Lo et al : Modeling performance of an airborne lidar survey system for anchovy 



265 



aerial survey (narrow swath, deeper penetration). 

 We discuss each. 



Precision of an airborne Hdar survey will depend 

 upon the number of transects flown and the probabil- 

 ity of encountering schools along them. The width of 

 the transect lines ( swath width ), may affect the prob- 

 ability of encountering schools and therefore could be 

 one of the few factors affecting precision that involve 

 instrument design. Swath width could be increased 

 in a variety of ways (flying higher, scanning or opti- 

 cally expanding the laser beam), but such changes 

 are accompanied by disadvantages (loss in penetra- 

 tion depth, reduced resolution, increased instrument 

 cost and weight). In our study, we modeled how the 

 width of the swath (width of transect line) cut by the 

 survey instrument affects the probability of encoun- 

 tering fish schools, and therefore the precision of the 

 survey estimate, assuming fish are uniformly dis- 

 tributed in the water column. 



The accuracy of a biomass survey depends on the 

 extent to which the entire stock is vulnerable to the 

 counting technique and on the variability in size of the 

 uncounted fraction (Gunderson, 1993). The key issue 

 for accuracy of a lidar survey is the vulnerability of 

 a stock to being counted in the vertical plane. Depth- 

 specific detection by a lidar depends upon laser power, 

 sensitivity of the detection system, the rate of exponen- 

 tial decay of the laser pulse with water depth, the way 

 the fish-detection function of the instnament changes 

 with signal attenuation, fish size and reflectivity, school 

 packing density, and, of course, the vertical distribu- 

 tion of the fish. Using a set of models and taking into 

 account many of these variables, we evaluated the 

 effect of instrument and survey design on the accuracy 

 of an aerial lidar survey for measuring fish abundance. 

 We considered how variations in laser power, school 

 size, diel changes in vertical distribution of schools 

 and school packing density (number of fish per m'^) 

 would affect the accuracy of the survey. We also used 

 these models to estimate the maximum depth at which 

 schools might be detected by a single lidar pulse. For 

 our study, we chose to use anchovy because more data 

 exist on anchovy schools than most other species. Lo et 

 al.- have, however, recently applied the same models 

 to sardine and hemng schools. 



Materials and methods 



We used various models to evaluate the potential 

 effects of instruments on survey design. To evaluate 



how swath width may affect survey accuracy, simu- 

 lation runs were used for a school-group encounter 

 model. To evaluate the relation between laser power 

 and maximum detection depth for fish schools, we 

 computed the probability of detecting schools as a 

 function of the signal-to-noise-ratio and estimated 

 laser power and the laser attenuation coefficient. 

 School parameters, size, distribution and density, 

 and survey area (46,204 km^=333 km (180 nmi) x 

 138.75 km (75 nmi)) were taken from acoustic sur- 

 veys of northern anchovy in the Southern California 

 Bight (Mais, 1974; Fiedler, 1978; Smith, 1981; Mac- 

 CalF). For daytime profiles, vertical distributions of 

 schools were based on northern anchovy off Califor- 

 nia (Holliday and Larsen, 1979); for nighttime pro- 

 files we used the distribution of early stage anchovy 

 eggs (Pommeranz and Moser, 1987) and acoustic 

 data for anchoveta off Peru (Castillo Valderrama, 

 1995). Signal-to-noise ratio was based on informa- 

 tion on packing density of schools provided by Aoki 

 and Inagaki ( 1988) and Graves ( 1977). 



Many pelagic fish schools form distinct aggrega- 

 tions or school groups (Cram and Hampton, 1976; 

 Fiedler, 1978). The area of an anchovy school (ex- 

 pressed by school diameter in our study) is highly 

 variable, as are the size and number of schools with- 

 in a school gi'oup. Because of this complexity, simula- 

 tions were used to compute the probability of encoun- 

 tering anchovy schools in a survey area (Fiedler, 

 1978). 



In the simulation, school groups were randomly 

 assigned in the survey area. The sizes of the anchovy 

 schools within a gi'oup were generated from the fre- 

 quency distribution of the diameters of northern 

 anchovy schools in the Southern California Bight 

 (Fiedler, 1978; Smith, 1981; Table 1). The number of 

 anchovy schools within an anchovy school group was 

 generated from the area occupied by the group and 

 the density of schools. Both the diameters of school 

 groups and the density of schools within a school 

 group were assumed to follow the lognormal distri- 

 butions measured for anchovy in the Southern Cal- 

 ifornia Bight (Fiedler, 1978; Smith, 1981) (Fig. 1). 

 Simulations were used to compute the encounter 

 probability {py'f for various swath widths iy). 



The locations of school groups were randomly allo- 

 cated in north-south ( n-s ) and east-west ( e-w ) direc- 

 tions. When school groups overlapped (intersected) 

 in the north-south directions, they were combined 

 as a "single" school group for computing the encoun- 



-' Lo, N. C. H., J. R. Hunter, and J. H. Churnside. 1999. Mod- 

 eling properties of airborne lidar surveys for epipelagic fish. 

 Admin. Rep. LJ-99-01. Southwest Fish. Sci. Ctr. NMFS, NOAA. 

 P.O. Box 271, La Jolla, CA 92037. 



■* MacCall, A. 1975. Anchovy population survey simulation: a 

 report ofCalCOFI Anchovy Workshop Group on methods of esti- 

 mating anchovy abundance, July 21-22. 1975, Contribution 4, 

 9 p. Marine Life Research Group. Scripps Institution of Ocean- 

 ography, 9500 Oilman Drive, La Jolla, CA 92037-0227. 



