© 2005 International Council for the Exploration of the Sea
Development and evaluation of catch per unit effort indices for southern blue whiting (Micromesistius australis) on the Campbell Island Rise, New Zealand
a National Institute of Water and Atmospheric Research PO Box 893, Nelson, New Zealand
b National Institute of Water and Atmospheric Research PO Box 14-901, Kilbirnie, Wellington, New Zealand
*Correspondence to S. Hanchet: tel: +64 35457739; fax: +64 35481716. e-mail: s.hanchet{at}niwa.co.nz.
This paper develops standardized commercial cpue indices for a highly aggregated spawning fishery in New Zealand waters, and verifies the indices using fishery-independent data. Indices were calculated for all vessels using three different measures of effort, and for vessel subsets based on processing type (surimi and dressed), and relative experience in the fishery. Trends in cpue were consistent with trends in fishery-independent acoustic surveys, age composition of the commercial catch, and recent stock assessment results. In particular, the cpue indices tracked the more than fourfold increase in abundance from 1993 to 1996 associated with the recruitment of the strong 1991 year class, and the decline in relative abundance as this year class was fished down. Despite this being a highly aggregated spawning fishery, there was little evidence for hyperstability. There were also significant differences in fishing strategies of the fleets between periods of high and low fish abundance.
Keywords: fishery-independent, fishing strategies, southern blue whiting, standardized cpue
Received 24 June 2004; accepted 16 February 2005.
 |
Introduction |
|---|
 Top Introduction MethodsResults and discussionReferences |
|---|
Despite their well-documented shortcomings, commercial catch per unit effort (cpue) indices of relative abundance continue to be used for monitoring a large number of fisheries worldwide (Harley et al., 2001). Reasons why cpue may not be proportional to abundance have been investigated by simulation (Gillis and Peterman, 1998), and through examination of empirical data (Rose and Leggett, 1991). The most common form of non-proportionality involves cpue remaining high while abundance declines, which is known as hyperstability (Hillborn and Walters, 1992). This is believed to be a particular problem for highly aggregated fisheries, but is also a problem for a wide range of other fisheries (Harley et al., 2001). Many stocks are monitored using cpue in New Zealand, although few have been validated through fishery-independent data (Dunn et al., 2000). A recent cpue analysis provided the opportunity to test the relationship between cpue and abundance for the southern blue whiting fishery in New Zealand waters.
Southern blue whiting (Micromesistius australis Norman) is a small southern hemisphere gadoid closely related to the blue whiting (Micromesistius poutassou Risso) found in the northern hemisphere. Southern blue whiting is found in commercial quantities off both coasts of South America, around the Falkland Islands, and in waters around southern New Zealand (Bailey, 1982). Within New Zealand waters they aggregate to spawn in August and September on distinct spawning grounds on the Campbell Island Rise, Bounty Platform, Pukaki Rise, and Auckland Islands Shelf (Figure 1). Since 1986, commercial fishing vessels have targeted these spawning aggregations with large midwater trawls, and total catches from New Zealand waters have averaged 32 000 t, with 11 000–35 000 t (60% on average) coming from the Campbell Island Rise spawning fishery (Hanchet et al., 2003a). The fishery has been under quota since 1993.
|
Unstandardized and standardized commercial cpue analyses have been carried out for the Campbell Island Rise spawning fishery since the early 1990s, but there has always been concern that the cpue may not be monitoring abundance (Hanchet et al., 2003a). This is because of the highly aggregated nature of the fishery, and the associated difficulty in locating and maintaining contact with the highly mobile spawning schools in some years. Therefore, standardized cpue analyses have not been updated since 1998, and the indices have not been included in stock models or assessments since 1997 (Hanchet et al., 2003a). During the past decade, a time-series of acoustic indices has also been developed for this stock (Hanchet et al., 2003b). This time-series has been used in conjunction with catch-at-age data to assess the stock (e.g. Hanchet et al., 1998, 2003a). The main aim of the current study was to determine cpue indices for the Campbell Island Rise stock, and, if possible, to verify these with fishery-independent abundance data. A secondary aim was to determine how fleet fishing strategies changed between periods of low and high abundance.
|
Methods |
|---|
|
TopIntroductionMethodsResults and discussionReferences |
|---|
Data selection and definition of the fishery
Tow-by-tow data were extracted from the Ministry of Fisheries FSU (Fisheries Statistics Unit) database from 1986 to 1989, and from the TCEPR (Trawl, Catch, Effort and Processing Returns) database from 1989 to 2002. Data analysis was restricted to the years 1986–2002, because the fishery changed substantially in 1986, making comparison with earlier years difficult.
The area defined for the analysis is based on the boundaries used for managing the stock (Hanchet, 1999). The fishery operated mainly to the north and east of the shelf edge around Campbell Island (Figure 1). This area was split into four subareas (1–4) based on prior knowledge of the fishery: spawning aggregations are typically fished in subareas 2 and 4, whereas prespawning aggregations and postspawning fish are often found in subareas 1 and 3, respectively (Chatterton, 1996). The spawning southern blue whiting fishery is of relatively short duration, with most of the catch taken from 1 September to 7 October, so this time period was chosen for the analysis.
Description of the variables
The data extracted for each tow included time, location, catch details, and fishing gear parameters, and details of the variables included in the models are listed in Table 1. Other explanatory variables were offered to the models, but were not significant. Day in season was defined as the difference between the date of fishing and the date of the onset of spawning (defined as when 10% of the females were running ripe) in each year (after Chatterton, 1996). Catch rates vary both diurnally and with height of the groundrope above the bottom, so start and end times of tow were each grouped into 4-h time categories.
|
Vessel was included as a categorical variable in the analysis, following Dunn (2002). Other vessel characteristics were also included. Vessel experience represented the cumulative number of years that each vessel had completed at least 5 tows in the fishery. Preliminary examination of the data showed that there were two main fleets: a Japanese surimi fleet and a predominantly ex-Soviet dressed fleet. Many vessels that participated in the fishery were involved for a short period or conducted a limited number of tows. Core vessels were defined for each processing method as those that had participated in the fishery for at least 3 years, with a minimum of 5 completed tows per year.
Commercial catch and effort data commonly include many errors, such as missing or invalid codes, or implausible data. Hence, all data were checked for errors before analysis.
Model structure
Effort can be measured in several ways, as catch per tow, catch per hour, or catch per day. Initial analyses showed that model results were similar for the different measures, so we chose catch per tow, and tested the sensitivity of the final results to the measure of cpue used.
Generalized linear models were used for modelling ln-transformed cpue as a function of explanatory variables (see Table 1), assuming normal error distributions. Estimates of relative year effects provided abundance indices, following the procedures of Gavaris (1980) and Vignaux (1996). We used the Proc GLM (General Linear Modelling) procedure of the SAS statistical software (SAS, 1999). Explanatory variables were progressively added to the model in order of the percentage of variance explained, until <1% improvement was seen. Only 6% of the tows made zero catches of southern blue whiting, so following Dunn (2002), we used only positive tows. Previous work showed that the year effects were relatively insensitive to the inclusion of zero tows (NIWA, unpublished data). Interaction terms were not considered.
Some continuous explanatory variables may have a non-linear relationship with cpue. Variables entered into the model as polynomials if the quadratic or cubic transformations explained >1% of the variability in the model. Residual plots were examined for evidence of significant departures from model assumptions, and to determine the fit of the regression model to the data.
Model validation and verification
We compared the indices resulting from the cpue analysis with a time-series of fishery-independent acoustic survey indices, with the age composition of the commercial catch, and with estimates of biomass derived from a recent stock assessment. During the past decade, a time-series of acoustic indices has been developed for this stock (Hanchet et al., 2003b). Surveys were carried out during the spawning season, when the fish school into large, dense, easily identifiable, mono-specific aggregations (Hanchet et al., 2003b), annually from 1993 to 1995, then biannually from 1998 to 2002. The surveys generally follow a Jolly and Hampton (1990) design, with a random starting point and parallel transects placed at right angles to the depth contours, but they have also incorporated various adaptive strategies. Adult biomass estimates for most surveys had CVs in the range 17–36%, although the 2002 survey had a very high CV of 68%. More details of the design and analysis of recent surveys are given in Hanchet and Grimes (2001) and Hanchet et al. (2003b).
Annual stock assessments have been carried out for the Campbell Island Rise stock since 1991. Although cpue indices were used to tune some of the earlier stock assessments (e.g. Hanchet et al., 1998), more recent stock assessments have used only the acoustic biomass estimates as indices of abundance (Hanchet et al., 2003a). We used the results of the most recent assessment carried out in 2003 (Hanchet et al., 2003a) for the comparison with cpue indices. This assessment was carried out using Bayesian estimation, with the NIWA modelling program CASAL v1.02 (Bull et al., 2002). In this assessment an age-structured population model was fitted to a time-series of catch-at-age data from the commercial fishery for the period 1979–2002, and to the time-series of acoustic indices from 1993 to 2002. Because this was the first assessment using CASAL, a number of initial runs were made to explore the sensitivity of the results to changes in the various model assumptions and priors. We compared the results of the cpue indices with the estimates of mid-season spawning-stock biomass (SSB) from the population model for the base case assessment. The estimates of SSB were the medians of the posterior estimates (see Hanchet et al., 2003a, for details). To make the data sets comparable, the cpue indices and estimates of SSB were each centred on their means.
For analysis and interpretation of the indices, it is generally assumed that there is a simple direct relationship between cpue and abundance. However, the fishery for southern blue whiting is a highly aggregated spawning one, and fishers actively search for the dense aggregations with sonar, then target those marks. There is, therefore, potential for the cpue/abundance relationship to be hyperstable (Dunn et al., 2000; Harley et al., 2001). To examine this, we carried out a simulation exercise using parametric bootstrap regression, and assuming a non-linear relationship between cpue and SSB of the form:
|
|
Understanding fishing strategies, the decisions that fishers make in deciding when and where to fish, are important as part of a general understanding of the fishery (Hillborn and Walters, 1992), and it may help identify why cpue is monitoring abundance. To examine fishing strategies, we calculated correlations among the fleet summary data and estimates of SSB.
|
Results and discussion |
|---|
|
TopIntroductionMethodsResults and discussionReferences |
|---|
Fishery description
The annual number of tows in the analysis fluctuated between 201 and 1183. Both surimi and dressed vessels fished in all years, but the relative importance of the two fleets varied between years. Although there have been 23 different surimi vessels in the fishery, the 13 core vessels accounted for 90% of the surimi catch. Both mean duration of fishing and mean catch per tow have more than doubled since 1986. This coincided with a change to smaller surimi vessels in the early 1990s. Of the total of 117 dressed vessels, only 25 were core vessels, and these accounted for about 60% of the dressed catch. In contrast to the surimi vessels, tow duration and mean catch per tow of dressed vessels remained relatively constant over time. There has been an increase in vessel experience in both fleets since 1995.
Cpue indices
The variables vessel, year, end time of tow, and day in season were significant in the all vessels catch per tow model, and explained 32% of the variance (Table 2). The vessel variable (factor) alone explained >25% of the data variability and suggested a wide range of relative fishing powers, with expected catch rates ranging from about 3 to 25 t per tow. The end time of tow variable (factor) showed that catch rates were maximized during the day, while the continuous day in season variable indicated that cpue was maximized at peak spawning. The year effects generally declined from 1986 to 1992, increased to 1996, then declined again to 2002 (Figure 2). The diagnostic normal quantile–quantile plots showed a deviation from the normal distribution of the residuals at the lower and upper ends, indicating that very small and very large values of catch rate are not well predicted (Figure 3).
|
|
|
Similar variables were significant in the all vessels catch per hour model but explained slightly less of the variance than catch per tow (Table 2). The variables vessel, number of tows per day, year, tow distance, and day in season entered the all vessels catch per day model, together explaining 45% of data variability (Table 2). Number of tows per day and tow distance had positive relationships with cpue in the catch per day model. The year effects from both catch per hour and per day models followed similar trends to the catch per tow model (see Figure 2).
Models were run on the all dressed, core dressed, all surimi, and core surimi data subsets. The year effects were almost identical between the all vessel and core vessel data subsets for a particular processing type, so only the results of core data subsets are presented. Variables entering the models and the variation explained by the models were again similar to the all vessels catch per tow model (Table 2). However, duration of tow and difference between headline and groundrope depth were significant for the core surimi vessels. The year effects for these models showed similar trends to the all vessel model (Figure 2). Regression diagnostics indicated that these models were a less satisfactory fit to the data than the all vessels models, and that the standard errors were high for the core surimi model (Table 3).
|
The three cpue models with different measures of effort had similar significant explanatory variables, similar r2 values, similar diagnostics, and very similar year effects. Vessel was the most important factor in most models, which is consistent with cpue analyses carried out on other fisheries (e.g. Dunn, 2002). The other variables that entered into the model were also significant in earlier cpue analyses of the fishery (Chatterton, 1996), and are consistent with our understanding of the fishery. cpue peaked as the fish started spawning, and also during daylight, when the fish are more highly aggregated. Both these variables were also significant in standardized cpue analyses of the related west coast South Island hoki spawning fishery (Dunn, 2002). The percentage of variation explained by the models was rather low, but it is in the range of similar cpue analyses (e.g. Large, 1992; Vignaux, 1996).
Comparison of the cpue indices with fishery-independent data
Although a number of model runs and sensitivity analyses were carried out in the 2003 assessment, the SSB trajectories were almost identical when standardized to their means, so only the trajectory for one of the runs (base case 2) is shown (see Figure 2). The cpue indices show similar trends to the population model, with a decline from 1986 to 1992, an increase to 1996, and a decline to 2002. However, when examined in more detail, the relationship between the cpue models and the abundance over these three time periods appears to be subtly different. Over the period 1986–1992, the catch per tow and per day models appear to show some degree of resistance to the change in underlying abundance. The year effect reaches its minimum in 1992 in all cpue models, consistent with the population model and with the perception of the stock status at that time. In that year, more than 50 vessels were fishing the Campbell fishery for more than a month, and the total catch was just 14 000 t. The month before, about 35 of these vessels had caught >60 000 t of southern blue whiting on the Bounty Platform (NIWA, unpublished data).
A very strong year class (spawned in 1991) first recruited to the commercial fishery as 3-year olds in 1994 (Figure 4). This year class completely dominated the catch by both number and weight for the rest of the 1990s. It also led to a fourfold increase in SSB from 1993 to 1996, when it became fully recruited to the fishery (see Figure 2). None of the cpue models can match the large rate of increase in the abundance estimated by the population model, all the 1995 year effects being substantially less than would be expected. This could be a feature of the availability and distribution of the new recruits in relation to the fishing fleet. Most fishing in 1995 was in subarea 2, whereas a substantial number of the new recruits were farther south in subarea 4, as indicated by the 1995 acoustic survey results (NIWA, unpublished data). By 1996, most of the cpue indices had increased to the level of the stock abundance.
|
Stock abundance declined from 1996 to 2002 as the 1991 year class was fished down and no new strong year classes entered the fishery. The cpue indices tend to be more erratic over this period, showing varying levels of decline, but most end up close to the modelled abundance by 2002. Note that the stock assessment model was unable to fit the 2002 acoustic survey estimate, which had a high CV and may have overestimated the adult biomass (Hanchet et al., 2003a, b). A feature of all the cpue models is a marked peak in the year effect in 1999, similar to those seen in 1989 and 1991. The reason for this is unclear but does not appear to be an artefact of the analysis, because the number of tows carried out in that year were about average and the raw (unstandardized) cpue for that year was also high.
For the analysis and interpretation of the indices, we have assumed a simple direct relationship between cpue and abundance. This assumption was tested by modelling the relationship between cpue and abundance using a hyperstability parameter (Harley et al., 2001). Estimates of ß had a mean of 1.14, with 95% intervals of 0.83–1.57, and a median of 1.12 (Figure 5). There was therefore no evidence for a significant hyperstable relationship between SSB and cpue, in contrast to the strong evidence of hyperstability found in most fish stocks examined by Harley et al. (2001).
|
Fishing strategies
The presence of two fishing fleets allows opportunity to investigate fishing strategies between fleets and between periods of high and low abundance. Surimi vessels can process up to 200 t live weight of southern blue whiting per day, while dressed vessels can process about 50–80 t live weight per day. Vessels from both fleets use catch sensors in this fishery, so can haul their nets when the desired quantity of fish has been caught. Vessel skippers presumably want to maximize their daily catch (subject to processing limitations), and the simplest way to achieve this is by altering the number of tows per day and/or the tow duration.
In periods of high fish abundance (i.e. high model biomass), surimi vessels make significantly longer tows (Table 4, p < 0.05). The cpue indicators (catch per hour, per tow, and per day) are all positively correlated with stock abundance (p < 0.05). Therefore, the catching capacity of surimi vessels appears to be relatively unconstrained by their processing capabilities. In contrast, dressed vessels make significantly shorter and fewer tows when fish abundance is high, and only mean catch per hour is positively correlated with stock abundance (p < 0.05). This suggests that catching capacity is constrained by the processing capacity on those vessels, and that vessels are targeting a particular size of catch for each tow. It also bears out comments from fishing industry representatives that dressed vessels prefer smaller catches, ensuring better quality fish for processing.
|
In years of low stock abundance, dressed vessels simply increase tow duration to achieve their targeted catch per tow, and the number of tows per day to achieve their targeted daily catch. In contrast, surimi vessels make significantly shorter tows, and tend to make more of them. This is counter-intuitive, because in general there is a positive relationship between tow duration and catch per tow, so surimi vessels might be expected to make longer tows. However, this strategy could be a response to a change in fish distribution in periods of low abundance. It has been reported previously that fish have a more patchy distribution and that fish schools are smaller at times of lower abundance (Rose and Kulka, 1999). Therefore, reducing the length of the tow could reflect a smaller school size, whereas increasing the number of tows could reflect increased searching for other schools.
Clearly, for both fleets there is a strong correlation between catch per hour and abundance. The fact that we obtained the same results for a range of different models, and in particular for two different fleets with different fishing strategies, suggests a causal relationship between cpue and abundance. The next question to address is why catch per hour is proportional to abundance. Although it is an aggregated fishery, the underlying fish distribution must vary between years of high and low abundance. School size, fish density within the school, and distance between schools will presumably influence the strategies employed by the vessels to maximize their catch per hour. Examination of data recorded by Scientific Observers has shown that longer tows are achieved by making multiple U-turns through a fish school (NIWA, unpublished data). When fish abundance is low, either the fish densities decrease or the schools become so small that it is no longer worthwhile making multiple passes through the same school. The latter may be more likely, as several authors have shown that during periods of low abundance, fish generally have a smaller spatial extent (Rose and Kulka, 1999; Salthaug and Aanes, 2003). Moreover, as schools become smaller, interference competition between fishers is likely to increase (Gillis and Peterman, 1998), which may further reduce catch rates.
Future research should focus on fine-scale analysis of acoustic data to determine factors such as school size, fish density within the school, and distance between schools under conditions of high and low abundance. Research should also focus on the spatial and temporal fleet fishing dynamics, following Vignaux (1996) and Salthaug and Aanes (2003). This would help evaluate the effects of factors such as vessel competition and vessel movement on catch per hour. Simulations could then be carried out to determine under what conditions catch per hour would monitor abundance.
|
Acknowledgements |
|---|
We thank the Ministry of Fisheries for financial support under project code SBW2001/01, and two anonymous referees for comments on an early draft.
|
References |
|---|
|
TopIntroductionMethodsResults and discussionReferences |
|---|
-
Annala J. H., Sullivan K. J., Smith N. W. Mc. L., Griffiths M. H., Todd P. R., Mace P. M., Connell A. M. (2004) Reports from the fishery assessment plenary, April 2004: stock assessments and yield estimates (Unpublished report held in NIWA library, Wellington, NZ). (Comps.) 940 pp.
Bailey R.S. (1982) The population biology of blue whiting in the North Atlantic. Advances in Marine Biology 19:257–355.[Web of Science]
Bull B., Francis R. I. C. C., Dunn A., Gilbert D. J. (2002) CASAL (C++ algorithmic stock assessment laboratory): CASAL user manual v1.02.2002/10/21. NIWA Technical Report, 117. 199 pp.
Chatterton T. D. (1996) Catch per unit effort (cpue) analysis of the Campbell Island Rise southern blue whiting (Micromesistius australis) trawl fishery from 1986 to 1993. New Zealand Fisheries Assessment Research Document, 96/1 (Unpublished report held in NIWA library, Wellington). 26 pp.
Dunn A. (2002) Updated catch-per-unit-effort indices for hoki (Macruronus novaezelandiae) on the west coast South Island, Cook Strait, Chatham Rise, and sub-Antarctic for the years 1990 to 2001. New Zealand Fisheries Assessment Report, 2002/47. 51 pp.
Dunn A., Harley S. J., Doonan I. J., Bull D. (2000) Calculation and interpretation of catch-per-unit-effort (CPUE) statistics. New Zealand Fisheries Assessment Report, 2000/1. 44 pp.
Gavaris S. (1980) Use of a multiplicative model to estimate catch rate and effort from commercial data. Canadian Journal of Fisheries and Aquatic Sciences 37:2272–2275.
Gillis D.M. and Peterman R.M. (1998) Implications of interference among fishing vessels and the ideal free distribution to the interpretation of cpue. Canadian Journal of Fisheries and Aquatic Sciences 55:37–46.
Hanchet S.M. (1999) Stock structure of southern blue whiting (Micromesistius australis) in New Zealand waters. New Zealand Journal of Marine and Freshwater Research 33:599–610.[Web of Science]
Hanchet S. M., Dunn A., Stevenson M. L. (2003a) Southern blue whiting (Micromesistius australis) stock assessment for the Campbell Island Rise for 2003. New Zealand Fisheries Assessment Report, 2004/59. 42 pp.
Hanchet S. M. and Grimes P. J. (2001) Acoustic biomass estimates of southern blue whiting (Micromesistius australis) from the Campbell Island Rise and Pukaki Rise, September 2000. New Zealand Fisheries Assessment Report, 2001/58. 37 pp.
Hanchet S. M., Grimes P. J., Coombs R. F., Dunford A. (2003b) Acoustic biomass estimates of southern blue whiting (Micromesistius australis) for the Campbell Island Rise, August–September 2002. New Zealand Fisheries Assessment Report, 2003/44. 38 pp.
Hanchet S.M., Haist V., Fournier D. (1998) An integrated assessment of southern blue whiting (Micromesistius australis) from New Zealand waters using separable sequential population analysis. In Funk F., Quinn T.J., Heifetz J., Ianelli J.N., Powers J.E., Schweigert J.F., Sullivan P.J., Zhang C-I. (Eds.). Fishery Stock Assessment Models(University of Alaska, Fairbanks) Alaska Sea Grant College Program Report, AK-SG-98-01. 1037 pp.
Harley S.J., Myers R.A., Dunn A. (2001) Is catch-per-unit-effort proportional to abundance? Canadian Journal of Fisheries and Aquatic Sciences 58:1760–1772.
Hillborn R. and Walters C.J. (1992) Quantitative Fisheries Stock Assessment(Chapman & Hall, New York).
Jolly G.M. and Hampton I. (1990) A stratified random transect design for acoustic surveys of fish stocks. Canadian Journal of Fisheries and Aquatic Sciences 47:1282–1291.
Large P.A. (1992) Use of a multiplicative model to estimate relative abundance from commercial cpue data. ICES Journal of Marine Science 49:253–261.
Rose G. and Kulka D.W. (1999) Hyperaggregation of fish and fisheries: how catch-per-unit-effort increased as the northern cod (Gadus morhua) declined. Canadian Journal of Fisheries and Aquatic Sciences 56:Suppl. 1, 118–127.
Rose G.A. and Leggett W.C. (1991) Effects of biomass-range interactions on catchability of migratory demersal fish by mobile fisheries: an example of Atlantic cod (Gadus morhua). Canadian Journal of Fisheries and Aquatic Sciences 48:843–848.
Salthaug A. and Aanes S. (2003) Catchability and the spatial distribution of fishing vessels. Canadian Journal of Fisheries and Aquatic Sciences 60:259–268.
SAS. (1999) SAS/STAT User's Guide, Version 8(SAS Institute Inc., Cary, NC, USA) vols. 1, 2 and 3:.
Vignaux M. (1996) Analysis of vessel movements and strategies using commercial catch and effort data from the New Zealand hoki fishery. Canadian Journal of Fisheries and Aquatic Sciences 53:2126–2136.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||