276 



Fishery Bulletin 98(2) 



and settings of the lidar as well. 

 Some generalizations can be made. 

 During the night the lidar field of 

 view can be large because less back- 

 ground light exits to interfere with 

 the signal. Under these conditions, 

 the lidar attenuation coefficient (a) 

 will be very nearly equal to the dif- 

 fuse attenuation coefficient. For the 

 Jerlov open-ocean water types at 532 

 nmi, a varies from about 0.05/m (type 

 I) to about 0.11/m (type III). The 

 values for the Jerlov coastal water 

 types range from 0.15 (type 1) to 

 about 0.53 (type 9). From our analy- 

 sis, we expected that most anchovy 

 schools would be detected during 

 the night in the open ocean for a < 

 0.1. 



During the day, the situation is 

 more complicated. A lidar system 

 with a large field of view will have 

 a smaller signal-to-noise ratio be- 

 cause scattered sunlight reaches the 

 receiver. We could have increased the signal to noise 

 by decreasing the field of view, but this would tend 

 to increase a. Besides, an increase in the signal-to- 

 noise ratio during the day would have little effect on 

 q, because the vertical distribution of schools during 

 the day has a long tail extending down to 155 m 

 (Castillo Valderrama, 1995). At night, an increase 

 in the signal strength that extends the maximum 

 depth of a return by 10 m or so could have an impor- 

 tant consequence because of shallow vertical distri- 

 bution (829f of the school are in the upper 20 m). 



Comparisons with vision-based methods 



We compared the ability of a human observer to count 

 fish schools with the capability of a lidar. Hara ( 1990) 

 reported that an observer flying at 500 m would be 

 able to detect sardine schools along a 1600-m swath 

 and to a depth of about 4 m. We assumed that all 

 schools at 4 m could be detected visually and none 

 was detected below that depth. For the lidar, we used 

 a swath width of 7 m, and we considered the prod- 

 uct of encounter probability (p^,) (Eq. 1) depicted by 

 swath width and the maximum proportion of schools 

 detected (q) depicted by depth (Eq. 17) to be a mea- 

 sure of the overall performance. 



A human observer detects somewhat more schools 

 in the horizontal plane than does a lidar system 

 because of the relatively large swath width provided 

 by aerial viewing (Table 5). The encounter probabil- 

 ities in the horizontal plane (Eq. 1) obtained from 



simulation for a population of 32,000 schools was 

 0.51 for visual detection and 0.42 for lidar detection. 

 In the vertical plane, however, our analysis demon- 

 strated that lidar is superior in detecting schools 

 both during the day and night. During the night, 

 the proportion of anchovy schools detected (c/) by the 

 lidar was 0.65, whereas it was 0.28 for the human 

 observer. The difference between visual detection 

 and lidar detection was much greater during the 

 day (g was 0.63 for the lidar), whereas that for the 

 observer was only 0.095. 



The p^q (an overall measure of detection perfor- 

 mance) for a lidar was at least 1.9 times that of 

 an aerial observer. This means on the average that 

 the proportion of anchovy schools detected during a 

 survey would be about twice as great for a lidar as it 

 would be for an aerial observer. 



We have considered here, however, only one aspect 

 of the two systems — detection rates. Many other 

 differences also exist in relation to species identifi- 

 cations, biomass, and effects of environmental con- 

 ditions on the observing system. 



Laser power and penetration depth 



A set of parameters used to compute the laser power 

 and penetration depth for schools of anchovy is 

 listed in Table 3. The lidar signal was computed 

 by using Equation 8, and the penetration depth 

 was computed from the attenuation coefficient esti- 

 mated from Equation 12. The relation of penetration 



