300 



Fishery Bulletin 101(2) 



Series 1 10spp Series 2 5 spp Series 3 4 spp Series 4 23 spp Series 5 22 spp Series 6 25 spp 



1976 1978 1980 1984 1990 1997 1983 1990 1999 1993 1995 1998 1993 1995 1997 1999 



Series 7 1 5 spp Series 8 18 spp Series 9 25 spp Series 10 16 spp Series 11 15 spp Series 12 13 spp 



\. 



4.0 

 3 5 

 3-0- 

 2.5 

 20 

 1.5 

 1 01 



~P^ 



1991 1993 1996 1992 1996 2000 1984 1988 1992 

 Series 13 18 spp 



1991 1993 1995 1993 



1995 



1992 



1996 



Series 14 5 spp Series 1 5 25 spp Series 16 5 spp Series 17 5 spp 

 5 



,/ 



1992 1997 1992 1997 1984 1987 1990 1989 



Year 



1994 



1 



1986 1991 1996 



Figure 5 



Mean ranks (calculated as illustrated in Table 2) for each trawl survey data set. The solid horizontal line in each panel indicates the 

 overall mean rank; mean ranks outside the broken lines (circled pointsi indicate extreme survey years. 



masses changed by so much. For example, for series 1 

 the median change in biomass index between 1979 and 

 1980 (calculated over 10 species) was a factor of 3.4. It is 

 not plausible to say that the biomass of so many species 

 changed by that much in just one year. A second example 

 is years 1988 to 1990 for series 9. Here the median change 

 (over 25 species) was a halving, from 1988 to 1999, followed 

 by a doubling, from 1989 to 1990. Again, it is not plausible 

 to say that the actual biomasses changed by this much. 



Are there consistencies between data sets? 



The range of estimated trawl survey catchabilities is very 

 wide, covering more than two orders of magnitude, from 

 0.0035 to 1.6 (Fig. 6). Although the theoretical maximum 

 value for a trawl survey f/ is 1 , the two values that exceed 

 this may not be of concern if we allow for estimation error. 

 However, the lowest values an- of concern. If these are accu- 

 rate, then it would seem inappropriate to use trawl surveys 

 to assess these stocks. For example, a q that is less than 

 0.01 means that more than 99^7, of the stock is, in some 

 sense, not available to the trawl survey — either because 



it is outside the survey area (low areal availability), does 

 not encounter the trawl (low vertical availability ), or easily 

 avoids it (low vulnerability). For two species the range of 

 values was implausibly wide: for species F the four esti- 

 mates varied by a factor of 79 (0.0039 to 0.31); for species 

 E the factor was 49 (0.0035 to 0.17) (the next widest range 

 was for species C, a factor of just 2.8). 



There is only limited scope for between-series com- 

 parisons because the years or seasons covered by different 

 series may not overlap and, in any case, only about one 

 in six years is labeled as extreme. There are three years 

 which were labeled as extreme for more than one series: 

 1984. 1989, and 1995. In two of these three, the labels are 

 consistent: series 9 and 15 (both of which were deepwater 

 surveys targeting orange roughy, Hoplostethus atlanticus, 

 in different areas) agree in finding catchability to be high 

 in winter 1984 and low in winter 1989. For 1995, two 

 surveys found low catchability (series 5 in depths 20-400 

 m in February and March, and series 8 in 200-800 in in 

 January) and two found high (series 10 in 750-1500 ni in 

 Octoix'r and November, and series 13 in depths 20-400 ni 

 in March ami April). (liven the differences in depth ranges 



