THE PRECISION OF SIMULATED TRANSECT SURVEYS OF 

 NORTHERN ANCHOVY, ENGRAULIS MORDAX, SCHOOL GROUPS 



Paul C. Fiedler' 



ABSTRACT 



Simulated transect surveys of model anchovy populations were compared in terms of precision and 

 efficiency. The precision of systematic surveys varies inversely with the distance between transects. 

 Systematic surveys give more precise population estimates than random surveys, due to the large 

 positive correlation between closely spaced transects. The precision of stratified systematic surveys is 

 not significantly different from that of the unstratified surveys when the school groups are randomly 

 distributed in the survey area. However, stratified systematic surveys are more precise when the school 

 groups are clumped in one end of the survey area. The results of the simulations show that the patchy 

 distribution of anchovy schools can be a major source of error in population estimates. 



Any sampling program intended to estimate the 

 size of a population is subject to a variety of errors 

 which may reduce the accuracy or precision of the 

 estimate. Precision is the reciprocal of the varia- 

 tion of replicate estimates. Successful manage- 

 ment of a northern anchovy fishery in California 

 will require the monitoring of changes in the popu- 

 lation size. Acoustic survey techniques are cur- 

 rently being developed to obtain population and 

 biomass estimates independent of the fishery 

 (Hewitt et al. 1976). As in the study of any biologi- 

 cal population, it will be important to avoid con- 

 fusing the variation of a series of estimates due to 

 sampling error with true fluctuations in the popu- 

 lation size. 



Precision may be affected by 1) the manner in 

 which the sampled population is distributed in 

 space, and 2) variations within the sampling 

 method itself Several studies have shown that the 

 patchy distribution of individuals in a population 

 may cause considerable variation in replicate 

 population estimates and that the variation is re- 

 lated to sample design. Winsor and Clarke ( 1940) 

 studied the variation of catches in series of 

 plankton net tows. Although they did not separate 

 the components of between-tow variation due to 

 factors (1) and (2), it was observed that oblique 

 tows were more precise than vertical or horizontal 

 tows. Barnes and Marshall (1951) took an exten- 

 sive series of replicate pump samples and attri- 

 buted the considerable variation observed to the 



■Scripps Institution of Oceanography, Universitv of Califor- 

 nia, San Diego, La Jolla, CA 92093. 



nonrandom distribution of the zooplankton since 

 the volumes filtered were known accurately. Taft 

 ( 1960) analyzed the variance of sardine egg counts 

 in a grid of closely spaced stations. The distribu- 

 tion of eggs was extremely patchy (the densities 

 between samples ranged over more than four or- 

 ders of magnitude) and the relative 95^}^ con- 

 fidence limits for an estimate of the egg population 

 in the area of the grid from a single sample were 

 represented by a factor of 62. A simulation study 

 by Wiebe (1971) showed that the precision of zoo- 

 plankton population estimates depends both on 

 the sampling design (net size and tow length) and 

 the distribution of the population (size and loca- 

 tion of patches). 



Similar studies have investigated the precision 

 of sampling fish populations. Taylor ( 1953) discuss- 

 ed the implications of the patchy distribution of 

 fish for the optimum design of trawl surveys to 

 estimate population size. Cram and Hampton 

 ( 1976) demonstrated that the patchy distribution 

 of pilchard schools can cause imprecision suf- 

 ficient to render a population estimate useless for 

 management. 



The anchovy population is patchy on two levels: 

 individual fish are aggregated in schools and 

 schools themselves tend to be aggregated in school 

 groups. This patchiness, or nonrandomness, is ex- 

 pected to be a major source of variation in popula- 

 tion estimates. The present study simulated sur- 

 veys of model anchovy populations to determine 

 the effect of patchiness on the precision of popula- 

 tion estimates. Three transect survey designs 

 were compared: systematic, random, and strat- 

 ified systematic. These are merely different 



Manu.-icnpt accepted .January 1978. 

 FISHERY BULLETIN: VOL. 76. NO. 3. 1978. 



679 



