182 PAUL J. MARKWICK
Fig. 2. The relationship between the numbers of species and genera at each site used in the analysis. (a) Non-
avian tetrapod species and genera by continent. (b) Species and genera by taxonomic group.
Kingdon 1990; Conant & Collins 1991; Grenard
1991; Cogger 1992; Iverson 1992; Redford &
Eisenberg 1992; Strahan 1992). A 50 km radius
circle is drawn around each station, and an
occurrence registered where the taxon's distri-
bution intersects this circle. A radial limit of 50
km was chosen as it represents a typical decor-
relation distance for precipitation, which is the
most sensitive climate parameter to spatial
heterogeneity. This means that the faunas and
floras can be confidently assumed to have
experienced the climate assigned to them,
except in areas with rapid relief changes (such as
the Alps), where this methodology mixes high-
and low-elevation faunas. These points are
found to fall off the derived regressions but do
not significantly affect results. Ecologically, this
approach removes local, small-scale faunal and
floral heterogeneities, and thereby emulates the
spatial and temporal time-averaging in the fossil
record, with which this modern dataset can
thereby be directly compared. The 50 km radius
also approximately equates with the scale of
regional general circulation models (0.5° X 0.5°).
This approach differs from existing diversity
gradient studies (e.g. Currie 1991) in two ways.
Firstly, the use of point (station) rather than
gridded data reduces the area effects implicit in
quadrat techniques (Anderson & Marcus 1993),
and allows a direct comparison with climate (the
use of latitudinal zones is to be avoided because
it ignores longitudinal effects such as 'continen-
tality'). Secondly, the method integrates data
from more than one continent, thus reducing the
potential effects of regional biogeographical or
historical artifacts.
The diversity data have been plotted on
present-day maps using Arc View GIS for each
taxonomic group in the database, as well as for
each habitat and diet category. This provides a
qualitative indication of spatial similarities
between biogeography, taxonomic diversity and
environmental factors. The Spearman rank test
(using the Statview software; Haycock et al
1992-1993) was used to investigate correlations
between environmental variables and global
taxonomic diversity (Table 2), and regional
diversity (Tables 3-7). The relationship of
taxonomic assemblage composition to environ-
mental factors was examined with correspon-
dence analysis using the CANOCO software
(Ter Braak 1987-1992). The use of generic
rather than species-level presence-absence
information for this analysis is due to computa-
tional limitations. However, analysis of the
dataset has shown the close relationship
between the numbers of species and genera (Fig.
2), such that results derived from either taxo-
nomic level are comparable. Generic level
assignments are probably more robust for fossil
taxa.
Results
General patterns
Plots of diversity against absolute latitude are
shown in Figure 3. For comparison, a selection of
environmental variables is plotted against
absolute latitude in Figure 4 (see Table 1 for an
explanation of each variable). Total non-avian
tetrapod species diversity (amphibians + reptiles
