Final Technical Report: - Southwest Fisheries Science Center - NOAA

etween cetacean presence and oceanographic variables for dolphins in the ETP. CCA allowed
the geographic mapping of dolphin habitats for the first time. Forney (2000) used GAMs to
determine the relationship of cetacean encounter rates with oceanographic and geographic
variables. However, none of these approaches allow the geographically explicit estimation of
cetacean density. Ferguson and Barlow (2001) used a stratification approach to finer scale
density estimation, but found that sample sizes still required that they use relatively large areas.
Hedley et al. (1999), Hedley (2000), and Marques (2001) developed the first spatially explicit
methods for modeling density from cetacean line-transect data. The GAM-based framework is
now clearly established as a method for modeling cetacean density as a function of fixed
geographic and stochastic habitat variables.
Although analytical methods are clearly necessary for geographically explicit modeling
of cetacean density, another requirement for the development of accurate models is a large
amount of survey data collected using rigorous line-transect methods. Ever since line-transect
methods were first established (Burnham et al. 1980), the SWFSC has been a leader in the
application and improvement of line-transect methods to estimate cetacean abundance (Holt and
Powers 1982, Holt 1987, Barlow 1988, Barlow et al. 1988, Holt and Sexton 1989, Gerrodette
and Perrin 1991, Wade and Gerrodette 1993, Forney and Barlow 1993, Barlow 1994, Barlow
1995, Forney et al. 1995, Barlow et al. 1997, Forney and Barlow 1998, Carretta et al. 1998,
Barlow 1999, Ferguson and Barlow 2001, Barlow et al. 2001). Here we base our models of
cetacean densities on SWFSC ship line-transect data collected from 1986 to 2006. These
surveys include over 17,000 sightings of cetacean groups on over 400,000 km of transect line.
In addition to cetacean line-transect data, our model development is dependent on having
measures of the oceanographic conditions that define cetacean habitat. Since 1986, the SWFSC
has consistently gathered basic oceanographic data on virtually all of their cetacean line-transect
surveys (Reilly and Fiedler 1994) and has been increasingly gathering additional data on midtrophic
levels, including plankton and neuston net tows and acoustic backscatter measurements
(Fiedler et al. 1998). Although we also build models of cetacean density with remotely-sensed
oceanographic data, the concurrent collection of line-transect data and cetacean habitat data
ensures a closer correspondence between the real-time distribution of cetaceans and their
measured habitat variables and has allowed us to sample more aspects of their habitat than is
possible with remotely-sensed data.
Most of our shipboard line-transect data were collected during summer and fall, and these
data cannot be used directly to build models for other seasons. However, SWFSC has conducted
aerial surveys at other times of the year in portions of the California Current. This region is
known to have pronounced seasonal variation in the distribution and abundance of marine
mammals (Forney and Barlow 1998). The aerial survey data contain too few sightings to build
predictive environmental models, but we use these data to evaluate whether models constructed
for summer/fall using the extensive shipboard sighting data are applicable to other seasons. This
3