510 



Fishery Bulletin 105(4) 



and fall with the same line-transect survey methods 

 from two National Oceanographic and Atmospheric 

 Administration (NOAA) research vessels: the 53-m RV 

 McArthur and the 52 m RV David Starr Jordan. A third 

 ship, the 62-m RV McArthur II, was also used for a 

 very short time in 2005. Transect lines followed a grid 

 that was established before each survey to uniformly 

 cover waters between the coast and approximately 556 

 km (300 nmi) offshore. Surveys were designed with a 

 uniform grid of transect lines anchored by a randomly 

 chosen start point. Ships traveled at 16.7-18.5 km/h 

 (9-10 kt) through the water. The 1991 and 1993 surveys 

 only covered waters off California, but the subsequent 

 surveys also included waters off Oregon and Washington 

 (Fig. 1). 



Experienced field biologists (henceforth referred to 

 as "observers") searched for cetaceans from the flying 

 bridge deck of the ships (observation height -10.5 m 

 for the two primary vessels, 15.2 m for the RV McAr- 

 thur II). Typically, six observers rotated among three 

 observation stations (left station, where 25x binoculars 

 were used; forward station where the data recorder was 

 positioned; and right station, where 25 x binoculars were 

 used). Each observer and recorder watched for 2 hours 

 and then rested for 2 hours. The recorder searched 

 with unaided eyes (and occasionally 7x binoculars) and 

 entered effort and sighting data using a data entry 

 program on a laptop computer. The observers were se- 

 lected on the basis of previous experience searching 

 for and identifying marine mammals at sea; at least 

 four observers on each ship had previous line-transect 

 experience with cetaceans and at least two were ex- 

 perts in marine mammal identification at sea. Before 

 each survey, observers were given a refresher course in 

 marine mammal identification and group size estima- 

 tion. Group size and the percentage of each species in a 

 group were estimated and recorded independently and 

 confidentially by each on-duty observer. Generally, after 

 a group of cetaceans was seen, observers took as much 

 time as necessary to estimate group size and species 

 composition. Starting in 1996, at least one hour was 

 allocated to group size estimation for sperm whales to 

 provide reasonable confidence that all members of the 

 group surfaced at least once. Species determinations 

 were recorded only if observers were certain of their 

 species identification; otherwise, animals were identified 

 to the lowest taxonomic level or general category (e.g., 

 large whale or baleen whale) that an observer could de- 

 termine with certainty. Observers were also encouraged 

 to record separately the most probable species if the ac- 

 tual species could not be determined with certainty. In 

 this article, we used probable species identifications if 

 certain identifications were missing, rather than prorat- 

 ing the unidentified sightings into species categories, as 

 done in other studies (Gerrodette and Forcada, 2005). 

 If probable species identifications were not available, 

 species were classified as unidentified delphinoids, small 

 whales, beaked whales, rorqual whales, or large whales. 

 Common and scientific names for all species are given 

 in Table 1. 



Most surveys were conducted in closing mode during 

 which the ship diverted from the trackline as necessary 

 to allow closer estimation of group size and species 

 composition. The ship was not diverted if observers felt 

 that group size and species could be determined from 

 the transect line, as was frequently the case of nearby 

 sightings of Ball's porpoise or large baleen whales. To 

 investigate potential biases associated with the use 

 of closing mode surveys, every third day of effort in 

 1996 was conducted in passing mode during which the 

 ship did not divert from the trackline except for spe- 

 cies of particular interest (sperm whales, short-finned 

 pilot whales, and Baird's beaked whales). No consis- 

 tent biases were found between the two survey modes; 

 however, group size estimation and species determina- 

 tion suffered in passing mode, and therefore the latter 

 (passing mode) was not undertaken during subsequent 

 surveys. 



Frequently, a fourth observer searched for cetacean 

 groups that were missed by the primary team of three 

 observers. Sightings made by this fourth observer were 

 recorded after the group had passed abeam and had 

 been clearly missed by the primary team. The data 

 from the fourth observer were considered conditionally 

 independent of the primary team (conditioned on the 

 animals not being seen by the primary team) and were 

 used to estimate the proportion of sightings missed by 

 the primary team. 



Calibration of group size 



Individual observers may tend to over- or under-esti- 

 mate group sizes, and their estimates can be improved 

 by calibration based on a subset of groups with known 

 size (Gerrodette and Forcada, 2005) or based on compari- 

 son to data from an unbiased observer (Barlow, 1995). 

 Calibration factors were used to correct estimates made 

 by observers who were previously calibrated by using 

 aerial photographic estimates of group size taken from 

 a helicopter on dolphin surveys in the eastern tropical 

 Pacific (Gerrodette and Forcada, 2005). These calibra- 

 tions were not applied to groups whose size was outside 

 the range of sizes used in the calibration study. A direct 

 helicopter calibration could not be used on these west 

 coast surveys because the weather was too rough and 

 the water is too turbid. Therefore, we used an indirect 

 calibration method (Barlow, 1995) to calibrate these 

 remaining observers in relation to the directly cali- 

 brated observers. The indirect calibration coefficient, fig, 

 for a given observer was estimated by comparison to 

 calibrated estimates of directly calibrated observers by 

 using log-transformed, least-squares regression through 

 the origin: 



\nS'=Pg-\nS, 



(1) 



where S* = the observer's best estimate of group size; 

 and 

 S = the mean of calibrated, bias-corrected esti- 

 mates for all other calibrated observers. 



