than random fluctuations about the mean values. The August- 

 October 1975 and April-June 1976 demersal trawl surveys were 

 conducted during a period of cold temperatures that occurred 

 from January 1971 to December 1976. 



In general, surface seawater temperatures recorded at lat. 

 57°N, long. 170°W during the 4-6 mo preceding both the 1975 

 and 1976 surveys were unusually cold (Fig. 80). During August- 

 October 1975, however, surface temperatures were only slightly 

 cooler (anomalies of 0.0° to - 1.0 °C) than the 1962-75 mean. In 

 comparison, cold winter sea surface temperatures (anomalies of 

 - 1.8° to - 2.0 °Q continued through the April-June 1976 survey 

 period. In both 1975 and 1976, June bottom water temperatures 

 in the southeastern Bering Sea were quite cold compared to obser- 

 vations during most other years. 



As a third measure of environmental conditions during the 

 1976 survey period, Figure 81 compares the distribution of sea ice 

 observed during April- June 1976 to the extreme southern exten- 

 sions of ice cover observed during 1954-70. In general, the extent 

 of ice cover during April and May 1976 was at least equivalent to 

 the 17-yr (1954-70) extremes. 



Before discussing the biological results and comparability of the 

 1975 and 1976 eastern Bering Sea surveys, and relationships to the 

 above model, it is appropriate to reevaluate basic objectives, 

 assumptions, and limitations of the overall program. The funda- 

 mental objective of the surveys was to obtain two short-term 

 descriptions of characteristics (including density distribution, total 

 size, age structure, etc.) of the mixed species populations caught 

 by the sampling gear and to compare these characteristics between 

 time periods. Target populations were implicitly defined by the 

 location of the sampling area and its boundaries, and selective 

 characteristics of the sampling gear and field methods. 



The real identity of the target populations of a trawl survey is 

 frequently confused with the desire to measure total population 

 size, true population density, and the commercially fished popula- 

 tions. Clearly, the validity of intentions to measure total popula- 

 tion abundance is dependent upon relationships between the area 

 surveyed and species range. Because of biasing characteristics of 

 sampling gear and methods (i.e., imperfect efficiencies), estimates 

 of true population density are usually expressed as measures of 

 available population density (where available population density is 

 usually some fraction of true population density), unless the 

 sources of error have been identified and corrections can be ap- 

 plied. The true target populations of a survey, therefore, are 

 usually boundary and gear dependent. 



The relationships of the target populations of a survey to com- 

 mercially fished populations are then dependent upon 1) the 

 character of temporal and geographical overlap between survey 

 and commercial fishing activities, 2) selectivity resulting from gear 

 and methods, and 3) the importance of population movements 

 within, into, and from the regions of survey and commercial 

 fishing. 



The accuracy of estimates obtained from a trawl survey may 

 have several meanings based upon different references. Estimates 

 of relative abundance (i.e., catch per unit of gear operation) are 

 used to evaluate potential catch rates available to commercial 

 fishing and to measure variations in the relative densities of the 

 biological populations in space and time. The accuracy of point 

 estimates of relative abundance is dependent upon constant pro- 

 portionality between abundance indices and the actual abundance 

 of the populations. Overall estimates of relative abundance are 

 also affected by the relation of survey design to the distributions 

 of target populations, and the validity of stationary distributions 

 and closed populations assumptions. 



Estimates of absolute abundance (i.e., population density per 

 unit area) are used to assess the size of populations within a de- 

 fined area and to estimate their potential absolute yields (weight, 

 or numbers per unit time). The accuracy of these estimates is de- 

 pendent upon 1) the accuracy of the estimates of relative abun- 

 dance from which the estimates of absolute abundance are de- 

 rived, 2) biases due to sampling inefficiencies of the trawl gear, 

 and 3) potential biases during expansion from per unit to overall 

 estimates. 



The accuracy of trawl survey estimates, then, may refer to 

 either 1) the fidelity of sample estimates to reflect constant pro- 

 portionality to the real world, 2) the departure of estimates of ab- 

 solute properties from real world values, or 3) the departure of 

 estimates of absolute properties from true, but indefinable, 

 characteristics of artificial available populations determined by 

 the fishing gear. 



Comparison of Results Between Surveys 



Two comparisons of the overall results of the 1975 and 1976 

 surveys are shown in Tables 64 and 65. Table 64 compares indices 

 of relative fish abundances observed during the two surveys. Table 

 65 compares the absolute population estimates obtained for each 

 species, uncorrected for differences in geographical coverage be- 

 tween surveys. Population estimates are also compared in Table 

 65 against total commercial catches in the eastern Bering Sea dur- 

 ing 1975. 



Mean relative abundances showed large differences between 

 surveys, both within individual subdivisions of the survey area 

 and overall (Table 64). Differences in overall apparent densities 

 may have been due to 1 ) changes in sampling efficiencies and sam- 

 pling biases, 2) true population growth due to recruitment, or 

 decline due to mortality, and 3) population in-migration to, or 

 out-migration from, the overall survey area. The importance of 

 these potential effects has previously been discussed in the presen- 

 tation of survey results for each species. Differences in apparent 

 mean densities within individual subdivisions may then have been 

 caused by 1 ) all of the above potential effects and 2) changes in 

 the geographical distribution of populations between subareas 

 due to seasonal or between-year shifts and migrations. 



Comparisons of absolute population estimates between surveys 

 also showed large differences (Table 65), reflecting changes in ap- 

 parent mean densities (Table 64), and effects due to the reduced 

 area surveyed in 1976 (Table 1). The accuracy of these estimates, 

 relative to either true or available population references, cannot 

 yet be well assessed due to the unavailability of estimates from 

 other sources using independent procedures. 



Although comparisons of the 1975 and 1976 survey estimates 

 with the total 1975 commercial fish catches provide perhaps unex- 

 pected results (Table 65) — where one year's total removals of 

 walleye pollock, Pacific cod, Greenland turbot, and arrowtooth 

 flounder represent a large proportion of, or exceed, the survey 

 population estimates — these inconsistent results may indicate 

 poor comparability due to differences in space and time dimen- 

 sions. First, there are questions regarding the degree of 

 geographical overlap between the survey and commercial catch 

 data sets. Secondly, in making comparisons of short-term survey 

 population estimates against a year's total removals, somatic 

 growth, recruitment, and in-migration must also be considered as 

 potential sources of differences. In particular, in-migration pro- 

 cesses could be quite important if individuals from outside — such 

 as deep or midwater populations — migrate to replace populations 

 removed from preferred grounds within the survey area. 



112 



