Conservation Priority Maps
In addition to
distribution maps for individual species, we have combined
predicitions for species that have high conservation priority
or ecological similarity. The methods used to generate each
composite score are given below. These maps provide tools
for identifying areas of high bird conservation priority.
The maps differ
from one another in the emphasis placed on different categories
of birds. Like the individual species maps, these have been
divided into 8525 cells of 4 km x 4 km. The color code for
each cell depends on a "score" calculated according
to one of the methods below. These scores are combinations
of the probabilities of each species occurring in a cell,
multiplied by a weighting factor. Different weighting factors
give different scores and, consequently, different map patterns.
This method tends to emphasize species that have restricted
distributions within the study area, but does not take into
account global distributions or independent assessments of
conservation status. The calculation involves four steps:
1. Calculate a ratio that is equal to the [predicted probability
of presence for species i in cell j] divided by the [median
predicted probability of presence for species i across all
This ratio indicates the degree to which a cell is likely
to contain a given species compared to the overall likelihood
of species predicted presence in the entire map. The median
is used as the denominator instead of the mean because the
distributions of predicted probabilities were skewed for most
From a conceptual standpoint, we can think that cells with
ratios =1.0 to represent places that are not of great conservation
priority for species. By omitting these ratio values from
the subsequent computation of the score, cells having greater
predicted probability than the median for a species represent
a threshold of usefulness for the identification of priority
areas. Species that are common tend to have median predicted
probabilities that are high, and the ratio for these species
tends to be low. Use of this ratio therefore tends to give
more weight to rarer species in cells where these species
are predicted to occur.
2. Compute the natural logarithm of the ratio.
This is necessary because there is a large range in the values
of the ratio. A few species exhibit very large values of the
ratio for some cells because of low median predicted probabilities.
Palm Warbler stands out as the species with the maximum ratio
(38), while most species had maximum ratios below 10. If we
didn't dampen the influence of this species, the resulting
map of conservation scores would have looked a lot like the
Palm Warbler map. The logarithm of the ratio was not computed
for ratios =1.0. Instead, these values were manually set to
3. Calculate the mean of the natural logarithm of the ratios
across all species for each cell.
For this step we include species with ratios =1.0 so that
the mean log_ratio for each map cell is based upon the same
number of species in the denominator. If logarithm had been
calculated for ratios <1, the log_ratio would have been
negative in several cases. It didn't seem that negative values
(low ratios) should be able to subtract from the conservation
value of a cell.
4. Compute the conservation priority score by back-transforming
the mean of the log_ratios for cells.
Back-transformation has the effect of putting the scores
in the original scale of the ratio. For example, if the conservation
priority score is 1.5 for a cell, then species will tend to
be predicted at a probability 50% greater than their median
predicted probability, on average. This is the score that
has been mapped.
We used 43 species (of 82 evaluated overall) to calculate
the log ratio. These species yielded predictive models that
discriminated presence or absence of individuals at the p
< 0.10 level.
Black-throated Green Warbler
This score is the number of species that have a
in Flight (PIF) conservation priority score >18 in
Bird Conservation Region 12 and a predicted probability of
occurrence in the cell of at least 0.20 (20%). Grassland or
open country species and species that had poor predictive
models were excluded, leaving 16 species for this analysis:
Black-throated Green Warbler
This score uses the same 43 species as described
above for the Log Ratio scores. The predicted probability
for each species in each 4 km x 4 km cell is multiplied by
the corresponding PIF conservation score. The mean of these
weighted scores gives a composite score for the entire cell.
Species with low PIF scores or low predicted probabilities
will tend to have low influence on the PIF weighted score
for any given cell.