Canary rockfish (Sebastes pinniger) COSEWIC assessment and status report: chapter 12
Population sizes and trends
- Population size
- Population trends from surveys in B.C. waters
- U.S. triennial survey
- West Coast Vancouver Island (WCVI) shrimp survey
- Regression analysis of triennial survey and WCVI shrimp survey
- Queen Charlotte Sound (QCSd) shrimp survey
- Queen Charlotte Sound groundfish survey
- Hecate Strait assemblage survey
- Goose Island Gully Pacific Ocean perch survey
- Abundance trends from Canadian commercial trawl catch per unit effort (CPUE)
- Other stock asssessments of the Canadian population(s)
- Trends in biological characteristics
- Summary of current abundance and trends in B.C. waters (southern, central and northern areas)
- Summary of population status
- Population trends and assessments in U.S. waters
Average recent total landings of at least 840 t/y with a mean weight of landed canary rockfish of 2.03 kg, equates to over 413 000 pieces landed each year, composed predominantly of mature individuals (GFBio: unpublished data). The population has sustained a continual harvest of this magnitude for over 30 years. In the absence of evidence of imminent collapse in abundance, or size and age composition (see below), it seems likely that the current standing population of adults is at least in the low millions. Certainly it cannot be in the low 100 000s. While an estimate of abundance with this uncertainty falls well short of characterizing the status of the population, it assists the discussion of whether the population is at risk to such issues as genetic drift.
An alternative low or underestimate of the standing population can be made by summing the area-expanded bottom trawl catch rates in recent British Columbia (B.C.) surveys (WCVI: unpublished data for 2004; QCSd: see Table 6; HS: unpublished data for 2005). These surveys are designed to monitor relative abundance of bottom dwelling fish species. They are conducted with Atlantic Western IIA bottom trawls and use a random stratified design. They survey bottom depths from 50-500 m, spanning the depth range of adult canary rockfish (Fig. 11; Table 3).
The resulting biomass estimate of 2 563 t assumes a catchability (between the trawl doors) of 1.0. U.S. research by Millar and Methot (2002) indicates a likely range for canary rockfish catchability in the U.S. triennial survey of 0.15-0.35. Applying this range to the B.C. surveys expands the 2 563 t to 7 300-17 100 t of canary rockfish biomass in B.C. survey areas. This does not include populations on the west coast of the Queen Charlotte Islands and inshore waters, which implies that this estimate is likely to be low. Given a mean weight of trawl caught canary rockfish of 2.03 kg, the range of expanded biomass estimates translates into a current abundance of 4 to 8 million adults in B.C. waters, given that the majority of the canary rockfish catch in the survey (by weight) is composed of mature fish.
Figure 11: Locations of trawl surveys that provide indices of canary rockfish abundance
All surveys target groundfish except two shrimp trawl surveys conducted in QCSd and off the WCVI.
Survey | Start Year |
End Year |
Number of Surveys | Depth Range (m) |
Bottom Trawl Gear Used |
---|---|---|---|---|---|
Hecate Strait Groundfish | 2005 | 2005 | 1 | 11-230 | Atlantic Western llA |
West Coast Vancouver Island ShrimpFootnote a | 1975 | 2005 | 31 | 15-258 | NMFS Standard Shrimp |
West Coast Vancouver Island Groundfish | 2004 | 2004 | 1 | 46-750 | Atlantic Western llA |
U.S. TriennialFootnote b | 1980 | 2001 | 8 | 55-477 | Noreastern |
Queen Charlotte Sound Shrimp | 1999 | 2004 | 6 | 15-309 | NMFS Standard Shrimp |
Queen Charlotte Sound Groundfish | 2003 | 2005 | 3 | 37-543 | Atlantic Western llA |
Goose Island Gully Pacific O. perch | 1966 | 2005 | 16 | 146-218 | various |
Hecate Strait AssemblageFootnote c | 1984 | 2003 | 11 | 18-232 | Yankee 36 |
The following discussion summarizes existing indices that can be used to infer abundance trends for canary rockfish in Canadian waters. These indices are:
- U.S. triennial bottom trawl survey (U.S. triennial survey)
- West Coast Vancouver Island shrimp trawl survey (WCVI shrimp survey)
- Queen Charlotte Sound shrimp trawl survey (QCSd shrimp survey)
- Queen Charlotte Sound bottom trawl survey (QCSd groundfish survey)
- Hecate Strait Assemblage survey (HS assemblage survey)
- Goose Island Gully Pacific Ocean perch survey
The U.S. triennial survey began in 1977 and typically covered northern California to the U.S./Canada border in northern Washington (Weinberg et al. 2002). For the years 1980, 1983, 1989, 1992, 1995, 1998, and 2001, it also extended into southern B.C. waters. The first two of these surveys extended to 49°15' N; the latter five surveys extended further north to 49°40' N (Fig. 12).
Figure 12: Set locations from the U.S. triennial survey conducted in 2001
The U.S. triennial survey indices for canary rockfish show a declining trend over the period of the survey, with the amount of decline depending on which area is considered (Fig. 13, Table 4).
Figure 13: Three biomass estimates for canary rockfish in the INPFC Vancouver region (total region, Canadian waters only and U.S. waters only) with 95% bias corrected error bars estimated from 5000 bootstraps
The trend for this species from the US-Vancouver section is -7% per year since 1980 while the trend in the Canada-Vancouver section is -14% per year, for an overall decline of about 95% (Fig. 14). The overall trend for the total Vancouver section is also a decreasing trend of -4% per year.
Fitting a log-linear regression to the Canada-Vancouver index values gives a regression significantly different from 0 and an overall decline of 96% over the series. Survey data are considered the most reliable method for monitoring demersal marine species, but typically (as here) give large error bars. Annual biomass estimates can be highly leveraged by 1-2 large tows. However, despite these caveats, this survey is considered to be a reliable index of population status in this area.
Area | Year | Mean bootstrap biomass | Lower bound biomass | Upper bound biomass |
---|---|---|---|---|
Total Vancouver | 1980 | 7 633 | 427 | 28 611 |
Total Vancouver | 1983 | 11 063 | 4 976 | 19 812 |
Total Vancouver | 1989 | 7 918 | 3 389 | 16 711 |
Total Vancouver | 1992 | 1 654 | 801 | 2 884 |
Total Vancouver | 1995 | 293 | 109 | 594 |
Total Vancouver | 1998 | 2 233 | 1 275 | 3 472 |
Total Vancouver | 2001 | 622 | 271 | 1 151 |
Canada Vancouver | 1980 | 8 082 | 306 | 30 811 |
Canada Vancouver | 1983 | 6 241 | 1 078 | 14 815 |
Canada Vancouver | 1989 | 4 814 | 1 303 | 13 362 |
Canada Vancouver | 1992 | 1 310 | 555 | 2 469 |
Canada Vancouver | 1995 | 253 | 88 | 504 |
Canada Vancouver | 1998 | 1 805 | 957 | 2 888 |
Canada Vancouver | 2001 | 351 | 75 | 850 |
U.S. Vancouver | 1980 | 158 | 0 | 390 |
U.S. Vancouver | 1983 | 4 647 | 1 726 | 8 963 |
U.S. Vancouver | 1989 | 3 104 | 1 106 | 6 165 |
U.S. Vancouver | 1992 | 344 | 138 | 801 |
U.S. Vancouver | 1995 | 40 | 12 | 103 |
U.S. Vancouver | 1998 | 427 | 242 | 707 |
U.S. Vancouver | 2001 | 271 | 102 | 508 |
The bootstrap estimates are based on 5000 random draws with replacement.
Figure 14: Biomass estimates for canary rockfish from the U.S. triennial survey grouped for the different zones
The lines represent an exponential fitted curve through the point estimates.
Note the improbable change in the U.S. Vancouver series from 1980 to 1983. It shows that this survey for this species can easily indicate population changes over the short term that are extremely unlikely. Even the low end of the error range for 1983 requires at least a 4X increase from the upper end of the 1980 estimate. There was no evidence of a large year class entering the fishery at this time.
Survey indices for canary rockfish are available from the WCVI shrimp survey which spans 1975 to 2006 (Fig. 15). This is the longest series available to monitor this species in Canadian waters and was conducted nearly annually over the entire period of record. These survey data were analysed, following the recommendations made by Starr et al. (2002), by post-stratifying the data into two areas, Areas 124 and 125, and treating the tows as having been randomly selected. Tows were selected in areas that had been consistently covered across depths over all years and the analysis was confined to a consistent set of vessels and survey months.
Figure 15: Canary rockfish index from the west coast Vancouver Island shrimp survey, 1975 to 2006
The bottom panel plots the index on the Ln scale, with the dotted line showing the least squares linear fit. The middle panel plots the same index but includes the 95% confidence intervals around each point. The top panel lists the total number of tows (N), the number of tows that captured canary rockfish (n), and the weight of canary rockfish captured (W, in kg) in each year of the survey.
The survey data were analysed using equations consistent with a random stratified survey and uncertainty was estimated by resampling the survey data with replacement for 1000 bootstrap iterations. Area stratum 125 was not surveyed in two of the survey years (1989 and 1991) so the mean catch rate from area stratum 124 in those years was used in its place to ensure comparability over all survey years.
Estimated biomass levels for canary rockfish from the WCVI shrimp survey have declined throughout the history of this survey, although there is considerable variability around the trend line, with some years of relatively high biomass estimates associated with high levels of relative error (e.g. 1977, 1983, 1994). Biomass levels appear to be gradually increasing since the late 1990s, but these indices also have high uncertainty, and there have been periods of increase or stability earlier in the series followed by a continuing decline.
Fitting a log-linear regression to the series of indices provides a regression significantly different from zero and an overall decline over the period of 78%, consistent with the pattern observed in the triennial survey.
The proportion of tows with canary rockfish shows a consistent trend towards increasing canary rockfish in recent years, following a period of decline, such that the proportions are now above the long term average (Fig. 16). The power of this index to detect real changes in abundance is unknown, although there is evidence that the frequency of non-zero catches is a valid alternative index and may sometimes be superior (Bannerot and Austin 1983). For this assessment, the biomass estimates are considered more reasonable as an index of abundance.
Trends in the WCVI shrimp survey catch rate indices were analysed following a step function methodology presented by Stanley and Starr (2004). The survey series was blocked into two or three periods of approximately equal length (Fig. 17 and 18). An alternative interpretation blocked the series into four periods (Fig. 19) which attempted to capture a beginning and ending cluster of 5 years, separated by two decadal groupings. The choice of the periods over which to summarize is obviously arbitrary, but it is easy to examine Figures 17-19 to assess the impact of using alternate groupings.
The average of the survey indices in each period was calculated in one of two ways: either as a simple average or by using the inverse of each survey CV (relative error) as a weighting factor (Table 5). This second approach down-weights indices which are associated with high relative error. Plots are presented for the two step, three step, and four step analyses using the inverse weighting assumption (Figs. 17-19). The analyses presented in this document estimate that recent abundance from this survey is 39% to 61% of the long term mean, or is 23% to 45% of the earliest period in the series (Table 5).
Figure 16: Proportion of tows with canary rockfish by year for the WCVI shrimp survey
The average proportion is shown by the solid line.
1) Recent abundance relative to overall mean abundance | 2) Recent abundance relative to abundance in earliest period | |||
---|---|---|---|---|
Simple average | Inverse weighting | Simple average | Inverse weighting | |
2-step | 0.56 | 0.48 | 0.38 | 0.28 |
3-step | 0.40 | 0.39 | 0.23 | 0.23 |
4-step | 0.51 | 0.61 | 0.45 | 0.39 |
Two averaging schemes were used for each comparison period: a) a simple average for the period; and b) an average where each index is weighted by the inverse square of the survey CV to account for differences in survey reliability. The period averages are scaled either by the mean of the entire survey series or by the mean of the first period.
The step approach presented above represents an alternative to a simple regression to characterize trends over time. If simple linear regression is fit to the shrimp survey data, it indicates a point estimate of decline over the entire period (1975-2005) of about 80% (consistent with the log linear analysis noted above), but this drops to 60% if the 1983 estimate is removed. The series does not appear monotonic, so the rationale for fitting a linear regression can be questioned. The step function approach may be more robust to the outlier index points which are present in this series and it makes fewer assumptions about the continuity of the series.
Figure 17: Two step function for the WCVI shrimp survey index, plotted relative to the mean of the survey series, weighted by the inverse of the CV2 for each survey
Figure 18: Three step function for the WCVI shrimp survey index, plotted relative to the mean of the survey series, weighted by the inverse of the CV2 for each survey
Figure 19: Four step function for the WCVI shrimp survey index, plotted relative to the mean of the survey series, weighted by the inverse of the CV2 for each survey
Population indices from the WCVI shrimp survey and the U.S. Triennial survey were analysed with a log linear regression model to estimate the rate of decline of canary rockfish as indexed by these surveys. The slope estimates were statistically significant in both cases at p <0.05. The WCVI index indicated a decline of 78% over the 31 year time series. The triennial survey indicated a 96% decline over the 22 year time series. An analysis combining both indices indicated no significant difference in slope between the surveys, and the estimated decline was 86% over the 31 year time period covered by both surveys.
A swept-area shrimp survey of QCSd has been conducted yearly since 1998 (Boutillier and Olsen 2000). Although the original design employs uniform sampling stations and uses spatial interpolation to estimate biomass, we re-analysed the surveys as if they were randomly stratified to arrive at the canary rockfish biomass estimates given in Table 6 and Fig. 20. The points indicate a rising trend for the central coast since 1999, but the survey is obviously imprecise, took canary rockfish in a low proportion of sets, and, in common with the other surveys summarized in the following section, covers only a short time period.
Year | Biomass (t) | Lower CI (t) | Upper CI (t) |
---|---|---|---|
1999 | 5.4 | 0.9 | 25.3 |
2000 | 0.8 | 0.0 | 2.3 |
2001 | 0.7 | 0.0 | 2.1 |
2002 | 9.5 | 2.9 | 22.6 |
2003 | 14.2 | 5.3 | 28.0 |
2004 | 2.4 | 0.0 | 7.3 |
Confidence intervals are at the 95% level.
Figure 20: Bootstrapped biomass estimates (t, bottom panel) and biomass + 95% confidence intervals (t, middle panel) for canary rockfish caught in the QCSd shrimp survey, 1999 to 2004
The top panel indicates: N = the number of sets conducted; n = the number of sets in which canary rockfish were caught; W = the total weight (kg) of canary rockfish caught.
A large-scale groundfish bottom trawl survey of QCSd was initiated in 2003 and repeated in 2004 and 2005 (Fig. 11) (Stanley et al. 2004). Funded primarily by the trawl industry, the current plan is to continue it on a biennial rotation. The survey is based on approximately 240 successful tows. Results indicate an increasing trend over the three years (Table 7, Fig. 21) but, as with the other surveys for this species, is obviously imprecise, although it captures a much larger number of canary rockfish than other surveys.
Year | Biomass (t) | Lower CI (t) | Upper CI (t) |
---|---|---|---|
2003 | 1326 | 709 | 2861 |
2004 | 1493 | 784 | 3313 |
2005 | 1701 | 349 | 5232 |
Confidence intervals are at the 95% level.
Figure 21: Bootstrapped biomass estimates (100's t, bottom panel) and biomass + 95% confidence intervals (100's t, middle panel) for canary rockfish caught in the QCSd groundfish survey, 2003 to 2005
The top panel indicates: N = the number of sets conducted; n = the number of sets in which canary rockfish were caught; W = the total weight (kg) of canary rockfish caught. The methods used to calculate the confidence intervals are the same as those used in the analysis of the QCSd shrimp survey.
DFO conducted a bottom trawl "assemblage" survey in HS in 1984-2003. However, it was conducted in waters which are too shallow for canary rockfish, resulting in catch rates which are extremely low. Canary rockfish were observed in only 1-11 sets/y of the 85-146 sets/y. The trend, such as it is, is downwards, although heavily leveraged by one high point in 1984 and two low points in 2002 and 2003 (Table 8, Fig. 22). We attach little confidence 38 to this trend owing to the low catch rates in the survey. This survey was re-designed in 2005, which added a few more tows in deeper water. It may prove to be more useful for tracking canary rockfish than the previous survey but it is still likely to be imprecise.
Year | Biomass (t) | Lower CI (t) | Upper CI (t) |
---|---|---|---|
1984 | 246 | 79 | 913 |
1987 | 23 | 3 | 87 |
1989 | 32 | 5 | 124 |
1991 | 159 | 25 | 659 |
1993 | 49 | 14 | 196 |
1995 | 39 | 6 | 115 |
1996 | 14 | 2 | 57 |
1998 | 37 | 1 | 244 |
2000 | 57 | 10 | 202 |
2002 | 1 | 0 | 3 |
2003 | 5 | 1 | 14 |
Confidence intervals are at the 95% level.
A Pacific Ocean perch (POP) survey in the Goose Island Gully of QCSd was conducted with reasonable frequency and the same design, vessel and gear from 1966 until 1984Footnote 6. It was then abandoned for 10 years, to be restarted in 1994 with different vessels, gear, and design, but then stopped again in 1995 (Hand et al. 1995; Yamanaka et al. 1996).
More recently, DFO and the trawl industry have initiated a much larger-scale multiple-species survey in 2003, which was repeated in 2004 and 2005 and will be continued on a biennial frequency starting in 2007 (Olsen et al. 2007).
The original POP survey was mostly in waters too deep for significant catches of canary rockfish. The low catches of canary rockfish, the long gaps, and the problematic assumption of constant catchability in the face of numerous re-designs discouraged us from exploring these data. However, the lack of survey information on population trends in the northern part of B.C.'s coast merits an examination of these data. This survey may also throw light on the question of whether the foreign trawl fisheries (USSR and Japan) in the 1960s and 1970s may have depleted the canary rockfish population in B.C. waters.
Figure 22: Bootstrapped biomass estimates (t) (bottom panel) and estimates + 95% confidence intervals (t) (middle panel) for canary rockfish caught in the HS assemblage survey between 1984 and 2003
The top panel indicates: N = the number of sets conducted; n = the number of sets in which canary rockfish were caught; W = the total weight (kg) of canary rockfish caught. The methods used to calculate the confidence intervals are the same as those used in the analysis of the QCSd shrimp survey.
The area common to all years corresponds to Goose Island GullyFootnote 7 (Figure 23) from depths of 146-218 m. Our attempt to standardize fishing power were limited to correcting for doorspread and average speed (Table 9, Figure 24).
Assuming a mean size of 2 kg, total catch ranged from about 8-370 fish over the entire time period. There were catches of canary rockfish in 2-16 tows from 1966-2005. The low catches of an aggregating species contribute to the implied large interannual variance.
The nominal results indicate a 56% decline based on the log-regression. The recent point for 2005 exerts significant leverage. Without it, the data indicate a 23% decline.
Figure 23: Goose Island Gully depth strata and tow locations within those strata, from historic POP surveys
The box inset shows the location of Goose Island Gully on the B.C. coast.
Year | Bootstrap Results (t) | RE | Num. Tows | Canary Tows | Catch Weight (kg) |
Doorspread (m) |
Speed (km/h) |
||||
---|---|---|---|---|---|---|---|---|---|---|---|
Index (t) | Mean | Median | Lower CI | Upper CI | |||||||
1966 | 198 | 199 | 198 | 0 | 565 | 0.81 | 8 | 2 | 246 | 62 | 5.6 |
1967 | 66 | 66 | 66 | 23 | 113 | 0.35 | 14 | 10 | 54 | 62 | 5.6 |
1969 | 87 | 87 | 82 | 33 | 226 | 0.47 | 18 | 9 | 163 | 62 | 5.6 |
1971 | 42 | 43 | 41 | 23 | 88 | 0.33 | 18 | 8 | 61 | 62 | 5.6 |
1973 | 88 | 88 | 85 | 18 | 294 | 0.67 | 18 | 8 | 104 | 62 | 5.6 |
1976 | 42 | 42 | 40 | 12 | 122 | 0.57 | 17 | 5 | 40 | 62 | 5.6 |
1977 | 422 | 417 | 400 | 87 | 1348 | 0.59 | 25 | 16 | 737 | 62 | 5.6 |
1979 | 60 | 60 | 58 | 24 | 165 | 0.46 | 28 | 5 | 184 | 71 | 5.9 |
1984 | 75 | 76 | 74 | 25 | 150 | 0.41 | 23 | 10 | 105 | 59 | 5.9 |
1984 | 118 | 122 | 124 | 31 | 152 | 0.20 | 6 | 4 | 70 | 43 | 5.6 |
1994 | 51 | 51 | 47 | 7 | 210 | 0.74 | 32 | 4 | 106 | 54 | 5.9 |
1995 | 274 | 277 | 267 | 16 | 884 | 0.67 | 32 | 6 | 416 | 54 | 6.1 |
1995 | 408 | 410 | 388 | 112 | 1093 | 0.53 | 34 | 8 | 464 | 53 | 4.8 |
2003 | 29 | 29 | 28 | 7 | 107 | 0.65 | 31 | 7 | 61 | 72 | 5.7 |
2004 | 44 | 44 | 44 | 0 | 166 | 0.89 | 17 | 2 | 46 | 72 | 5.7 |
2005 | 10 | 10 | 10 | 3 | 29 | 0.55 | 25 | 5 | 17 | 72 | 5.7 |
Figure 24: Biomass estimates from the Goose Island Gully Pacific Ocean perch survey and the QCSd groundfish survey (1966-2005)
Data standardized only by doorspread and towing speed.
The most reliable or comparable portion of the time series indicates no change from 1966-1984 (1% increase). As this period spans the period of foreign fishing, it can be inferred that the foreign fleets did not significantly deplete the stock, at least in QCSd. Using the survey data is obviously problematic; nevertheless it provides the only insight into a long term trend for the central coast.
Analysis of commercial trawl CPUE has been restricted to the period April 1996 through March 2005. The beginning date of this analysis corresponds to the start of atsea observer records, thus ignoring the earlier period of catch history that relied on fisher logs and sales slips. Catch rate data prior to April 1996 are not comparable over time, owing largely to the significant and varying degrees of misreporting. During this period a large number of landing events exist for which the fishing logs and sales slips were obviously falsified. It was apparent at the time that many, possibly the majority, of sales-slips (and logbooks) were completed to accommodate official species' trip limits. Furthermore, the trip limits were varied widely over time, thus the directions of the biases would vary from one year to the next, or over groups of years. The dysfunction in the catch reporting system and the resulting inability to manage to quotas was the primary reason that the Department of Fisheries and Oceans imposed 100% observer coverage on the trawl fishery in 1996. While the degree of misreporting was never documented in a manner which would support these concerns, catch rates from this period are not considered reliable.
Even with good catch data in the period 1996+, CPUE can be expected to be "hyper-stable" within the context of an individual vessel quota (IVQ) fishery (IVQs were introduced in 1997). As canary rockfish abundance varies, fishers in an IVQ fishery are likely to alternate between targeting and avoiding this species in response to changes in abundance, thus making CPUE appear to be stable. However, we assume this tendency towards hyper-stability would be overwhelmed by large-scale changes in abundance, particularly for declines because, at some point, IVQs will not be caught if abundance declines significantly. This should be manifest in the CPUE as well. Therefore, these analyses were conducted to examine whether there was evidence of a decline large enough to overcome the tendency for hyper-stability.
Trawl catch/effort data pertaining to canary rockfish from the DFO PacHarvTrawl database were analysed using two general linear regression models (GLM): one assuming a log-normal distribution based on the non-zero catches of canary rockfish and the other assuming a binomial distribution based on the presence/absence of this species in the catch. This analysis begins from April 1, 1996, which represents the period when the quality of data had been vastly improved through the imposition of 100% observer coverage on all the major trawl operators. The analysis was also restricted to tows at optimal depths for canary rockfish and confined to vessels which had been in the fishery for at least three years for a minimum of five trips per year. The analysis considered two fisheries for canary rockfish: the WCVI (Areas 3C+ 3D) and QCSd (5A+5B). A comparison of the two areas for each type of GLM analysis shows that there are similarities between series across areas (Fig. 25).
A comparison of the two areas for each type of GLM analysis shows that the binomial series are very similar for the two areas, with each area showing a strong increase between 1996/97 to 1997/98 and remaining fairly flat since. The QCSd binomial series shows a drop in the most recent fishing year while the WCVI series does not. The two sets of lognormal series differ more, with the QCSd series showing an increase in the first half of the series while the WCVI series shows an increasing trend in the latter half of the series. The WCVI canary fishery has a higher catch rate and a higher proportion of nonzero tows. These series of relative abundance indices should be interpreted with caution as they are derived from fishery dependent data and are subject to between-year effects which may originate from sources other than fish abundance.
Three of the four sets of CPUE abundance series (two models: lognormal and binomial for each of two areas outlined above) show an increasing trend of 5-6% per year, depending on the area and the regression model applied. The QCSd binomial model has a decreasing trend of -1% per year. Simple two-parameter models are not a substitute for a stock assessment model and are provided as one indicator of the overall trend over the analytical period. It is not possible to predict a "three generational" change for these populations because such a prediction would require a complex analysis and strong assumptions of stability over long periods which are unlikely to be met. Nevertheless, these data, with their limitations, do not indicate a decline in abundance in these areas, since 1996.
Figure 25: Comparison of two sets of CPUE indices each based on different regression model assumptions for each of three areas
Each series has been standardized relative to the geometric mean of the period 1996/97 to 2004/05. The error bars show ± 95% confidence bounds.
Stanley (1999) provided stock assessment advice for canary rockfish. The author conducted a catch curve analysis after blocking the age observations into groups of years to account for aging error. The resulting estimates of Z (instantaneous rate of total mortality) in all the periods for areas 3C+3D and 5A+5B (females: 0.046-0.10 and males: 0.03-0.07) were not significantly different from the range of possible M, indicating, by subtraction (F=Z-M), that the fishing impact was likely to be low. Even the most recent period (1996-1998) analyzed indicated that the estimates of Z were 0.092 and 0.095 for females from Areas 3C+3D and 5A+5B respectively and the Z estimates for males were 0.047 and 0.053 for the same two areas, indicating that the Z estimates continued to be near the plausible values for M. While the weaknesses of conducting catch curve analysis in isolation are well documented (Ricker 1975), the implied estimates of F in various epochs did not indicate an unsustainable level of fishing nor were they increasing over time for the two main regions. Thus, existing quotas at that time (Table 2) appeared sustainable and they have not been changed since then.
The recommended quota range tended to bracket historical mean landings. In the absence of quantitative risk analysis, the intent of the upper and lower bounds presented in Table 2 was to provide qualitative guidance to managers. Harvests less than the minimum level would incur negligible risk, while harvests above the maximum level could not be defended as being sustainable and may put the stock at risk. Walters and Bonfil (1999) provide two alternative stock assessments of canary rockfish. The first was based on an expansion of catch rates in the commercial fishery and used an area-swept biomass approach. However, they had no knowledge of catchability of the trawls and commented that they were "less than satisfied with the technique". Nevertheless, they estimated "minimum" biomasses of 3246-4932 t for the years 1994-1996, for the areas that were heavily trawled.
Their second method involved a single stock Bayesian assessment procedure. This procedure modelled populations over various assumptions of starting biomass (B0) and was tuned to the 1980-1996 qualified commercial trawl CPUE, in spite of the fact that those authors noted that the data indicated unrealistic trends in CPUE. As noted above (and in the bocaccio assessment, Stanley et al. 2001), catch and CPUE data were neither accurate nor comparable over this period owing to a variable management regime and trends in misreporting.
Walters and Bonfil provided a useful contribution by indicating the impact those trends would have as a tuning index for stock assessment but results should be interpreted with caution (Stanley 1999). The canary rockfish assessment, along with the other assessments in that work, were highly leveraged by the sudden drop in CPUE near the end of the time series (mid-1990s) which was associated with improvements in the reporting of catch data, the advent of the dockside monitoring program (DMP) in 1994 and complete observer coverage in 1996. Nevertheless, their analyses suggested that the ratio of current biomass (B1996) to unfished biomass (B0) in 19 trawl localities was 0.29-0.77 with a mean proportion of 0.49.
Length and age composition observations for commercial catches in Canadian waters are summarized in Figs. 5-8 and 26, shown separately for Area 3C+3D, 5A+5B, and 5E (there are too few data from 5C+5D). Since these data are collected “opportunistically” from the commercial fishery, the actual spatial distribution of these samples, within these areas, varies among years and may not be entirely representative of the fishery. This brings into question the comparability of these data over time and the specific possibility that stability in mean length or age might be an artefact of harvesters gradually finding relatively unexploited sub-stocks within these areas. However, this possibility is unlikely as the known areas of canary abundance in 3C+3D and 5A+5B are relatively small and have been continuously exploited since the late 1950s. Thus it is unlikely that serial depletion in recent decades would act to camouflage overall declining trends in mean size or age within these areas.
At larger spatial scales, however, this effect is more likely and this is why the data have been separated into regions. For example, Area 5E has only been fished since about 1977, thus pooling the samples from this area into a coastwide summary would cause the above artefact. Table 10 summarizes the available canary rockfish age samples and shows that the number of samples is too sparse to permit detailed exploration of how varying characteristics of each sample (see above), such as season, depth, or source (port sample versus at-sea), may influence comparability over time. However, the modest increase in presence of small fish in recent years (Figs. 6 to 8) may have resulted from some at-sea samples taken from shallower depths. Removing these samples results in larger mean sizes and ages in recent years (compare Figs. 6-8 with Fig. 26). Thus, while a serial depletion effect is unlikely to be present, there is evidence of more catches coming from shallower water and affecting the comparability of samples over time. This underlines the weakness of trend analysis in samples taken from opportunistic sampling. The recently initiated set of fishery independent surveys will provide more comparability in population samples, although these will not be representative of commercial catches.
Both nominal (unweighted) and weighted trends in mean length and age composition are presented. The weighted versions pool the same samples, while weighting each sample by the catch of canary rockfish associated with the sample (Figs. 27 and 28).
There is an apparent decrease in mean length for males in Area 3C+3D, but not for Area 3C+3D females. There is no overall trend for Area 5A+5B in either sex, although mean length may be increasing in recent years. The time series is short for Area 5E.
Mean age in Area 3C+3D shows a decline for both sexes from late 1970s until 1990 then no trend. The Area 5A +5B is without trend. The mean age of 3C+3D is lower than Area 5A+5B in recent years, although it appears similar to Washington State collections (Methot and Stewart 2005). The one 5E sample collected in 1977 indicated an unexploited age composition. Samples from this area now show a lower mean age, which is generally consistent with areas to the south.
Year | Port n |
Port N |
Observer n |
Observer N |
Research n |
Research N |
Total n |
Total N |
---|---|---|---|---|---|---|---|---|
1978 | 1 | 104 | 1 | 104 | ||||
1979 | 2 | 201 | 2 | 201 | ||||
1980 | ||||||||
1981 | ||||||||
1982 | 2 | 50 | 2 | 50 | ||||
1983 | 2 | 225 | 2 | 225 | ||||
1984 | 3 | 212 | 3 | 212 | ||||
1985 | 1 | 296 | 3 | 75 | 4 | 371 | ||
1986 | 2 | 75 | 2 | 75 | ||||
1987 | ||||||||
1988 | 1 | 50 | 1 | 50 | ||||
1989 | 1 | 25 | 1 | 25 | ||||
1990 | 1 | 33 | 1 | 33 | ||||
1991 | 2 | 102 | 2 | 102 | ||||
1992 | ||||||||
1993 | 3 | 151 | 3 | 151 | ||||
1994 | 1 | 52 | 1 | 52 | ||||
1995 | 4 | 211 | 4 | 211 | ||||
1996 | 1 | 62 | 3 | 135 | 4 | 197 | ||
1997 | 4 | 117 | 4 | 117 | ||||
1998 | 6 | 346 | 11 | 551 | 17 | 897 | ||
1999 | 2 | 108 | 7 | 321 | 9 | 429 | ||
2000 | 1 | 62 | 3 | 180 | 4 | 242 | ||
2001 | 3 | 165 | 3 | 165 | ||||
2002 | 1 | 59 | 4 | 152 | 5 | 211 | ||
2003 | 2 | 113 | 2 | 94 | 4 | 207 | ||
2004 | 3 | 153 | 7 | 299 | 10 | 452 | ||
Total: | 38 | 2461 | 44 | 2014 | 7 | 304 | 89 | 4779 |
Port = samples obtained at the offloading port; Observer = samples obtained at sea by on-board observers; Research = samples obtained at-sea during research cruises; n = the number of samples; N = the number of aged specimens.
Year | Port n |
Port N |
Observer n |
Observer N |
Research n |
Research N |
Total n |
Total N |
---|---|---|---|---|---|---|---|---|
1978 | 4 | 387 | 4 | 387 | ||||
1979 | 1 | 100 | 1 | 100 | ||||
1980 | 1 | 100 | 1 | 100 | ||||
1981 | 1 | 24 | 1 | 24 | ||||
1982 | 1 | 27 | 1 | 27 | ||||
1983 | 1 | 25 | 1 | 25 | ||||
1984 | ||||||||
1985 | ||||||||
1986 | ||||||||
1987 | ||||||||
1988 | 2 | 166 | 2 | 166 | ||||
1989 | ||||||||
1990 | 4 | 141 | 4 | 141 | ||||
1991 | 4 | 206 | 4 | 206 | ||||
1992 | 2 | 109 | 2 | 109 | ||||
1993 | 1 | 81 | 1 | 81 | ||||
1994 | 7 | 365 | 7 | 365 | ||||
1995 | ||||||||
1996 | 1 | 40 | 1 | 40 | ||||
1997 | 2 | 106 | 3 | 154 | 5 | 260 | ||
1998 | 1 | 59 | 1 | 48 | 2 | 107 | ||
1999 | 2 | 118 | 2 | 86 | 1 | 29 | 5 | 233 |
2000 | 3 | 165 | 1 | 49 | 4 | 214 | ||
2001 | 5 | 322 | 1 | 24 | 6 | 346 | ||
2002 | ||||||||
2003 | 2 | 109 | 2 | 60 | 4 | 169 | ||
2004 | 1 | 40 | 1 | 46 | 2 | 86 | ||
Total: | 45 | 2650 | 12 | 507 | 1 | 29 | 58 | 3186 |
Port = samples obtained at the offloading port; Observer = samples obtained at sea by on-board observers; Research = samples obtained at-sea during research cruises; n = the number of samples; N = the number of aged specimens.
Year | Port n |
Port N |
Observer n |
Observer N |
Research n |
Research N |
Total n |
Total N |
---|---|---|---|---|---|---|---|---|
1978 | 1 | 100 | 1 | 100 | ||||
1979 | ||||||||
1980 | ||||||||
1981 | ||||||||
1982 | ||||||||
1983 | ||||||||
1984 | ||||||||
1985 | ||||||||
1986 | ||||||||
1987 | ||||||||
1988 | ||||||||
1989 | ||||||||
1990 | ||||||||
1991 | ||||||||
1992 | ||||||||
1993 | ||||||||
1994 | ||||||||
1995 | ||||||||
1996 | ||||||||
1997 | 1 | 51 | 1 | 51 | ||||
1998 | 2 | 93 | 2 | 93 | ||||
1999 | ||||||||
2000 | 1 | 56 | 1 | 48 | 1 | 50 | 3 | 154 |
2001 | 2 | 158 | 2 | 158 | ||||
2002 | ||||||||
2003 | ||||||||
2004 | 3 | 125 | 3 | 125 | ||||
Total: | 6 | 332 | 5 | 299 | 1 | 50 | 12 | 681 |
Port = samples obtained at the offloading port; Observer = samples obtained at sea by on-board observers; Research = samples obtained at-sea during research cruises; n = the number of samples; N = the number of aged specimens.
Figure 26: The effect of shallow samples on canary rockfish proportions-at-age
Panel (a) identifies 4 shallow samples from Area 3C+3D. Removal of these samples from the proportions-at-age analysis yields the figure shown in panel (b). Compared to the original proportions-at-age plot shown in Fgure 6, this plot has markedly fewer fish in the younger age classes, for the years in which the shallow samples were removed. A similar pattern exists for Area 5A+5B (panels (c) and (d)) and Area 5E (panels (e) and (f)).
Figure 27: Trends in mean fork length for canary rockfish from (a) Area 3C+3D, (b) Area 5A+5B, and (c) Area 5E
The grey lines show the effect of weighting each sample by the total catch weight of canary rockfish from which the sample was taken. Sample catch weights are only available for more recent years.
Figure 28: Mean age versus year for canary rockfish from (a) Area 3C+3D, (b) Area 5A+5B, and (c) Area 5E
The grey lines show the effect of weighting each sample by the total catch weight of canary rockfish from which the sample was taken. Sample catch weights are only available for more recent years.
Estimates of abundance inferred from annual landings and from trawl surveys indicate that adult canary rockfish abundance in Canadian waters is probably at least several million adults. With respect to trends in relative abundance, information from different regions is presented, although the indices have been standardized to a common mean and are presented in combined graphs (Figs. 29-31).
Index | Units | Coverage | Signal | Weight/comments |
---|---|---|---|---|
1. US Triennial (Canada-Vancouver) | Biomass (t) | US border north to mid-Vancouver Island 1980-2001 | Log-linear regression: Decline 96% | High (gear and design appropriate) |
2. WCVI shrimp survey | Biomass (t) | Off west coast of Vancouver Island 1975-2006 | Log-linear regression: Decline 78% | High (gear and design appropriate) |
3. WCVI shrimp survey | Positive tows | Off west coast of Vancouver Island 1975-2005 | Decline then increase – no overall trend | Low (power to detect abundance changes unknown) |
4. WCVI shrimp survey | Biomass (t) | Off west coast of Vancouver Island 1975-2005 | Step functions: varying levels of decline (to 23-45%) | Low (promising approach) |
4. Combined 1 and 2 | Index | As above | Log-linear regression: Decline 86% | High (as per individual surveys) |
5. WCVI Commercial CPUE | Catch per unit effort | Off west coast of Vancouver Island 1996-2005 | Inspection: Increase | Low (commercial catch rate influenced by factors other than abundance) |
6. QCS shrimp survey | Biomass (t) | Queen Charlotte Sound 1999-2004 | Inspection: Increase/no trend | Low (short time series) |
7. QCS groundfish survey | Biomass (t) | Queen Charlotte Sound 2003-2005 | Inspection: Increase | Low (short time series) |
8. Hecate Strait assemblage survey | Biomass (t) | Hecate Strait 1984-2003 | Inspection: Decline | Low (canary depths not covered) |
9. Goose Island Gully Pacific ocean perch survey | Biomass index | Goose Island Gully 1966-2005 | Log-linear regression: decline 56% (23% without last point) | Low (depths not covered, changes in sampling methods over time) |
10. QCS commercial CPUE | Catch per unit effort | Queen Charlotte Sound 1996-2005 | Inspection: No trend | Low (commercial catch rate influenced by factors other than abundance) |
All are based on trawling. Numbers 1-5: southern part of range; numbers 6-10: northern part of range.
Figure 29: Relative biomass indices for canary rockfish from four longer term fishery independent surveys
All indices have been scaled such to a common mean calculated over the period 1983-2001.
Figure 30: Relative indices for canary rockfish from shorter term commercial trawl data in Areas 3C+3D and 5A+5B and from two fishery independent surveys in QCSd
All indices have been scaled to a common mean calculated over the period 1999-2004.
Figure 31: Proportions of non-zero tows of canary rockfish in six fishery independent surveys
A log-linear regression analysis of the WCVI shrimp survey information shows a decline of 80% over the period (1975-2006). Step-function analyses of this index show recent values of 39-61% of the long term mean, or 23-45% of the earliest period. Examining the trend in the proportion of non-zero tows (Figs. 16 and 31) from the same survey indicates that the index is at the same or higher levels than it was at the beginning of the survey. The U.S triennial survey catch rate (Figs.13 and 29) shows a fitted decline of 95%, while the average of four more recent surveys (1992-2001) in comparison with the average of the three surveys from 1980-1989 period indicates a decline of 92%. Both surveys (as all trawl surveys for species which aggregate) are influenced by occasional large tows which exert significant leverage on the annual survey estimates. This effect is illustrated by the results in the U.S.-Vancouver zone between 1980 and 1983, when the index rose from near zero to the largest index value of the time series. The time series of proportion of non-zero tows from this survey has not shown a consistent strong decline (Fig. 31).
South coast commercial trawl catch rates, as elsewhere, appear to be stable, if not increasing, since 1996. However, any observed trend in commercial trawl CPUE may be an artefact of the target/avoidance response by fishers within the context of an ITQ fishery. Biological samples from the southern coast appear to indicate a decrease in mean size and age over the long term, but are stable in recent years.
There is no long-term index available for the central coast (Area 5A+5B), other than the Goose Island Gully series of surveys which is considered to sample canary rockfish poorly and whose design has changed over the years. The QCSd groundfish survey indicates a possible increasing trend over the first three survey years (2003-2005) (Figs. 21 and 30), while the QCSd shrimp survey is more variable. Commercial trawl catch rates appear to be stable since 1996, although the same caveat presented for 3C+3D also applies to these commercial data.
The point estimates of minimum biomass for QCSd from the 2005 groundfish survey (assuming a catchability of 1.0) indicate that there is likely to be at least 1795 t (95% confidence range: 433-5668 t) in 2005 compared to the 738 t (95% confidence range: 417-1390 t) estimated for the WCVI in 2004. While catchabilities cannot be assumed to be equal among both areas; the nominal results imply that there is a larger biomass of canary rockfish in the central region.
Biological samples from the central coast do not indicate a trend in mean size or age over the long term. Fishers have long reported that there is a significant population of canary rockfish in the north coast (Area 5E), although northern waters have generated few landings. Their opinions are based on significant acoustic sign of rockfish over untrawlable bottom in canary rockfish depths. This acoustic “sign” has also been noted by research staff and partially confirmed with tows of canary rockfish during numerous research trips. There are only a few places where canary rockfish can be captured by trawl, given the rough bottom topography, but fishers report that the low quotas in this area have prevented expansion of this fishery. Canary rockfish are frequently encountered when hook-and-line fishing in this region.
Biological samples from the north coast (from the West Coast of the Queen Charlotte Islands) are limited. Comparison of recent samples with one sample collected in 1978 possibly indicates that there has been an impact from exploitation. Current mean size and age are similar to southern and central coast samples. However, this interpretation may be affected by how the samples were obtained.
The two most reliable survey indices come from the WCVI shrimp survey and the U.S.A. triennial survey. Analysed separately with a log-linear model, the WCVI shrimp survey indicates a decline rate of -0.051 yr-1 and the triennial survey has a decline rate of -0.15 yr-1 (Appendix 1). In an analysis of covariance of these two surveys, with survey as a categorical variable, the interaction term is not significant (i.e. different slopes between surveys) and the intercept term is significant. The common slope estimate for the analysis of covariance is -0.064 yr-1, close to the WCVI survey index which has many more data points than the triennial.
The common slope (combined series) decline is considered to best represent the trend in population abundance. This gives a total decline of 86% over 30 years or 1.0-1.5 generations. Because information on numbers of mature individuals is not available over the entire time period for both surveys, total biomass is used here as a proxy for numbers of mature individuals. Given the apparent loss of older individuals over the period covered (Figure 6), the decline in biomass of mature individuals would probably be steeper than that observed for all individuals.
Fishing is the most likely cause for the observed decline; although a reduction in recruitment could have contributed, there is little clear evidence that this has occurred. Based on a simple production model, observed catches and plausible estimates of biomass early and late in the period of observed decline are consistent with the observed decline of 86% in the period for which information is available, suggesting that fishing alone could have caused the decline under “average” biological conditions. Asustained period of poor recruitment in the 1990s has been reported for many groundfish stocks in the Washington-California area (King 2005), and a period of poor recruitment for canary rockfish starting in the mid-1990s was anticipated in the most recent assessment (Stanley 1999). However, there is no strong indication of reduced recruitment in available age and length information (Figure 6). Thus although a reduction in recruitment due to environmental factors could have contributed to the decline, there is no strong evidence for such a reduction and it is not necessary to explain the decline.
There is substantial uncertainty in interpreting trend in population abundance in the most recent years (since 1995). The triennial survey stopped in 2001 (Figure 14); indices since the early 1990s were variable around a low level. The WCVI shrimp survey has provided annual indices over the entire period available (Figure 15). Because of the high year-to-year variability, indices from this survey since 1995 could be interpreted as continuing an earlier decline, stability, or even increasing.
Given the uncertainty about whether the decline has ceased, it appears that the most conservative and cautious interpretation of the trend is a continuing decline (as in the calculations of decline above and shown in Figure 15). Although there is some indication of an increase in younger individuals in recent years (Figure 6), there is no strong indication of increased recruitment which might drive an increase in abundance; and any increase in the index driven by an increase in young individuals would not reflect an increase in the mature population. Catches have not declined substantially over the past decade.
Because of the loss of older individuals over the time period covered, the observed declines would underestimate the decline in mature individuals, further suggesting that trend in recent years is most cautiously interpreted as a continuing decline.
There are several sources of uncertainty in interpreting population status. Uncertainty about recent trend in the most reliable index is discussed above. There is also uncertainty with respect to applying population trend indices from one part of the distribution (southwest, off the west coast of Vancouver Island) to the whole distribution. There is uncertainty about trends in biological characteristics (particularly age and length) since biological sampling has generally been at a low level, pattern has varied over time, and changes in fishing patterns could have influenced the data. Overall there is considerable variability in the information. However, the strong decline in two relatively reliable indices is considered to provide a relatively clear signal of population trend.
U.S. research staff have recently updated the assessment of canary rockfish from the Washington/B.C. border to southern California, which is treated as one stock (Methot and Stewart 2005). Data sources include catch, length- and age-frequency data from 10 fishing fleets and the U.S. triennial survey. These data were used in a catch-at-age analysis tuned with an index from the U.S. triennial survey, although in this case, the data from the entire triennial survey from California to the Canadian border were used. This series of survey data include additional surveys in 1977, 1986 and 2004 which did not venture into Canadian waters. Current stock status in the U.S. was summarized (Methot and Stewart 2005 p. 10) as:
“Canary rockfish were relatively lightly exploited until the early 1940’s, when catches increased and a decline in biomass began. The rate of the decline in spawning biomass accelerated during the late 1970s, and finally stabilized in the late 1990s in response to management measures. The canary rockfish spawning stock biomass reached an estimated low in 2000, but has been increasing since that time. The estimated relative depletion level in 2005 is 5.7% in the base model and 11.4% in the alternate model.”
The rebuilding plan for canary rockfish includes a number of measures including a closure on directed fisheries, closing the continental shelf to trawling shallow of 137 m and non-retention in hook-and-line fisheries. Estimated catches for 2004 were less than 38 t, which include estimates of discarding. The principal monitoring tool for this population, the U.S. triennial survey, is now conducted annually instead of the previous triennial frequency.
Catches from this population dropped from around 1500 t/yr in the mid-1990s to 30-50 t/yr in 2003-2004. Catches were usually above 3 000 t/yr in the 1980s (maximum 5 400 t in 1982) and totalled 150 000 t over the entire time period covered by the assessment (1916 to the present). Unfished biomass was estimated at 35 000 t at the beginning of the period of the assessment (1916), with a steady decline to the present; biomass was around 15 000 t in the 1970s (Methot and Stewart 2005).
There is a swept area survey conducted in southeastern Alaska but too few canary rockfish are captured to infer trends in abundance (Mark Wilkins, pers. comm.).
Page details
- Date modified: