HOLT: DENSITY OF DOLPHIN SCHOOLS 



ship offshore estimate was 2.71 schools/1,000 km^ 

 with a standard error of 0.334. 



DISCUSSION 



Onshore-Offshore Density 

 Gradients 



The onshore-to-offshore density gradient de- 

 creased based on aerial data in the inshore area 

 and comparison of inshore and offshore density 

 estimates. Offshore density estimates were only 

 about one-half the inshore estimates (Table 1). 

 Although sea state and sun glare conditions were 

 confounded with distance from shore, compari- 

 sons of detection rates in the two inshore density 

 bands for data stratified by Beaufort state or sun 

 conditions indicated lower rates in the outer band 

 (Fig. 5). 



Fit of Fourier Series Model 



Burnham et al. (1980) provided criteria for se- 

 lecting the appropriate number of terms in the FS 

 model. However, these criteria were not satisfac- 

 tory for use with the aerial and ship perpendicu- 

 lar distance distributions, which had pronounced 

 modes at the origin. Instead, I selected models 

 which had the fewest terms but provided a good 

 fit near the origin. This resulted in models with 

 large numbers of terms. However, to the degree 

 that the modes are representative of school den- 

 sity, my estimates of densities will be unbiased. 

 Alternate statistical models need development 

 which can fit data which lack a shoulder near the 

 origin (i.e., data with pronounced modes at the 

 origin). Buckland (1985) investigated several 

 models but concluded that reliable estimation is 

 not possible unless a shoulder exists. 



Line Transect Assumptions 



Aerial Data 



Confounding of aerial sea state and sun condi- 

 tion data with distance from shore made it impos- 

 sible to test the assumption that all trackline 

 schools were detected during all viewing condi- 

 tions. If viewing conditions had been homoge- 

 neous throughout the area, the density estimate 

 calculated for calm sea and good sun conditions 

 (12.64 schools/1,000 km") could be used for the 

 inshore area (Table 1). This estimate is over 7 

 times the rough sea and poor sun estimate (1.78 

 schools/1,000 km^). However, the calm seas and 



good sun condition effort occurred mostly in the 

 northern nearshore region of the inshore area 

 (Fig. 3, 4) where density may be high. 



Consequently, Holt (fn. 8) conducted an aerial 

 experiment in a relatively small area to test sea 

 state and sun effects upon LT density estimates. 

 The results indicated that sun glare adversely 

 affected estimates of school density. The density 

 estimate was 39% larger during good sun condi- 

 tions than during poor conditions. Although den- 

 sity estimates were larger for calm sea data than 

 for rough sea data, the differences were not signif- 

 icant. 



The aerial experimental data (Holt fn. 8) may 

 be used to estimate maximum bias for sun and sea 

 state effects. The adjusted density estimate (D^ ) 



is 



Da=^^D,P,j 



1=1 7=1 





where D, 



P„ = 



D', 



Density estimate in survey area 

 during ith sea state and jth sun 

 condition, 



Proportion of effort in survey area 

 with ith. sea state andj'th sun con- 

 dition. 



Experimental density estimate 

 during ;th sea state and jth sun 

 condition determined from Holt 

 (fn. 8). 



In addition, i equal 1 denotes calm sea states and 

 i equal 2 denotes rough sea states, and j equal 1 

 denotes good sun conditions and J equal 2 denotes 

 poor sun conditions. An estimate of the sampling 

 variance (Var(D^ )) using the Taylor approxima- 

 tion method is 



2 2 



Vdr(D^ ) = 2 S P'/ 



1=1 j=i 





Vdr(D'u) 



+ 



+ 







Vdr{D\ 



The adjusted inshore density estimate is 4.51 

 schools/1,000 km^ with a standard error of 1.107. 

 This is an 8% increase over the unadjusted esti- 



431 



