FISHERY BULLETIN: VOL. 86, NO. 3 



the usual transect lines (along the 18 m isobath) to 

 deeper waters. 



Helicopter Observations 



During survey 4, a Hughes 500-D helicopter was 

 used to collect information on harbor porpoise 

 behavior in response to the survey ship. The heli- 

 copter flew approximately 10 km ahead of the 

 vessel, and 3 observers in the helicopter looked for 

 harbor porpoise. Once a group of harbor porpoise 

 was sighted, the helicopter hovered at 200-300 m 

 while observers made behavioral observations and 

 periodically recorded the helicopter's position using 

 an on-board Loran system. Fluorescein dye pack- 

 ages were dropped in the water to allow the heli- 

 copter to maintain its position when harbor porpoise 

 were diving. Radio communication was maintained 

 with personnel on the ship who also kept records 

 of the helicopter position using radar distances and 

 bearings based on returns from an X-band radar 

 transponder in the helicopter. The ship changed 

 course, when necessary, to ensure that it passed in 

 close proximity to the porpoise that were being 

 observed. Porpoise observers on the ship were not 

 aware of the helicopter's activities and were not told 

 of sightings made by the helicopter observers (al- 

 though they were able to see dye patches in some 

 cases). Behavioral observations from the helicopter 

 included time spent at the surface, time spent div- 

 ing, and direction of porpoise movement. 



Density Estimation 



Line transect methods were used to estimate the 

 density of harbor porpoise from sightings. The 

 assumptions of these methods are considered in 

 detail in the discussion. The usual formula for 

 estimating density (D) based on line transect 

 surveys of small cetaceans is given by 



D = 



m  n- G 

 2  L 



(1) 



where /(O) 



n 

 G 



L = 



the probability density function for 

 sightings evaluated at zero perpen- 

 dicular distance, 

 number of sightings of groups, 

 average group size calculated as the 

 total number of individuals in all 

 groups divided by the number of 

 groups (iN/n), and 

 length of the transect. 



(Holt and Powers 1982; Hammond and Laake 1983; 

 Holt in press). I did not use mean group size explicit- 

 ly in abundance estimation, and density of harbor 

 porpoise individuals, D, was estimated as 



D = /(O)  iRI2) 



where R = the number of individuals 

 length of transect (XN/L). 



(2) 



seen per 



Equation (2) is functionally equivalent to Equation 

 (1), but it simplifies variance estimation. Typically 

 when using Equation (1), variances (and possibly 

 covariances) must be estimated for/(0), G, and n. 

 Using Equation (2), variances are needed only for 

 /(O) and R, and covariance between mean group size 

 and number of groups is handled implicitly. Sight- 

 ing distributions appear to be independent of group 

 size, G, (Results section), hence no adjustments were 

 made to /(O) for group size bias. 



The parameter /(O) is, in effect, a measure of sight- 

 ing efficiency and should not vary with porpoise 

 abundance. Sighting efficiency is, however, likely 

 to change with sighting conditions, such as Beau- 

 fort sea state. Given these expectations and because 

 relatively large sample sizes are needed to estimate 

 /(O) accurately, values for/(0) were estimated for 

 each survey by pooling all sightings within defined 

 sea state categories. In order to estimate density 

 on a finer scale, estimates of R were stratified by 

 geographic region and multiplied by the pooled esti- 

 mate of /(O). 



The sighting probability density function evalu- 

 ated at zero distance, /(O), was determined 

 empirically by fitting curves to the frequency 

 distribution of sightings as a function of perpen- 

 dicular distance from the trackline (Burnham and 

 Anderson 1976). Differences in distributions of 

 perpendicular distance were tested using the 

 Kolmogorov-Smirnov 2-sample test. To avoid bias 

 due to rounding error, angle and radial distance data 

 were "smeared" (Butterworth 1982; Hammond and 

 Laake 1983). Angles were smeared by adding a 

 uniformly distributed random number between - 5° 

 and -1-5° to angle estimates. Radial distances were 

 smeared by adding a uniformly distributed random 

 number between 0.2 and -(-0.2 times the estimated 

 distance. These smearing levels were based on the 

 degree of rounding that was apparent from the data 

 (Barlow^). 



^Barlow, J. 1987. Abundance estimation for harbor porpoise 

 (Phocoena phocoena) based on ship surveys along the coasts of 

 Cahfornia, Oregon, and Washington. Adm. Rep. LJ-87-05. 



420 



