Details - CALS Networking Lab - University of Arizona

Estimating Abundance
Estimating population size is a common goal
of biologists who are motivated by the desire
to reduce (pest species), increase (endangered
species), maintain (game species) or monitor
(indicator species) population size. Our surveys
at the park were generally focused on detecting
species rather than estimating population size.
In many cases, however, we present estimates
of “relative abundance” by species to provide
information on areas in which species might be
more or less common. Relative abundance is
an index to population size; we calculate it as
the number of individuals of a species recorded,
scaled by survey effort. If we completed multiple
surveys in comparable areas, we included a
measure of precision (usually standard error) with
the mean of those survey results.
Indices of abundance are presumed to
correlate with true population size but ecologists
do not typically attempt to account for variation
in detectability among different species or groups
of species under different circumstances. Metrics
(rather than indices) of abundance do consider
variation in detection probability, and these
include density (number of individuals per unit
area; e.g., one Arizona black rattlesnake per km 2 )
and absolute abundance (population size; e.g., 30
Arizona black rattlesnakes at the district). These
estimates are beyond the scope of our research.
While it is true that indices to abundance have
often been criticized (and with good reason, c.f.
Anderson 2001a), the abundance information that
we present in this report is used to characterize
the commonness of different species rather than
to quantify changes in abundance over long
periods of time (e.g., monitoring). As such,
relative abundance estimates are more useful
than detectability-adjusted estimates of density
for only a few species or raw count data for all
species without scaling counts by survey effort.
Sampling Design
Overview
Sampling design is the process of selecting
sample units from a population or area of interest.
4
Unbiased random samples allow inference to
the larger population from which those samples
were drawn and enable one to estimate the true
value of a parameter. The precision of these
estimates, based on sample variance, increases
with the number of samples taken; theoretically,
random samples can be taken until all possible
samples have been selected and precision is exact
– a census has been taken and the true value is
known. Non-random samples are less likely to be
representative of the entire population, because
the sample may (intentionally or not) be biased
toward a particular characteristic, perhaps one of
interest or convenience.
In our surveys we employed both
random and non-random spatial sampling
designs for all taxa. For random sites, we colocated
all taxonomic studies at the same sites
(focal points and focal-point transects; see
below for more information) because some
characteristics, especially vegetation, could be
used to explain differences in species richness
or relative abundance among transects. We also
used vegetation floristics and structure to group
transects into community types that allowed more
accurate data summaries. The location of nonrandom
study sites was entirely at the discretion
of each field crew (i.e., plants, birds, etc.) and we
made no effort to co-locate them.
Focal Points and Focal-point Transects: Random
Sampling
To account for differences in plant and animal
communities at different elevation zones (e.g.,
Whittaker and Niering 1965) at the district,
we used a stratified random design using
elevation to delineate three strata: 6,000 feet. We chose a stratified
design over a simple random design because
stratified sampling better captures the inherent
environmental variability within each stratum,
allowing for greater precision of parameter
estimates and increased sampling efficiency
(Levy and Lemeshow 1999). This design also
generates a better spatial dispersion of sampling
units. Further, we chose to delineate strata
based on elevation because it can be a good
predictor of changes in vegetation and animal