606 



Fishery Bulletin 101 (3) 



stranding records of mesoplodont whales from U.S. Atlantic 

 shores, sightings of Mesoplodon were probably True's (M. 

 mirus), Gervais's (M. europaeus) or Blainville's (M. densiro- 

 stris) beaked whales (Mead, 1989). In some cases cetaceans 

 could only be identified as large whales (>7 m long), small 

 whales (nondolphin, <7 m), dolphins, or odontocetes. 



Analytical techniques 



For each species or taxonomic category, abundance esti- 

 mates (N) were made with line-transect methods by using 

 the software program DISTANCE (Colorado Coop. Fish 

 and Wildlife Research Unit, Colorado State Univ., Fort 

 Collins, CO) (Buckland et al., 1993) with the equation 



A' = 



A n 5/(0) 

 2Lg{0) 



where A = size of the study area; 



n = number of on-efTort group sightings; 

 S = mean group-size estimate; 

 /!0) = sighting probability density function at per- 

 pendicular distance zero; 

 L = total length of transect line; and 

 giO) = probability of seeing a group on the transect 

 line. 



The log-normal 95% confidence interval was computed 

 for each abundance estimate because it was a product of 

 estimates and tends to have a skewed distribution. The 

 variance of N was estimated as 



\-dT(N) = N- 



var(») var(5) var[/(0)] var[i,'(0)] 

 »r "^ S' f(Of f(Of 



and the coefficient of variation (CV) was estimated as 



CViN): 



Vvar(/V) 

 N 



The sampling unit was the length of the transect completed 

 on-efTort each day when the Beaufort sea state was <5. The 



formula used to estimate each component of the variance is 

 given in Buckland et al. ( 1993). Var(«) was length-weighted 

 and based on the variation in the number of on-effort group 

 sightings between sampling units that ranged in length 

 from 39 to 229 km/day. 



Estimation of M) 



The perpendicular distance, >>, was estimated by using bear- 

 ing and reticle measurements. The reticle readings were 

 converted to radial sighting distances (/?) by the method 

 of Lerczakand Hobbs (1998), and the formula v = i?sin(6), 

 where b = angle between the sighting and the transect 

 line. Estimates of /iO) were made by using a hazard-rate, 

 uniform, or half-normal model with exact perpendicular 

 sighting distances. For each species group, outlying values 

 ofy were truncated to improve the fit of the model (Table 2). 

 Model selection was determined by using Akaike's informa- 

 tion criterion (AIC; Buckland et al., 1993). 



The number of groups sighted of most species was insuf- 

 ficient to obtain an estimate of /lO). Therefore, sightings of 

 species with similar sighting characteristics (i.e. body size, 

 group-size, surface behavior, blow visibility) were pooled to 

 estimate /(O) for five categories (Table 1). The abundance 

 for each species was estimated by using the pooled /1 0) and 

 var|/10)l for its category. The varl/!0)l was assumed to be 

 zero for the strip-transect estimates explained below. If 

 the individual detection functions of all species within a 

 category are indeed very similar, by pooling, the variance, 

 CV, and confidence interval of each abundance estimate 

 was probably underestimated because the variance of /lO) 

 was based on an artificially high sample size. On the other 

 hand, if the true detection functions of the species within 

 a category are highly variable, the variance of /lO) for an 

 individual species may be overestimated. 



During the study, effort was sometimes maintained while 

 in transit to and from ports or along the border of the study 

 area, but it usually occurred in a small range of water 

 depths (e.g. parallel to shore) and was excluded because 

 it could have biased abundance estimates. However, due 

 to the small number of sightings for the survey, y from the 

 "transit" sightings were pooled with the on-effort sightings 

 for estimates of /lO) (Table 2). 



