46 Monitoring Threatened Species and Ecological Communities
and highest likely variability in the species across its range, as well as the most
likely level. This was then converted to a single figure, assuming conceptual
similarity between steps in the 10-point scale, by averaging the mean of the lowest
and most likely score and the mean of the highest and most likely score.
Variation in detectability can bias wildlife surveys (Reid et al. 2013), so taxa
that are difficult to detect may be less likely to be monitored than those that can be
found easily. As with variability, taxa were scored against a 10-point scale based on
the survey effort required to be 90% sure of absence at a 10 ha site in usual habitat
and a time of day/year when the bird can be expected to be present in the region.
Scores ranged from ‘Exceedingly difficult to confirm presence or absence’ (score 1)
to ‘Detectable within five minutes’ (score 10). A 10 ha area is chosen as a standard
scale as being large enough for large timid species to remain despite the presence of
an observer (e.g. bustards in open grassland) but not too large to search for
nocturnal species or reclusive taxa in dense undergrowth.
Whether monitoring occurs can also be affected by accessibility with species
occurring close to population centres at sites to which access is unrestricted likely
to be easier to monitor than areas that are either physically or legally difficult to
access. The 10-point scale used to assess accessibility varied from there being legal
and/or physical access to <10% of the area occupied by the species (score 1) to >90%
(score 10).
Recovery plans have been prepared for many taxa and imply a greater degree of
attention is being given to the taxa for which they are written. Recovery plans were
assessed on a five-point scale: (0) no recovery plan, (1) no current recovery plan but
one promised, (2) approved Government conservation advice, (3) plan was in force
but now outdated, and (4) a current plan in force.
To examine the factors driving monitoring effort and quality, the inf luence of
the 10 variables (predictors) was modelled on: (1) whether or not any monitoring
occurred (a binary response in a logistic regression); and (2) overall average
monitoring quality, restricted to taxa where some monitoring does occur (i.e.
excluding zeros). IUCN Red List category, recovery plan, detectability, variability,
site accessibility, geography and taxon classification were included as categorical
variables, while range and population size were modelled as continuous variables.
Analysis took place using the ‘aod’ package (for over-dispersed counts of
proportions) in R version 3.3.1.
Next, the inf luence of each of the variables on overall monitoring quality was
modelled using multiple-linear regression. This analysis was restricted to those
species with some monitoring in place. Again, IUCN Red List category, recovery
plan, detectability, species and site accessibility were modelled as categorical
variables, while range size and population size were modelled as continuous
variables. Geography and taxon classification were modelled separately using
dummy variables to assess individual effects of each category within. Analysis took
place using the ‘lme4’ package in R version 3.3.1.