ICES Journal of Marine Science: Journal du Conseil Advance Access originally published online on July 11, 2007
ICES Journal of Marine Science: Journal du Conseil 2007 64(8):1512-1516; doi:10.1093/icesjms/fsm096
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
A strategy for developing scientific sampling tools for fishery-independent surveys of estuarine fish in New South Wales, Australia
1 NSW Department of Primary Industries, Cronulla Fisheries Research Centre of Excellence, PO Box 21, Cronulla, NSW 2230, Australia
2 Centre for Research on Ecological Impacts of Coastal Cities, Marine Ecology Laboratories A11, University of Sydney, NSW 2006, Australia
Correspondence to D. Rotherham: tel: +61 2 9527 8411; fax: +61 2 9527 8576; e-mail: douglas.rotherham{at}dpi.nsw.gov.au
Rotherham, D., Underwood, A. J., Chapman, M. G., and Gray, C. A. 2007. A strategy for developing scientific sampling tools for fishery-independent surveys of estuarine fish in New South Wales, Australia. – ICES Journal of Marine Science, 64: 1512–1516.The limitations of using fishery-dependent data, i.e. from commercial and recreational fisheries to assess harvested stocks of fish and invertebrates, are well known. Increasingly, fishery-independent surveys are used to validate data from fishery-dependent sources and to provide indices of recruitment and broader ecological information about species not normally retained in fishing operations. Any large-scale, long-term, fishery-independent study must develop sampling gear and designs that are standardized, representative, optimal with respect to the quantity and structure of catch, and replicated over relevant spatial and temporal scales. We present a strategy for achieving appropriate sampling designs. This involves: (i) identifying suitable sampling gears for target species; (ii) testing different configurations of gear and sampling practices to ensure that samples are optimal, representative, and cost efficient; (iii) understanding scales of spatial and temporal variability; and (iv) cost–benefit analyses to optimize replication. Examples of this strategy are illustrated, with brief considerations of the values of pilot research in developing fishery-independent sampling.
Keywords: cost–benefit analysis, fishery-independent survey, pilot study, spatial and temporal variation, standardized sampling
Received 30 August 2006; accepted 27 May 2007; advance access publication 11 July 2007.
 |
Introduction |
|---|
 Top Introduction Example of the strategy:...Step 1: identifying suitable...Step 2: testing different...Step 3: understanding scales...Step 4: cost-benefit analyses...ConclusionsReferences |
|---|
Fishery-independent surveys of fisheries resources play an important role in assessment and management of populations of fish and invertebrates (Gunderson, 1993; Pennington and Stromme, 1998). Research surveys are used to calibrate stock assessment models based on commercial catches (i.e. fishery-dependent data) and to provide empirical, independent checks of populations (Kline, 1996; Pennington and Stromme, 1998). Fishery-independent surveys are increasingly important: (i) to monitor aquatic resources where fishing has been modified as part of management; and (ii) to obtain better scientific assessments consistent with principles of ecologically sustainable development (ESD).
Fishery-independent data are often preferred over fishery-dependent data for monitoring the status of harvested populations because: (i) sampling is randomized rather than being concentrated where populations are (or are thought to be) most abundant; (ii) potentially, they provide more representative data on the entire size range of populations, rather than just retained components; (iii) there is no reliance on fishers reporting their catches and effort accurately; (iv) methodologies remain consistent over time; and (v) data can be collected on species not usually retained in commercial and recreational fisheries. Nevertheless, fishery-independent surveys that use inappropriate sampling gear or poorly designed sampling would also provide biased, inaccurate, and imprecise data.
Developing reliable, robust, and cost-effective sampling requires pilot studies to test specific hypotheses about the design and deployment of sampling gear (Andrew and Mapstone, 1987). Many experimental studies have: (i) compared methods of sampling fish or invertebrates (e.g. Guest et al., 2003; Olin and Malinen, 2003); (ii) tested the effects of biotic and abiotic factors on the performance of sampling gear (e.g. Acosta, 1994; Misund et al., 1999; Petrakis et al., 2001); and (iii) measured spatial and temporal variation in numbers of organisms across hierarchical scales (e.g. Morrisey et al., 1992a, b).
There are few examples, however, of necessary pilot work being done before a large-scale or long-term fishery-independent survey (Kennelly, 1989; Kennelly et al., 1993). We reviewed: (i) previous fishery-independent studies that have used pilot experiments to develop and optimize methods and designs; and (ii) key literature on surveys of fisheries resources (e.g. Gunderson, 1993) and design and analyses of ecological experiments (e.g. Andrew and Mapstone, 1987; Underwood, 1997). Many surveys were not consistent in their approaches to survey design. Therefore, we bring the elements together in a strategy for this type of preliminary work, following the approach advocated by Kennelly et al. (1993). The strategy involves: (i) identifying suitable gear to sample target species; (ii) testing configurations of gear and sampling practices to ensure that samples are optimal, representative, and cost efficient; (iii) understanding spatial and temporal variability; and (iv) cost–benefit analyses to determine optimal levels of replication (Figure 1).
|
|
Example of the strategy: developing fishery-independent surveys |
|---|
|
TopIntroductionExample of the strategy:...Step 1: identifying suitable...Step 2: testing different...Step 3: understanding scales...Step 4: cost-benefit analyses...ConclusionsReferences |
|---|
Estuaries in New South Wales (NSW), Australia, support commercial, recreational, and indigenous fisheries. Of 100 species landed from these estuaries, fewer than ten account for
80% of commercial catches. Stocks in NSW are currently assessed using fishery-dependent data, including catch and effort information supplied by the commercial sector, biological sampling of some species from commercial landings (e.g. Gray et al., 2002), and sporadic recreational creel surveys (e.g. Steffe and Chapman, 2003). The limitations of such data have led to increasing use of fishery-independent surveys to provide better information about the ecology and status of fisheries resources in NSW. A collaboration between the NSW Department of Primary Industries, the Centre for Research on Ecological Impacts of Coastal Cities at the University of Sydney, and the Fisheries Research and Development Corporation (FRDC 2002/059) is developing methods, procedures, and analyses for these surveys, so that more rigorous sampling designs can be implemented. Although commercial and scientific sampling gears are available, they are typically designed to be size-selective and species-specific. Often, such gears need to be modified to sample wider size ranges and greater diversities of fish. Many types of fish are targeted in estuaries in NSW; no single method can effectively estimate relative abundances and population structures for all species. Rather, a complementary suite of mobile (e.g. trawls) and static (e.g. gillnets) gears is required to estimate abundances, lengths, sex and age composition, reproduction, recruitment dynamics, etc.
We consider how the above strategy is being used in fishery-independent surveys of estuarine fish stocks in NSW. Surveys use several methods; we only illustrate one, a multimesh gillnet made of panels with different mesh sizes, designed to catch many species of different sizes and morphologies. Similar pilot work is being done with other methods (e.g. trawls).
|
Step 1: identifying suitable sampling gears for target species |
|---|
|
TopIntroductionExample of the strategy:...Step 1: identifying suitable...Step 2: testing different...Step 3: understanding scales...Step 4: cost-benefit analyses...ConclusionsReferences |
|---|
The first step was an experimental comparison of the utility and efficiency of multimesh gillnets and trammelnets. These gears were chosen because they are used widely in local fisheries. Much local knowledge and preliminary experiments not described here demonstrated that night-time sampling was less variable and more representative of size classes and diversity of species of fish than daytime sampling. All sampling described here was done at night. The gears differ: gillnets comprise a single panel of netting; trammelnets have two large-mesh panels enclosing a loosely hung centre panel of smaller mesh.
Replicate multimesh gillnets and trammelnets, each comprising five 30-m long panels of 38-, 54-, 70-, 90-, and 100-mm stretched mesh openings, were used in a NSW barrier estuary to test the hypothesis that catches of fish would differ between net types and mesh sizes (Gray et al., 2005). Analyses showed no statistically significant differences between the two types of net in compositions and structures of assemblages, abundance, or diversity of catches. Greater precision of catch per unit effort, ease of use, and smaller sampling effort made the multimesh gillnet the superior method.
|
Step 2: testing different configurations of gear and sampling practices |
|---|
|
TopIntroductionExample of the strategy:...Step 1: identifying suitable...Step 2: testing different...Step 3: understanding scales...Step 4: cost-benefit analyses...ConclusionsReferences |
|---|
The next step was to determine the most appropriate configuration and period of soak (i.e. length of time the gear is fished) and time of setting (i.e. time of night the gear is deployed). Experiments tested the hypotheses that catches and catch rates of fish were different between soak (1, 3, and 6 h) and setting times (18:00, 22:00, and 03:00), and net lengths (20-, 50-, and 120-m panels).
Univariate and multivariate procedures revealed that 20-m panels soaked for 1 h at any time of the night were optimal (in terms of catch and efficiency of sampling) and the best representative strategy for sampling populations, assemblages, and sizes of fish (details in Rotherham et al., 2006). The benefits of shorter soak times include greater replication, smaller costs, and potentially lower fish mortality (because the catch is processed and released as the gear is retrieved).
|
Step 3: understanding scales of variability |
|---|
|
TopIntroductionExample of the strategy:...Step 1: identifying suitable...Step 2: testing different...Step 3: understanding scales...Step 4: cost-benefit analyses...ConclusionsReferences |
|---|
Spatial and temporal variation in populations and assemblages of estuarine fish were examined using hierarchical sampling designs and nested analysis of variance (e.g. Morrisey et al., 1992a, b; Underwood, 1997). These analyses provide information about variation at different scales and the estimates of variances for cost–benefit analyses (Step 4). Here, one experiment on spatial and one on temporal variation were designed.
Experiment 1: spatial variation
Experiments investigating spatial heterogeneity in the abundance of organisms across hierarchical scales are relatively common (e.g. Green and Hobson, 1970; Kennelly, 1989), and the problems of spatial pseudoreplication (Hurlbert, 1984) are generally well understood. We examined patterns of variation of fish sampled at night using multimesh gillnets at several spatial scales at two depths (shallow: < 2 m; deep: 4–8 m) in an estuary. These depths were chosen as "candidates" for standardizing depth in future sampling; they were not chosen for detailed comparisons of depth gradients. The design incorporated spatial scales including randomly chosen zones (areas separated by 2–20 km) within estuaries, randomly chosen sites separated by at least 1 km within each zone, and replicate gillnets separated by 50–100 m (Figure 2). Fish abundances were hypothesized to differ at each spatial scale. To provide greater generality, the experiment was done in two large coastal lakes (Lake Macquarie and St Georges Basin), which are relatively well mixed and have no large salinity gradients or tidal ranges.
|
Components of variation were calculated from separate nested analyses of variance for each estuary and depth. For example, for tailor (Pomatomus saltatrix) in deep areas of Lake Macquarie, variability among replicates was greater than among sites or zones that were very similar (Table 1). In contrast, for tarwhine (Rhabdosargus sarba) in shallow areas, variability among replicates was similar to that among sites, but variation among zones was less (Table 1).
|
For most species, some components of variance were negative (Table 1), requiring pooling procedures (Fletcher and Underwood, 2002) or use of residual maximal likelihood (Robinson, 1987), which does not allow negative estimates. The two procedures give identical results when, as here, all levels of replication are balanced (Fletcher and Underwood, 2002). Negative components of variation at scales of sites or zones identifies that numbers of fish are less variable from site to site or zone to zone than among replicate nets. Negative sources of variation are pooled (as done for zones for flat-tail mullet, Liza argentea, in Table 1). Pooled components do not need to be sampled, because there is no variation in numbers of fish in this case, from zone to zone.
Negative components of variation were more common on the largest scale, so zones were often pooled with sites and re-analysed (Table 1). Most negative components then disappeared. There was a general pattern of more variation between replicate nets than between sites when components were scaled against the residual (to assess the magnitude of variation of each factor relative to the same reference, rather than relative to each other; see Underwood, 1997). The results indicate that it is unnecessary to include zones in future sampling; more effort should be put into sampling replicate nets and sites in these estuaries.
Experiment 2: temporal variation
Short-term variation (e.g. day-to-day and week-to-week) potentially confounds comparisons across longer scales (e.g. month-to-month and season-to-season). Differences from one time of sampling to the next cannot be interpreted as being associated on any larger scale (e.g. season-to-season), unless it has been demonstrated that differences on smaller scales (i.e. daily or weekly) are not as large (Morrisey et al., 1992b). The only alternative is the costly one of measuring variation at the larger scales (month-to-month or season-to-season) many independent times.
A second experiment was done to examine variation in populations and assemblages of estuarine fish across weeks, months, and seasons, again at two depths (shallow: < 2 m; deep: 4–8 m) in a coastal lake (St Georges Basin). Sampling also incorporated two spatial scales within the estuary: two sites 1 km apart nested in each of two zones (2–20 km apart; Figure 3) to measure spatiotemporal interactions and to provide greater generality. We did not attempt to measure variation among the nights sampled, because it takes several nights (in this case four) to sample all four sites across the two zones. Sites were sampled for two consecutive weeks, in 2 months in two consecutive "seasons" (July/August 2005, winter; October/November 2005, spring) to test the hypotheses that abundances of fish vary at each temporal scale.
|
For frequently abundant species (occurring in > 25% of samples), we analysed three factors (seasons, months, and weeks) separately for each site and depth, using nested analyses, then extracted the components of variation. As explained above, "zones" were not included in the model; instead, sites were treated as replicated experiments and analysed separately, giving four independent estimates of each component of variation. For most analyses, components of variation were negative at one or more scales, most frequently at the scales of weeks and months. Therefore, the data were pooled as four times of sampling in each season and re-analysed (Table 2).
|
Components of variation for species at deep sites were generally larger for seasons than for times within seasons. This pattern, however, was not consistent for the shallow sites. The only consistent pattern was that the residual variation (which is a spatial component) was greater than any temporal scale for every species (Table 2).
Fishery-independent surveys are often time-consuming, labour-intensive, and expensive. Because of these limitations, multiple sites and estuaries often cannot be sampled on the same night or even the same week. Therefore, any spatial comparisons are potentially confounded by time, because most sites (and estuaries) are sampled on different nights or in different weeks. Here, for most species, temporal variation was small compared with spatial variation. Therefore, as long as several sites within an estuary are sampled within a season, it is reasonable to interpret observed differences among sites as representing spatial rather than temporal variation. Further research, however, is needed to test whether similar patterns are observed over temporal scales longer than a few months.
|
Step 4: cost–benefit analyses to determine optimal levels of replication |
|---|
|
TopIntroductionExample of the strategy:...Step 1: identifying suitable...Step 2: testing different...Step 3: understanding scales...Step 4: cost-benefit analyses...ConclusionsReferences |
|---|
The final step in the strategy uses estimates of spatial and temporal variance (Step 3) to do cost–benefit analyses (e.g. Underwood, 1997). These well-known techniques allow determination of optimal levels of spatial and temporal replication (in terms of minimizing the imprecision of estimated means, the benefit) given restrictions of time, money, or both (the cost). Cost–benefit analyses for the present example are still in progress and will form the basis of a future publication. Applied examples focusing on marine biota, however, can be found in Green and Hobson (1970), Kennelly (1989), and Kennelly et al. (1993).
An example is shown for dusky flathead (Platycephalus fuscus; Table 2). As described above, weeks and months were pooled. Approximate costs are used solely to illustrate the analyses (costs are in Table 2). It was decided to optimize sampling, based on the precision of sampling, i.e. the standard error around each seasonal estimated mean number of fish. For example, 20% of the mean was chosen as the standard error. The number of samples, n, was then calculated from standard cost–benefit equations. The number of times, t, was determined using the standard error [see Underwood (1997) for all methods].
Of course, a different choice of precision would influence t. The value 20% is a compromise between excessive imprecision and excessive cost. In reality, precision in any sampling programme would be greater because it would be based on sampling from several seasons, increasing the degrees of freedom in estimated standard error, and reducing confidence intervals around estimates of means. For this example, optimal allocations of sampling in any season are to take n = 7 replicates. There should be four times of sampling in each season (Table 2).
Inevitably, different optimal strategies would result for different species, each with different patterns of spatial or temporal variance. Part of the strategy for surveys, therefore, will be to determine appropriate compromises and to understand how to "weight" imprecision for different species in accordance with their importance, commercial value, etc. This will be discussed elsewhere with details of the outcome of analyses for the fisheries in NSW estuaries.
|
Conclusions |
|---|
|
TopIntroductionExample of the strategy:...Step 1: identifying suitable...Step 2: testing different...Step 3: understanding scales...Step 4: cost-benefit analyses...ConclusionsReferences |
|---|
Using appropriately designed experiments to identify the most useful gear and to get estimates of variances at relevant spatial and temporal scales provides the essential information for planning fishery-independent surveys of fisheries resources. Optimal allocation of sampling units as replicates at different scales can then be done using standard methods of experimental design. Using a strategy designed to integrate knowledge of the most useful gear in the most effective sampling designs allows more reliable information from sampling that is independent of the fisheries themselves.
|
Acknowledgements |
|---|
This work is part of a cooperative study between the Wild Fisheries Program of the NSW Department of Primary Industries, the Centre for Research on Ecological Impacts of Coastal Cities at the University of Sydney, and the Australian Fisheries Research and Development Corporation (Project 2002/059). Sampling was done under NSW Agriculture Animal Care and Ethics approval 2003/019. We thank Daniel Johnson, Lachlan Barnes, and Paul Lokys for assistance with fieldwork, Matt Broadhurst for advice on gear, and anonymous referees for helpful comments.
|
References |
|---|
|
TopIntroductionExample of the strategy:...Step 1: identifying suitable...Step 2: testing different...Step 3: understanding scales...Step 4: cost-benefit analyses...ConclusionsReferences |
|---|
-
Acosta A. R. Soak time and net length effects on catch rate of entangling nets in coral reef areas. Fisheries Research (1994) 19:105–119.[CrossRef][Web of Science]
Andrew N. L., Mapstone B. D. Sampling and the description of spatial pattern in marine ecology. Oceanography and Marine Biology. An Annual Review (1987) 25:39–90.
Fletcher D. J., Underwood A. J. How to cope with negative estimates of components of variance in ecological field studies. Journal of Experimental Marine Biology and Ecology (2002) 273:89–95.[CrossRef][Web of Science]
Gray C. A., Gale V. J., Stringfellow S. L., Raines L. P. Variations in sex, length and age compositions of commercial catches of Platycephalus fuscus (Pisces: Platycephalidae) in New South Wales, Australia. Marine and Freshwater Research (2002) 53:1091–1100.[CrossRef][Web of Science]
Gray C. A., Jones M. V., Rotherham D., Broadhurst M. K., Johnson D. D., Barnes L. M. Utility and efficiency of multi-mesh gill nets and trammel nets for sampling assemblages and populations of estuarine fish. Marine and Freshwater Research (2005) 56:1077–1088.[CrossRef][Web of Science]
Green R. H., Hobson K. D. Spatial and temporal structure in a temperate inter tidal community, with special emphasis on Gemma gemma (Pelecypoda: Mollusca). Ecology (1970) 51:999–1011.[CrossRef]
Guest M. A., Connolly R. M., Loneragan N. R. Seine nets and beam trawls compared by day and night for sampling fish and crustaceans in shallow seagrass habitat. Fisheries Research (2003) 64:185–196.[CrossRef][Web of Science]
Gunderson D. R. Surveys of Fisheries Resources. (1993) New York: John Wiley and Sons, Inc. 248.
Hurlbert S. H. Pseudoreplication and the design of ecological field experiments. Ecological Monographs (1984) 54:187–211.[CrossRef][Web of Science]
Kennelly S. J. Effects of soak-time and spatial heterogeneity on sampling populations of spanner crabs Ranina ranina. Marine Ecology Progress Series (1989) 55:141–147.[CrossRef][Web of Science]
Kennelly S. J., Graham K. J., Montgomery S. S., Andrew N. L., Brett P. A. Variance and cost–benefit analyses to determine optimal duration of tows and levels of replication for sampling relative abundances of species using demersal trawling. Fisheries Research (1993) 16:51–67.[Medline]
Kline L. Fisheries research under fire: more than just a money issue. Fisheries (1996) 21:4–5.
Misund O. A., Luyeye N., Coetzee J., Boyer D. Trawl sampling of small pelagic fish off Angola: effects of avoidance, towing speed, towing duration, and time of day. ICES Journal of Marine Science (1999) 56:275–283.
Morrisey D. J., Howitt L., Underwood A. J., Stark J. S. Spatial variation in soft-sediment benthos. Marine Ecology Progress Series (1992) a 81:197–204.[CrossRef][Web of Science]
Morrisey D. J., Underwood A. J., Howitt L., Stark J. S. Temporal variation in soft-sediment benthos. Journal of Experimental Marine Biology and Ecology (1992) b 164:233–245.[CrossRef][Web of Science]
Olin M., Malinen T. Comparison of gillnet and trawl in diurnal fish community sampling. Hydrobiologia (2003) 506–509:443–449.
Pennington M., Stromme T. Surveys as a research tool for managing dynamic stocks. Fisheries Research (1998) 37:97–106.[CrossRef][Web of Science]
Petrakis G., MacLennan D. N., Newton A. W. Day–night and depth effects on catch rates during trawl surveys in the North Sea. ICES Journal of Marine Science (2001) 58:50–60.
Robinson D. L. Estimation and use of variance components. The Statistician (1987) 36:3–14.[CrossRef]
Rotherham D., Gray C. A., Broadhurst M. K., Johnson D. D., Barnes L. M., Jones M. V. Sampling estuarine fish using multi-mesh gill nets: effects of panel length and soak and setting times. Journal of Experimental Marine Biology and Ecology (2006) 331:226–239.[CrossRef][Web of Science]
Steffe A. S., Chapman D. J. A survey of daytime recreational fishing during the annual period, March 1999 to February 2000, in Lake Macquarie, New South Wales. (2003) NSW Fisheries Final Report Series, 52.
Underwood A. J. Experiments in Ecology: Their Logical Design and Interpretation Using Analysis of Variance. (1997) Cambridge: Cambridge University Press. 504.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  D. Rotherham, M. K. Broadhurst, C. A. Gray, and D. D. Johnson Developing a beam trawl for sampling estuarine fish and crustaceans: assessment of a codend cover and effects of different sizes of mesh in the body and codend ICES J. Mar. Sci., May 1, 2008; 65(4): 687 - 696. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||