The link between vegetation and climate at large spatial scales has encouraged the development of global vegetation models (eg BIOME 1: Prentice et al., 1992; BIOME 3: Haxeltine and Prentice,1996) which can be used to translate the output of AGCMs into maps of present (potential), future, or past vegetation (e.g. Prentice et al., 1992; Neilson et al., 1998; Harrison et al., 1995; Kutzbach et al., in press; Harrison et al., in press). The simulated vegetation distribution can be compared with modern vegetation maps (e.g. Prentice et al., 1992) or, in the case of past vegetation, reconstructions based on pollen and plant-macrofossil data (e.g. Harrison et al., in press; Jolly et al., in press). For this to be done effectively requires that palaeovegetation data are organized in a way that is compatible with model output. The aim of the BIOME 6000 project (Prentice and Webb, in press) is to translate fossil pollen assemblages into a form that allows such direct data-model comparison.

Biomization is a computer procedure that reconstructs biomes on the basis of plant functional types (PFTs, e.g., boreal summergreen tree) and their characteristic "signature" in the pollen record.  PFTs are based on their structural and functional features (e.g., leaf type, stature, bioclimatic tolerance). The biomization method involves assigning palynological taxa to one or more PFTs. The PFTs, in turn, are combined to define biomes following a specific set of algorithms or rules. The procedure for assigning a pollen spectra to a biome is as follows (see Prentice and others, 1996 for further details):

• Calculate the taxon scores (the square root of pollen %)
• Determine the relationship between the taxa and biomes (ues PFT to taxon relationship and PFT to biome relationship to create taxon to biome relationship)
• Sum the taxon scores for each biome
• Choose the biome with the highest score