Carretta et al. Abundance and depth distribution of Phocoena phocoena off northern California 
33 
eluded within the confidence interval of a second mean, 
have been shown to be biased, because a levels do not ap- 
proach the intended value of 0.05 (Lo, 1994). Therefore, 
we used a third method proposed by Lo (1994), based on 
the confidence interval of the difference (CI rf ), between two 
population means. Through computer simulation, we gen- 
erated 5000 log-normal pseudo-abundance estimates for 
the aerial and ship surveys (IV*), using the mean estimate 
and CV from each respective survey. The difference be- 
tween ship and aerial pseudo-estimates was calculated as 
d* - N* , -N* 
u ly ship ly air 
and a 95% confidence interval of the differences (CI d ) was 
calculated from the 5000 d * values with the percentile 
method. Aerial and ship survey estimates were considered 
significantly different if the resulting CI rf did not include 
zero. Under the alternative hypothesis that abundance 
estimates were significantly different, we estimated the 
statistical power of this test by constructing one thousand 
95% CI rf intervals through simulation, using the observed 
effect size and variance from the ship and aerial surveys in 
1995. The probability of committing a type-II error /3 was 
calculated as the fraction of 1000 intervals that included 
zero (indicating no significant difference at a=0.05). An 
initial power analysis at a = 0.05 revealed that the power 
to detect a difference as large as the one observed between 
aerial and ship estimates was low (=0.13). We therefore 
generated a power curve in order to objectively reselect an 
a level for the CI rf test that would provide an approximate 
power of 0.80 (Cohen, 1988). This resulted in an a level = 
0.10; therefore, all statistical comparisons between aerial 
and ship estimates were considered statistically signif- 
icantly different if the 90% CI rf did not include zero. 
Following Forney and Barlow ( 1998), we estimated the sig- 
nificance level for this comparison by iteratively construct- 
ing a range of confidence intervals from the simulated 
data (i.e. 80%, 90%, 95%-, 96%, 97%...) and we identified the 
threshold a level (two-tailed) where the CI (/ just included 
zero. 
Results 
We surveyed a total of 594 km of transect in California 
during calm sea states (Beaufort 0-2) and detected 170 
harbor porpoise groups within the truncation distance of 
1 km (Figs. 2 and 3, Table 1). Most survey effort (377 
km) occurred in northern California, where 153 groups of 
harbor porpoise were seen, mostly in the vicinity of Cape 
Mendocino. No harbor porpoise were seen within regions 1 
and 3 in central California, but the amount of survey effort 
in these regions (60 and 91 km, respectively) was low. 
Owing to persistent coastal fog, transect coverage within 
region 3 was limited to the offshore area near the Farallon 
Islands. In region 2, only 17 porpoise groups were detected 
within the 1-km truncation distance, but only 91 km of 
trackline was surveyed. 
The perpendicular distance data were best fitted with 
the half-normal model without adjustment terms and had 
125 124 123 122 121 120W 
Figure 2 
On-effort transect effort (594 km) shown as thick gray 
lines in central and northern California during calm 
(Beaufort 0-2) sea states. 
the lowest AIC value of all competing models ix 2 goodness- 
of-fit test, P=0.84, Fig. 4). Several truncation distances and 
interval groupings were explored when fitting a detection 
function to the distance data, and all fits resulted in esti- 
mates of abundance within 9% of each other. In general, 
the lowest abundances were obtained with hazard-rate 
models. Here, we report only the results obtained with the 
half-normal model with a truncation distance of 1 km. We 
used the observed mean group size of 2.45 for region 2 and 
2.65 for northern California to estimate porpoise abun- 
dance because the regression of the log of observed group 
size versus g(x) was not significant (r 2 =0.11, P<0.89), sug- 
gesting no school size bias within the truncation distance 
of 1 km. 
Abundance of harbor porpoise in northern California 
was estimated at 5686 (log-normal 95% CI=3498-9242; 
bootstrap CV=0.29, bootstrap 95% CI=2760-8394) out to 
the 91 m isobath, based on 377 km of transect effort and 
153 porpoise sightings in calm sea states (Beaufort 0-2). 
A statistical comparison of our estimate with the aerial 
