278 



Fisher/ Bulletin 98(2) 



single-engine aircraft. An aircraft something like the 

 DeHaviland Twin Otter may be necessary to accom- 

 modate the size and power requirements of the NOAA 

 lidar system. The system size and cost would prob- 

 ably increase significantly at an equivalent energy 

 of about 1 J. Thus, a practical range for penetration 

 depths would be between about 35 and 50 m. 



Discussion 



Interpretation of modeling results 



Our goal was to model various aspects of a lidar 

 survey system for anchovy with a focus on features 

 that might affect survey accuracy and precision. We 

 learned from our modeling of swath width that this 

 width has little or no effect on the rate at which 

 schools are encountered when they are aggregated 

 into school groups, as is commonly the case with 

 small pelagic fish, like sardines and anchovy, except 

 under very low biomass levels. Under conditions 

 of very low biomass, schools may become scattered 

 rather than aggregated, in which case encounter rates 

 would increase with swath width. Our analyses also 

 showed that the chance that all the transects would 

 not intercept any fish school was extremely small 

 because the number of transect lines is likely to be 

 much greater than 5, owing to the high speed of the 

 survey airplane. Thus, from the standpoint of survey 

 precision, swath width may be given a low priority. 



Lack of full vulnerability to the counting technique 

 is one of the most important potential sources of bias 

 for biomass sui-veys. Fish may not be fully vulner- 

 able because the survey does not extend over the full 

 geographic range of the stock and because there are 

 limitations to the counting system. Nearly all fishery- 

 independent surveys suffer to some extent from these 

 problems. In the case of an airborne lidar survey, the 

 depth limits of the sensing system could produce a 

 large potential bias, particularly if the system is used 

 during the day. Our model indicated that, on average, 

 36% of schools of anchovy during the day would be 

 expected to be below the maximum detection depth 

 of the lidar (2„,„j=41 m) (Figs. 3 and 8). Because 

 the vertical distribution of schools can vary consider- 

 ably between surveys, the undetected fraction would 

 vary, thus affecting survey accuracy. This computa- 

 tion is driven by our vertical distribution curve for 

 the daytime and the rapid attenuation of light in 

 water; packing density and fish size have negligible 

 effects. Thus, a reliable estimation of the biomass 

 of such small schooling fishes during the day in off- 

 shore waters does not seem practical unless a reliable 

 unbiased estimate of vertical distribution of schools 



is available. On the other hand, in water up to 30 

 m depth over the shelf, accurate estimates of bio- 

 mass for the daytime are practical because the verti- 

 cal movements of the fish would be restricted. 



If a lidar survey were restricted to night flights, 

 when schools are closer to the surface, the bias 

 caused by the uncounted fraction of deep schools 

 would be considerably reduced. During the night, 

 however, schools may become very diffuse and con- 

 sequently have a much lower target strength which 

 reduces their detectability. The good news from our 

 modeling work was that even the very diffuse schools 

 of 10-cm Japanese anchovy (0.53/m■^ Aoki and Ina- 

 gaki, 1988) at night were detectable over the upper 

 20 m. Our model indicated that 65% of all anchovy 

 schools would be detected during the night. 



In our study, we focused only on 10-cm anchovy 

 schools at two known and widely differing packing 

 densities and vertical distributions. We ran our models 

 with other packing densities and fish sizes, using data 

 from heiring, sardine, and mackerel but keeping the 

 vertical distributions the same as that for anchovy. - 

 These results indicated that packing density is an 

 important factor in the detection of schools at night 

 when the vertical distribution is shallow but is unim- 

 portant during the day when fish have a deeper verti- 

 cal distribution. The effect of packing density at night 

 can be significant. For example, at night at 30 m the 

 SNR for schools of 10-cm anchovy (packing density 

 0.53) was only 0.97, whereas that for 13-cm sardine 

 (4.0 packing density) was 12.38 (their detectability 

 was 13% and 77%, respectively). Schools of small 

 fishes may be inherently more detectable than those 

 of larger fish because the decline in average packing 

 densities of schools with increasing fish size is not 

 completely compensated by the increase in reflective 

 area of the fish ( packing density changes in propor- 

 tion to 1/L'^ (Misund, 1993), whereas the reflective 

 area changes in proportion toL-^). However, this theo- 

 retical relationship is eclipsed by the huge variation 

 in packing density due to behavioral factors. The 

 density of anchovy schools at night varies from com- 

 pact schools suitable for capture by the purse-seine 

 fishery ( Squire, 1972 ) to schools so diffuse that many 

 authors have concluded that schooling ceases (Whit- 

 ney, 1969; Baxter and Hunter, 1982). The packing 

 density used in our example of an anchovy school 

 at night from Aoki and Inagaki (1988) represents 

 such an extremely dispersed state, but dense con- 

 centrations of anchovy do occur at night, only their 

 packing density has not been measured. In fact, 4 

 fish/m'' used in the sardine night example might 

 be equally appropriate for anchovy. Unfortunately, 

 because field measurements of packing densities at 

 night are so infrequent, it is impossible to tell at this 



