Fróðskaparrit - 01.01.2007, Page 124

122
THE VEGETATION OF GRASS ROOFS IN THE FAROE ISLANDS AND
THE SURROUNDING GRASSLAND VEGETATION -A STUDY FROM SANDOY
appearing with low scores on axis 1, are
dominated by the grass species for which
this area is managed. They may also in-
clude taxa derived from the bryophyte
layer that is most commonly dominated by
Plagiomnium undulatum amongst other
mosses and liverworts.
Correlations of DCA axes with slope,
aspect and the presence of grazing
Correlations (Pearson product moment) of
the DCA axes with slope, aspect and graz-
ing reveal no signifícant result (Table 1)
when the turf-roof data alone are analysed.
A relationship becomes more apparent
(axis 1 only) when the data from the infield
and outfíeld areas are included, for the fac-
tors of slope (infield-outfíeld samples
p=12.3 o=9.0; roof samples p=39.9 o=2.4)
and grazing, but not aspect. However,
these correlations remain statistically in-
signifícant at the 95% confídence interval.
Table 1. R-values of correlations between the
sample DCA scores and the variables aspect, grazing
and slope for the roof data alone, and in combination
with data derived from the infield and outfíeld. All
values insignificant at the 95% confidence.
Variablc Axis 1 Axis 2
Roofs only Aspect 0.013 0.106
Grazing 0.195 0.086
Slope 0.192 0.169
Roofs and Aspect 0.162 0.004
Jields Grazing 0.672 0.256
Slope 0.637 0.238
Discussion
The signifícant and marked difference in
the diversity of the ground and roof-turf
data suggests that even if relatively homo-
geneous turfs dominated by favoured spe-
cies were specially chosen, the diversity of
the original turfs would have been higher
prior to their incorporation into the roof.
The ratio of plant species richness between
the ground and roof vegetation is around
3:1, and the most species-poor ground
quadrats contain more species than the
most species-rich turf-roof quadrats. This
observation refutes the third hypothesis
(that the species-poor nature of roof turfs is
a consequence of selection); this negative
inference is supported by the limited litera-
ture which relates that the propagules of
desirable species would be incorporated
into the roofs but not that turfs were specif-
ically cultivated for dominance by those
species. A certain degree of turf selection
remains a reasonable assumption, but the
diversity of the ground vegetation, as
measured within lm2 quadrats, is such that
it is unlikely that suffíciently large areas
would have been available for the easy se-
lection of homogeneous turfs for roofíng
en masse.
The axis length of DCA axis 1 in Figure
5, indicates further the small amount of
variation within the vegetation of the grass
roofs as there is not a complete turnover of
taxa. This small amount of variation is
demonstrated yet further in Figure 6, in
comparison with the samples from the
ground vegetation, where the turf-roof
samples form a relatively small, closely-
related grouping nested amongst the in-