPAN for analysis but has some rapid routines that greatly facilitate classification of the data sets.
Within each biome the accumulated data will be classified to produce units of vegetation. These units will be analysed together to produce an overall classification of the vegetation of southern Africa.
Ordination: relating vegetation units to the environment
The vegetation of southern Africa reflects the interaction of a large number of environmental factors the most important of which are climate and soil type (geology). No classification or map of the vegetation would be complete if these factors were not considered. Where possible environmental data will be obtained that will be used in conjunction with the vegetation data to investigate the gradients in the landscapes. The ordination of the data will be done using Canonical Correspondence Analysis (CCA) where the analysis of the vegetation data is constrained by the environmental data. This 'direct gradient analysis' method will enable the interpretation of the vegetation units in terms of their habitat and will facilitate the development of models for extrapolation of vegetation units. This procedure together with the classification will greatly enhance the accuracy of mapping the vegetation.