190 PAUL J. MARKWICK
Table 6. Spearman rank test rho values for environmental variables and mammal species diversity by region
n
Absolute latitude
Elevation
MAT
MART
CMM
WMM
Radiation
Cumulative T 0
Cumulative T 5
Annual precipitation
P range
Months T 10 P 40
PET
Mean annual NDVI
NDVI 1SD
South
America
72
-0.831
NS
0.715
-0.670
0.786
0.513
NS
0.714
0.715
0.565
0.712
0.713
0.633
0.636
-0.765
North
America
165
-0.658
0.511
0.615
-0.400
0.583
0.529
NS
0.628
0.627
NS
0.323
0.337
0.526
0.442
-0.511
Europe
204
-0.552
0.517
0.323
-0.321
NS
0.513
NS
0.347
0.444
NS
NS
0.411
0.385
0.616
-0.478
Arabia
18
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
Southern
Africa
21
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
Australia
40
NS
NS
NS
NS
NS
NS
NS
NS
NS
0.728
NS
0.643
NS
NS
NS
(p < 0.0001: NS, not significant, p > 0.0001)
America are treated separately, where axis 1
approximates axis 2 in this analysis). The
greatest differences in the faunas are between
southern Europe and southern North America,
consistent with the historical faunal inter-
changes being via the North Atlantic land
bridges that linked the two continents during the
Early Cenozoic. The second axis (Fig. 14)
accounts for 28% of the variance and correlates
most strongly with temperature and therefore
incident energy (Table 8). Axis 3 (13.4% of
variance) is dominated by the precipitation
pattern.
Application to the fossil record
Ostrom (1970) postulated the use of diversity
gradients as a palaeoclimate tool. In order to
examine the viability of this suggestion, an
experiment is made in which the linear relation-
ship between the proportion of the fauna rep-
resented by ectotherms and MAT (Fig. 11) is
used to reconstruct the palaeotemperature of
the Middle Eocene Messel Shale, Germany,
based on its fossil non-avian tetrapod fauna.
Messel, as a lagerstatten, is used in order to
minimize the effects of compositional biases
(taphonomy, taxonomy and collection). Species,
genus and family diversities for this fauna are
derived from the faunal lists given in Schaal &
Ziegler (1992). Additional linear relationships
between the proportion of ectotherms and
coldest month mean temperature (CMM) and
latitude have also been made. Errors from these
calculations represent one standard deviation of
the residuals of MAT, CMM and absolute
latitude (Fig. 12, Table 9), reflecting potential
effects of history, biogeography and taxonomy
on present-day data.
The results for this analysis of the Messel
Shale fauna are given in Table 9 and compared
with estimates from other climate proxies in
Table 10. The values for MAT and CMM cal-
culated using the proportion of reptiles or
ectotherms agree with the climate interpre-
tations based on the presence of fossil crocodil-
ians (Markwick 1998a) and palms (Markwick
1996), although they are somewhat lower than
the MAT estimated by Wilde (1989) based on
plant physiognomy. However, estimates of
absolute latitude are generally lower than the
44° palaeolatitude calculated for this site using
palaeomagnetic data (D.B. Rowley pers comm.
1995). Regressions based on the total number of
non-avian tetrapods give consistently lower
MATs and CMMs than those using the propor-
tion of reptiles or ectotherms; they also give
higher absolute latitudes (Table 9). In both cases
the results are similar to what would be expected
at the palaeolatitude of Messel in the Eocene
(44°N) if the Earth's latitudinal thermal gradient
was the same as today's, which would imply that
the absolute number of species of non-avian
tetrapods is indeed a function of latitude (and
thus of incident solar energy flux). However,
such an interpretation may be premature, since
it is based only on one fossil fauna. Variations in
predicted temperatures between taxonomic
