Relations between the distribution of herbaceous and woody species from the
flora of Israel and the variation (25 -900 mm) in mean annual rainfall are
described using a grid-based approach integrating multivariate techniques
(cluster analysis and DCA-ordination) and GIS. The analysis was based on an
extensive data base of vascular plant records in 10 km x 10 km grid cells
representing the climatic zones of Israel. Cluster analysis revealed three
geographically distinct clusters of grid cells which could be arranged alon
g the main rainfall gradient in the study area. The main subdivision is bet
ween cells from the Mediterranean region and cells from the more arid regio
n. The second subdivision separated cells from the arid region into two dis
tinct groups which differed significantly in rainfall. Clustering of herbac
eous and woody species was similar, but boundaries separating adjacent clus
ters based on woody species were consistently linked to more rainy areas. C
ell scores on the first DCA-axis were significantly correlated with mean an
nual rainfall, but this relationship was not linear. Log-transformation of
the rainfall data resulted in a higher correlation between the DCA-scores a
nd rainfall, indicating that variation in rainfall had a stronger effect on
species composition in relatively dry than in more mesic regions. Slopes o
f linear regression models relating cell scores on the first DCA-axis to me
an annual rainfall, as well as the corresponding R-2-values, were highest i
n the desert, intermediate in the dry Mediterranean, and lowest in the mesi
c Mediterranean territory. These results indicate that both the per-unit ef
fect of rainfall on plant species composition and its relative importance a
s a determinant of compositional variation decreased from relatively dry to
more mesic regions.
The length of the first axis, as well as the slopes of regression models re
lating cell scores on the first DCA-axis to mean annual rainfall were consi
stently higher in the case of woody species which are characterized by a hi
gher rate of compositional variation. Due to lack of comparable data, it is
difficult to evaluate whether our results can be generalized to other regi
ons. However, an analysis of data obtained from different parts of the worl
d points to a considerable variability in the degree to which rainfall affe
cts patterns of floristic variation. Our findings that compositional respon
ses to rainfall may vary between different groups of species and may depend
on the part of the rainfall range which is being considered may partially
explain this inconsistency.