Final Technical Report: - Southwest Fisheries Science Center - NOAA
Final Technical Report: - Southwest Fisheries Science Center - NOAA
Final Technical Report: - Southwest Fisheries Science Center - NOAA
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Encounter Rate and Group Size Models<br />
For each species or species group, we built separate models of cetacean encounter rate<br />
(number of sightings per unit of effort on the transect) and group size (number of individuals per<br />
sighting). In preparation for building the models, the cetacean sighting data and environmental<br />
data were summarized into segments of on-effort transect. Encounter rate models were built<br />
using all transect segments, regardless of whether they contained sightings. Group size models<br />
were built on only the subset of segments that contained sightings.<br />
Cetacean sighting data are essentially count data with relatively more zeroes than<br />
expected from a standard Poisson distribution. Therefore, we modeled encounter rate as a<br />
quasipoisson distribution with variance proportional to the mean and a logarithmic link function.<br />
The natural logarithm of segment length was included as an offset term to standardize each<br />
sample for effort.<br />
Cetacean group sizes can be highly variable, spanning up to three orders of magnitude.<br />
Estimating the mean group size associated with each line segment involved three steps. First, we<br />
computed an estimate of group size for each observer for each sighting based on the observer’s<br />
best, high, and low estimates of group size. Second, we computed the arithmetic mean of all<br />
observer’s group size estimates for each sighting. <strong>Final</strong>ly, we computed the arithmetic mean<br />
group size of all sightings in each line segment. This three-step process resulted in non-integer<br />
group size estimates. Given the wide range of cetacean group sizes and the fact that the group<br />
size estimates are continuous data, we constructed lognormal GAMs for group size, using the<br />
natural logarithm of group size as the response variable and an identity link function. It was<br />
necessary to apply a bias-correction factor to the group size predictions from the GAMs because<br />
the models were built in log space and then the results were transformed back to arithmetic<br />
space, converting the group size estimate to a geometric mean in the process (Finney 1941,<br />
Smith 1993). The ratio estimator was used to correct for this back-transformation bias (Smith<br />
1993).<br />
Density Computations<br />
To estimate cetacean density, the encounter rate and group size model results were<br />
incorporated into the standard line-transect equation:<br />
where,<br />
n/L = encounter rate (number of sightings per unit length of transect),<br />
S = expected (or mean) group size,<br />
18