Raum-Suryan and Harvey: Distribution and abundance of and habitat use by Phocoena phocoena 
and slopes in the study area, eight random points 
were plotted within a 2-km strip along the length of 
each 8-km transect {n = 73 transects, 584 points) and 
depth and slope were determined for each point. The 
number of random points chosen was determined by 
plotting precision (standard deviation/mean) against 
sample size until there was little variability in this 
measure (i.e. a plateau and subsequent leveling of 
the curve). 
Density and abundance estimates were calculated 
by using the line transect method as described by 
Burnham et al. (1980) and the computer program 
DISTANCE (Laake et al., 1993). Each transect was 
considered a replicate. Density and variance esti- 
mates of harbor porpoise sightings (rc=250) were cal- 
culated by replicate for each section (n = 12 to 15 
transects) and by replicate for all sections combined 
(n = 70 transects). Transects with Beaufort sea state 
of 2 (n= 3) were deleted from analyses because sight- 
ing rates of harbor porpoise in Beaufort 2 are less 
than Beaufort 0 or 1 (Barlow, 1988). Density was 
calculated as 
D = 
n x f( 0) x s 
2 L 
where n 
/TO) 
s 
L 
number of individual harbor porpoise 
sightings; 
the probability density function of dis- 
tances from the trackline evaluated at 
zero distance; 
average group size of harbor porpoise 
sightings, and 
total length of the trackline. 
Abundance was calculated as density multiplied by 
area of each section (A-E) and all sections (237 km 2 ). 
The parameter /TO) is essentially a measure of sight- 
ing efficiency and should not vary with porpoise abun- 
dance as long as sighting conditions (e.g. Beaufort 
sea state, visibility) remain the same. Because we 
surveyed only during optimal sighting conditions 
(Beaufort <1, no rain or fog) within all sections and 
because relatively large sample sizes are required to 
estimate /TO) accurately, values of /TO) for each sec- 
tion were estimated by pooling all sightings in all 
sections. Effective strip width is defined as 1//T0), 
which equals one-half the transect width, such that 
as many objects are detected outside the strip as re- 
main undetected within it (Buckland et al., 1993). 
Because group size was independent of distance from 
the trackline (determined through size-bias regres- 
sion analysis with DISTANCE software), average 
group size was used to calculate density. Average 
group size was estimated by section and for all 
sightings combined. 
Uniform, half-normal (hermite), hazard rate, and 
negative exponential models were compared with the 
frequency distributions of perpendicular sighting 
distance of harbor porpoise to trackline with DIS- 
TANCE. Several groupings and truncation points 
were investigated to achieve the best model fit. 
Buckland et al. (1993) recommend truncating 5 to 
10% of objects detected at the greatest distances from 
the trackline. The half-normal (hermite) model, 
grouped into 50 m intervals and truncated at 750 m 
(deleting 5% of sightings), was chosen on the basis 
of lowest Akaike Information Criterion (AIC; 
Buckland et al., 1993) score for all sections combined. 
The probability of detection at zero perpendicular 
distance, g( 0), was assumed to be one (all harbor 
porpoise on the trackline were assumed to be seen) 
because we were unable to estimate perception bias 
(bias resulting from animals available to be seen but 
that were not; Marsh and Sinclair, 1989). We did not 
have an independent observer to watch the trackline 
for porpoise that were missed by our two observers, 
therefore, a correction was not applied tog(0). It is 
likely g(0) was less than one but it is probably high 
(slow boat speed and excellent sighting conditions). 
Because g(0) was constant over the survey time pe- 
riod, the habitat correlations are valid; however, the 
abundance estimate is underestimated by an un- 
known amount. 
Seafloor depth and slope available in the study area 
in relation to areas of harbor porpoise occurrence 
were compared by using chi-square goodness-of-fit 
analyses. To test whether the frequency of occurrence 
of harbor porpoise was independent of frequency of 
tidal currents and surface temperature, we also used 
chi-square goodness-of-fit analyses. More surveys 
were conducted during flood tide (n=52) than during 
ebb tide (n=17); therefore, the number of harbor por- 
poise observed per minute during flood or ebb tide 
was used to standardize the data. A Mann-Whitney 
U, nonparametric two-sample test was conducted to 
examine differences in number of harbor porpoise 
observed per minute for each transect (n- 73) during 
flood and ebb tides. 
Power analyses (Cohen, 1988) were conducted on 
nonsignificant categories of chi-square goodness-of- 
fit analyses. Randomization statistics with the pro- 
gram Resampling Stats (Resampling Stats, 1995) 
were performed to assess the probability of detect- 
ing a difference between flood and ebb tides when 
the difference was determined to be nonsignificant. 
Additionally, power analyses were used to estimate 
the probability of detecting trends in abundance over 
time (Gerrodette, 1987). 
