FISHERY BULLETIN: VOL. 85. NO. 3 



to record position of sun relative to the platform 

 and for criteria used to define sun categories for 

 aerial data). Criteria used for ship data were 

 based upon observations recorded during a subse- 

 quent ship survey (Hohn^). Hohn found poor sun 

 conditions on the trackline only when horizontal 

 sun position was 12 and vertical position was 1, 2, 

 or 3 or when clouds were accompanied by fog or 

 rain. All other effort was defined as occurring 

 during good conditions. 



In order to apply the Fourier series (FS) model 

 to aerial and ship data, I structured the data by 

 1 ) selecting appropriate interval widths for 

 grouping the perpendicular sighting distribu- 

 tions (data cutpointsi, 2) choosing a maximum ob- 

 servation distance perpendicular to the trackline 

 (truncation point), 3) developing criteria to select 

 the appropriate number of terms for the FS 

 model, and 4) choosing the type of transformation 

 to use in compensating for measurement error in 

 the shipboard data. 



Based on a subset of the ship data (Holt^*^), I 

 used an interval width of 0.37 km (0.2 nmi) and 

 truncated the perpendicular distance distribu- 

 tions at 3.7 km (2.0 nmi). Since perpendicular 

 distance distributions for the ship data, and also 

 to a lesser extent for aerial data, have very promi- 

 nent modes or "spikes" at the origin, existing 

 criteria to select the appropriate number of terms 

 in the FS model were unsatisfactory. Therefore, I 

 selected the model which provided the best visual 

 fit to the distributions near the origin (Holt fn. 

 10). This technique was easily applied and was 

 consistent among data sets. For use of the tech- 

 nique I assumed that the sizes of the spikes near 

 the origins of the perpendicular distance distribu- 

 tions were indicative of relative density among 

 the data sets. To minimize the effects of recording 

 errors, the data were smoothed using the tech- 

 nique "smearing" (Butterworth 1982; Hammond 

 1984). 



Based on previous investigations of aerial data 

 (Holt and Powers 1982), I selected a truncation 

 point of 1.94 km (1.05 nmi) and an interval width 

 of 0.19 km (0.1 nmi) for the aerial data. I used the 

 same technique as used for ship data to select the 

 appropriate number of terms in the FS models; 



^A. Holin, Southwe.st Fi.sheries Center La Jolla Laboratory, 

 National Marine Fisheries Service, NOAA, P.O. Box 271, La 

 Jolla, CA 92038, pers. commun. January 1985. 



'"Holt, R. 1984. Estimation of density of dolphin schools in 

 the eastern tropical Pacific Ocean u.sing line transect meth- 

 ods. Southwest Fish. Cent. Adm. Rep. No. L.J-84-32. 

 72 p. National Marine Fisheries Service, NOAA, P.O. Box 

 271, La Jolla, CA 92038. 



however, the aerial data were not smoothed be- 

 cause there was no evidence that the data con- 

 tained estimation errors as did the ship data. 



An estimate of density in the total area (Z)^.) 

 was calculated by combining the aerial inshore 

 (Z),) and ship offshore (Dq) density estimates 

 weighted by the relative sizes of the inshore (A, ) 

 and offshore (Aq) areas as 



D. 



D,A, + DqAq 



The estimate of variance of D^ is 



A,^Vdr(D, ) + Ao-VdriDo) 



Vdr(D,) 



(A, + Aq)" 



RESULTS 



Factors Affecting Density 

 Estimates 



Aerial Data 



Density estimates for the aerial data in the in- 

 shore area during calm seas or with minimal sun 

 glare were more than twice the estimates for data 

 taken during rough seas or poor sun conditions 

 (Table 1). Differences in estimators were even 

 greater for sea state and sun glare interaction 

 effects. These differences may have occurred be- 

 cause observers failed to detect trackline schools 

 during poor conditions or because sea state condi- 

 tions were spatially confounded with distance 

 from shore. Therefore, these differences may be 

 reflecting a decreasing onshore-to-offshore den- 

 sity gradient. This was investigated by partition- 

 ing the inshore aerial data into "coastal" and 

 "offshore" bands for each Beaufort sea state 

 (Fig. 3) and sun glare condition (Fig. 4). Sufficient 

 data were not available in each band to stratify 

 detection rates by eaqh sun and sea state interac- 

 tion category. 



Sea conditions dufing the aerial surveys were 

 rougher offshore than nearshore. More searching 

 was done in the coastal band during low Beaufort 

 states, whereas tnore searching was done in the 

 offshore band at higher Beaufort states (Fig. 3). 

 The rates of detecting dolphin schools were 

 higher at each corresponding Beaufort state in 

 the coastal band than in the offshore band 

 (Fig. 5). The rates of detecting trackline schools 

 were generally higher in the coastal band; how- 

 ever, these rates were based upon very few 



424 



