DOHL ET AL.: COMMON DOLPHIN DISTRIBUTION AND ABUNDANCE 



variations in counts occurred randomly, with no pat- 

 tern to indicate in which method the higher counts 

 would occur. Small schools of <100 animals repre- 

 sented most of the sightings (53%). In medium-sized 

 schools, up to 300 animals, the variation was higher 

 (about 11%), and the photographs indicated probable 

 observer underestimation in 62% of the counts. The 

 largest underestimates occurred in large schools, 

 >300 animals, and were found in 76% of the observ- 

 er counts. These underestimates ranged up to 30% 

 in some circumstances. Within the large-school 

 category, two subcategories became evident: 1) Dis- 

 persed schools with multiple discrete subgroups of 

 animals gave the observers less of a problem than 

 2) the tightly grouped, rapidly moving, compact 

 schools. The dispersed large schools yielded under- 

 estimate values in the range of 14-16%, while the 

 compact, large groups were usually 21-23%. Ex- 

 tremely large schools of over 1,000 animals were 

 responsible for the highest error values of up to 30%; 

 these schools accounted for only 6.6% of total 

 sightings. 



Generally, we found that aerial estimates were 

 lower than numbers based on photographs and that 

 the larger the school, the higher the difference. We 

 attribute some of the difference to the time lag be- 

 tween when the count was made while circling the 

 school and the photo run over the center of the 

 school. Results of photo runs made either before or 

 after the counting effort did not vary significantly, 

 but occasionally, continued circling scattered larger 

 schools into several smaller subgroups. 



Sea surface glare affected observation efficiency 

 to some degree on about 10% of all survey days. Due 

 to the orientation of transect lines, glare conditions 

 could impair the search ability of only the left-side 

 observer on southwest-bound legs (up to 26% of total 

 search effort per survey day). Holt (1984 3 ) found 

 density estimates of dolphin schools to be 39% lower 

 under poor sun conditions than during good sun con- 

 ditions. Using his figure, we calculate that our over- 

 all seasonal density estimates might be low by about 

 1%. Because of the lack of any systematic bias 

 resulting from glare affecting density estimates in 

 one particular region or season more than another, 

 we made no corrections to adjust for this slight 

 underestimate. 



The perpendicular distance from the trackline to 

 the sighting was calculated from the declination 

 angle obtained using a hand-held inclinometer. Per- 



pendicular distances were recorded for 112 sightings 

 of common dolphin schools, representing 74.2% of 

 all sightings used in density calculations. 



Distributional Model 



Inspection of the first year's common dolphin 

 sighting numbers and plots of monthly distribution 

 indicated seasonal fluctuations of residency within 

 the Southern California Bight. 



Examination of the 3-yr database showed two 

 distinct seasons of occupancy for the species in the 

 SCB (Fig. 2). A comparison of the two sets of data 

 on a monthly basis show a significant statistical dif- 

 ference (^(1,34) = 7.66, P < 0.01). In view of these 

 observations, two seasons were defined for the 

 development of the distributional model: a summer- 

 autumn season (July through December) when com- 

 mon dolphin sightings were widespread in the SCB, 

 and a winter-spring season (January through June) 

 when most schools were confined to the southeast- 

 ern portion of the surveyed area. Common dolphin 

 sightings were assigned by their latitude and longi- 

 tude to 30' x 30' grid-cells (sampling quadrats) 

 centered on degree and half-degree lines of latitude 

 and longitude. Data were pooled to provide seasonal 

 estimates of common dolphin abundance for each 

 30' x 30' grid-cell. The estimate of density of groups 

 in cell i, D h was calculated from the relationship: 



Di = n l f{Q>)IZL l (Burnham et al. 1980) 



(1) 



3 Holt, R. S. 1984. Testing the validity of line transect theory 

 to estimate density of dolphin schools. U.S. Dep. Commer., 

 NOAA Admin. Rep., NMFS-SWFC LJ-84:31, 56 p. 



where n { is the number of groups encountered, /(0) 

 is the probability density function of perpendicular 

 distances evaluated at the ^-intercept, and h % is the 

 sum of all transect lengths in cell i contributing to 

 the seasonal estimate. The value of the f(0) term 

 was calculated using the nonparametric Fourier- 

 series estimator of Crain et al. 1978 (see Burnham 

 et al. 1980 for a complete discussion of this esti- 

 mator). Computations were made employing the 

 program TRANSECT (Laake et al. 1979). For calcu- 

 lation of the/(0) term, the perpendicular distance 

 of each sighting was reduced by one-half the width 

 of the exclusion area under the aircraft, where 

 visibility was obstructed by the fuselage (total ex- 

 clusion area = 530 ft at 1,000 ft ASL). This ap- 

 proach, in effect, moves the transect centerline out- 

 board to the point of nearest possible sighting 

 distance— a point where it is assumed that all 

 animals present will be seen and counted. The ques- 

 tion of how to deal with the problem of restricted 

 downward visibility and line transect theory has 

 been considered by others; however, the best treat- 



335 



