Geoscience Reference
In-Depth Information
However, the real advance in terms of GC is when one links the attribute data represented in tree
and matrix data structures to the taxon data. This allows one to consider genetically and phyloge-
netically weighted indices (Bickford et al., 2004; Rosauer et al., 2009b).
The most commonly applied form of phylogenetically weighted analysis is phylogenetic diversity
(PD; Equation 6.4; Faith, 1994). PD is the phylogenetic analogue of SR, simply comprising the sum
of the branch lengths spanned by the taxa occurring within a neighbourhood, measured from termi-
nal branches to root node. In many cases, the branch lengths correspond to the number of unique
features represented by that branch, so PD is operating at a finer level of detail than the taxon while
also correcting for features shared between related taxa. It is a simple matter to extend the method
into a measure of phylogenetic endemism (PE; Equation 6.5), something that is derived from a com-
bination of WE and PD (Rosauer et al., 2009b). As with species endemism metrics, one calculates
the relative range restriction, but in this case, it is for the nodes in the tree, with the weight calcu-
lated by multiplying the branch length of each node by the fraction of its range represented in the
neighbourhood. One can then either explore the relative weights of each node or sum the values to
obtain the aggregate PE score. The interpretation of this result depends on the nature of the tree. For a
chronogram, for example, one will have a metric of the spatial concentration of evolutionary history:
∑
PD
i
=
L
(6.4)
c
cC
∈
i
r
R
∑
c
PE
i
=
L
(6.5)
c
c
cC
∈
i
where
C
i
is the set of branches in the minimum spanning path joining the taxa in neighbourhood
i
to
the root of the tree
c
is a branch (a single segment between two nodes) in the spanning path
C
i
L
c
is the length of branch
c
R
c
is the geographic range of branch
c
(the union of the ranges of taxa under branch
c
)
r
c
is the local range of branch
c
restricted to neighbourhood
i
A comparison of the aforementioned metrics is given in Figure 6.2, using species distribution data
obtained from the Red List of Threatened Species version 2010.2 from the International Union
for Conservation of Nature (IUCN) (http://www.iucnredlist.org), aggregated to 100 km cells in an
Albers equal area coordinate system and a phylogenetic tree extracted from Bininda-Emonds et al.
(2007) with branch lengths representing millions of years of evolution. The SR and PD patterns are
comparatively similar, as are the WE and PE scores. However, where the WE score is the sum of the
weighted species, the PE score tells us how many millions of years of unique evolutionary history
is found in each of the cells. The cells with the highest CWE scores represent, on average, 25% or
more of the ranges of species that are found within them.
One GC application area that is worthy of further exploration is the use of more complex spatial
neighbourhoods, for example, using process-based knowledge (Laffan, 2002). Alternately one can
use an agglomerative cluster analysis of the data to define groupings based on some taxonomic,
genetic or phylogenetic index, a process enabled within Biodiverse (Laffan et al., 2010). Each of
these clusters, at each level of the tree, contains some set of geographic units, and this can be treated
as a neighbourhood for which other indices can be derived. Clearly some care needs to be taken in
terms of analytical circularity. Many indices are closely related and are often slight modifications of
each other. Calculation of spatial indices from the same set of taxa used to determine the clustering
will not be informative in many cases. However, indices of their traits might well be, for example,
the fruit size of endemic taxa.
