Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 20
IMPROVING THE COLLECTION, MANAGEMENT, AND USE OF MARINE FISHERIES DATA 2 Summer Flounder: Review and Insights INTRODUCTION The committee examined the summer flounder assessments with two objectives in mind. The first was to address specific issues raised by Congress and stakeholders regarding the quality and nature of these assessments and to gain a direct understanding of the significance these issues have for managing summer flounder. The second objective was to provide examples to illustrate the broader issues associated with the collection and use of fisheries data that affect the science and management of other stocks. To focus these tasks, the committee elicited comments from representatives of the commercial and recreational fishing communities and environmental advocates concerning data collection and assessment methods applied to summer flounder. The committee reviewed the summer flounder data and the NMFS assessments and conducted its own independent assessments (Appendix C and Appendix D) and an examination of modeling assumptions (this chapter) to generate its findings. The committee provided opportunities at its two Washington, D.C. meetings for public input on summer flounder issues. Although more general issues were raised at the committee's third meeting in Seattle, Washington, some of the concerns raised there also had relevance to summer flounder. Commercial and recreational fishermen, scientists from the National Marine Fisheries Service (NMFS), representatives from the Mid-Atlantic Fishery Management Council, representatives from the Atlantic, Gulf, and Pacific States Marine Fisheries Commissions, and representatives from environmental groups made presentations to the committee at these meetings. Commercial fishermen expressed a variety of views that indicated a general lack of trust in the management of summer flounder and other marine fisheries. Many recreational fishermen and their representatives were invited to the committee's meetings. Unfortunately, the committee was unable to elicit significant input from this sector regarding the summer flounder assessments or other data issues. Although the issues raised by stakeholders are not all resolved in this report, the report does explore a number of the more significant and persistent issues in the hope that NMFS and the regional councils will take steps to fix problems and correct misperceptions. The committee's public sessions revealed
OCR for page 21
IMPROVING THE COLLECTION, MANAGEMENT, AND USE OF MARINE FISHERIES DATA many perceptions of some commercial fishermen that affect their beliefs about fisheries data and stock assessments. Many fishermen believe that fisheries science and fisheries assessments are not objective scientific exercises; instead, they reflect political agendas. There are several steps between a stock assessment and the final management advice. Many fishermen believe that precaution is built into each step of the assessment process, so that the final results are too conservative. Fishermen advocate that scientists should provide completely objective advice and that managers should build in precaution only at the final management stage. Regulations change too frequently and before it has been determined whether management objectives have been achieved. This is especially problematic in cases that result in changes in gear regulations, because fishermen need to undertake the expense of replacing or modifying their gear with each change. The annual changes in the mesh size requirements for summer flounder were noted as an example. The fishing gear and survey techniques used by NMFS in its fishery-independent surveys have remained relatively constant over time. Fishermen no longer use the same kind of gear and consequently believe that NMFS ' use of outdated fishing gear and fishing practices for surveys result in stock assessments that estimate smaller populations than the industry believes are realistic. Many fishermen question a number of assumptions about the biology and population dynamics used in specific stock assessments (e.g., in the case of summer flounder, this includes the assumed level of natural mortality and the assumption of constant selectivity for the older age groups). Whether or not these concerns are valid from a scientific perspective, management of marine fisheries will be more difficult if these perceptions of fishermen are not addressed. Because the National Research Council (NRC) was requested by Congress to focus on summer flounder, the committee examined the 1996 and 1999 summer flounder assessments and makes a number of recommendations in this chapter for improving assessments. It should be recognized, however, that NMFS cannot afford to conduct similar investigations for all the species it manages. Therefore, Chapter 4 highlights only those recommendations that should be applied broadly in fisheries data collection, management, and use in the United States. It is the responsibility of NMFS and the Mid-Atlantic and New England Fishery Management Councils to determine whether the steps recommended for summer flounder are a priority given the value of summer flounder compared to other fisheries. The lack of certainty in assumptions described below may hinder the summer flounder stock from recovering to its 1980s peak, although environmental factors may also be important deterrents to full stock recovery. SUMMER FLOUNDER ASSESSMENT ISSUES Here we examine the facts about summer flounder, determine the merits of the concerns, and recommend to Congress, NMFS, and the councils ways to correct problems and misperceptions. The committee investigated questions related to (1) the biology and population dynamics of summer flounder; (2) summer flounder sampling; and (3) the information content of the model and model assumptions currently in use (Box 2-1). In some cases, data were not available for the committee to investigate the issues. Some of the concerns about the scientific assessment and management of summer flounder are specific and are dealt with in the following sections. Others result from the fishing industry's lack of trust in and an alienation from the management process. Such alienation suggests that there is a wider problem of cooperation and outreach, which will be addressed in Chapter 4.
OCR for page 22
IMPROVING THE COLLECTION, MANAGEMENT, AND USE OF MARINE FISHERIES DATA BOX 2-1 Concerns About Summer Flounder Assessments Investigated by the Committee Questions related to the biology and population dynamics of summer flounder Do the summer flounder in waters north of Cape Hatteras comprise a unit stock of fish? What natural mortality rate is appropriate to use in summer flounder assessment models? Are there differences between the growth and mortality of male and female summer flounder (sexual dimorphism) and, if so, how do the differences affect the assessments? Questions related to summer flounder sampling What are the appropriate survey and commercial catchabilities for summer flounder? Do problems with determining the age of summer flounder discredit age-based assessments? Are effort data used appropriately and are the effects of effort changes incorporated properly? Is the observer program for summer flounder adequate? Can and should state surveys be standardized? Is the catch from recreational fishing estimated properly? Can precision of data be increased? Questions related to the information content of the model and model assumptions currently in use What information does each model structure require and how do these requirements relate to the information in the data? Additional things to consider when answering this question are (1) the weight the model gives to the different sources of information that come into it, (2) how uncertainty is expressed in the assessment output provided to managers, and (3) the level of precision managers can expect with the data currently available. QUESTIONS RELATED TO THE BIOLOGY AND POPULATION DYNAMICS OF SUMMER FLOUNDER Summer flounder (Paralichthys dentatus; also called fluke) are found primarily from Cape Fear, North Carolina, to Cape Cod, Massachusetts (see Figure C-1) on a variety of seafloor substrates, including mud and sand (Able and Kaiser, 1994; Packer and Hoff, 1999). Summer flounder can reach a length of about 30 inches and a weight of some 20 pounds, and the maximum age is thought to be approximately 20 years. In a pattern typical for adults of many species of estuary-dependent fish, adult summer flounder move inshore in the spring and summer to feed in the bays and sounds and migrate offshore to spawn in fall and overwinter at the shelf break (Able and Fahay, 1998). Summer flounder appear to migrate northeast along the continental shelf as they grow larger. Considered lie-in-wait predators, they eat a large variety of prey, usually smaller fish, crabs, and shrimp. Commercial catches of summer flounder along the Atlantic Coast have fluctuated between 2 and 18 million kg over the past 50 years. Recreational catch has been tracked only since 1981 and during this period, catch has fluctuated between
OCR for page 23
IMPROVING THE COLLECTION, MANAGEMENT, AND USE OF MARINE FISHERIES DATA FIGURE 2-1 Commercial catch since 1950 and recreational catch since 1981 (earlier recreational data are unavailable) from Maine to North Carolina. 1.4 and 12.7 million kg annually. Amendment 2 to the Fishery Management Plan for summer flounder, approved in 1992, allocated 60 percent of the total allowable catch (TAC) to commercial fishermen and 40 percent to recreational fishermen. However, recreational catch has equaled commercial catches in several years and exceeded commercial catch in 1997 and 1998 (Figure 2-1), 69% over the recreational TAC. The proportion of fish in the population that are three or more years old has increased from 4 percent in 1993 to 43 percent in 1998 (Terceiro, 1999). Do the summer flounder found in waters north of Cape Hatteras comprise a unit stock of fish? National Standard 3 of the Magnuson-Stevens Act (Sec 30 [a]) states “to the extent practicable, an individual stock of fish shall be managed as a unit throughout its range, and interrelated stocks of fish shall be managed as a unit or in close coordination.” For ideal management, a species' distribution would be split into a set of discrete stocks, each of which spawns separately, has similar growth and mortality characteristics, and is fished by a different set of fishermen. This ideal is rarely achieved and the reality is that stock management areas have to be large, often comprising several separate, but usually overlapping, breeding populations and nursery areas. There are obvious administrative advantages of this approach and scientifically there may be advantages to combining sub-stocks rather than having to estimate migration rates among them. However, problems may result from combining a number of sub-stocks, potentially spanning a large gradient of latitude. The most important problem is that a combined stock may appear to be in reasonable shape even though some component is being fished to extinction. Combining sub-stocks may also hinder stock assessments if growth, recruitment, natural mortality, or fishing mortality differ substantially among the sub-stocks of a combined stock. This may result in difficulties of estimating average weights at age or mortality rates. Thus, the adequacy of stock definition should be questioned in every assessment (see Appendix D of NRC, 1998a). In the case of summer flounder, Able and Kaiser (1994) published an extensive review of the species' life history. Three possible interpretations of the number and range of summer flounder stocks were presented:
OCR for page 24
IMPROVING THE COLLECTION, MANAGEMENT, AND USE OF MARINE FISHERIES DATA Two major stocks: Middle Atlantic Bight and South Atlantic Bight Two major stocks, both in the Middle Atlantic Bight Three stocks: Middle Atlantic Bight, South Atlantic Bight, and trans-Cape Hatteras The committee was not able to determine which of these interpretations is correct, but provides comments that may be useful for multi-stock management in this context. Fogarty et al. (1983) used morphological characteristics and discriminant analysis to explore summer flounder stock structure. They found significant morphological differences among New York-New Jersey, Mid-Atlantic, and Cape Hatteras stocks, but were uncertain whether these differences had any biological significance. Although Jones and Quattro (1999) identified some genetic differences between summer flounder from Massachusetts and Rhode Island, it is unlikely that these are different stocks of fish that should be managed separately. Jones and Quattro pointed out that stock identification should be based not only on genetic data but should also consider life history attributes and morphological characteristics. The existence of genetic, morphological, or other differences does not automatically mean that different populations should be treated as separate stocks for management purposes because (1) morphological or genetic differences may not translate into significant ecological differences and may simply reflect short-term or small-scale fluctuations in population structure and ultimate survival of a species; (2) the differences may not be related to factors that are important in stock management; and (3) the value of the fishery may not justify the cost of treating the stocks separately. Conversely, it is possible for populations to be genetically and morphologically identical throughout a vast geographic range and yet it may be preferable to manage them as separate stocks because of differences in growth, reproduction, and mortality. Such differences can be driven by differences in water temperature, productivity of phytoplankton and zooplankton, abundance of natural predators, fishing pressure, and other factors. What are the implications for stock assessments and management if the unit stock assumption is wrong? If separate stocks are fished differently, the fishing mortality rate estimated for the combined stock will not be applicable to the separate stocks. Very intensive fishing of a small stock may have little impact on the overall fishing mortality. For example, the Arcto-Norwegian cod population has two components, one that feeds in the Barents Sea and one in the Bear Island-Svalbard area. Both components apparently spawn together on the Norwegian coast, mostly at Lofoten (it is not known whether they interbreed). Typically, the Barents Sea component is twice the size of the Svalbard component. Hence, if the former were not fished and the latter fished to extinction, the combined fishing mortality rate might be 0.3, which might seem reasonable, but the Svalbard stock component would have been eliminated. This is more or less how the North Sea herring collapsed, as separate units of a stock over its entire range were successively fished out. This scenario is more likely with schooling fish because fishermen can maintain high catch rates on declining, contracting populations rather than move to more abundant stock components. A similar situation exists with the various species of salmon on the west coast of North America. Species that are listed as endangered in California, Oregon, and Washington are abundant in Alaska. Thus, if the mortality of all salmon stocks is calculated together, the resulting number is not alarming. However, the number also does not reflect the true situation of the health of the salmon stocks. Northern cod provide another example (Rose et al., 2000). If, on the other hand, a unit stock splits into two separately assessed stocks without accounting for movement between them, differential migration between the components will distort estimates of fishing mortality and population size. For example, some cod spawned at the southwest corner of Iceland drift as larvae to Greenland, where there
OCR for page 25
IMPROVING THE COLLECTION, MANAGEMENT, AND USE OF MARINE FISHERIES DATA is a separate fishery. They return as spawning adults to Iceland, where the extra numbers accruing to the Icelandic stock distort the fishing mortality rates downward if this factor is not taken into account (Shepherd and Pope, 1993). Committee Investigations The committee could not determine whether the summer flounder population should be considered multiple stocks. NMFS told the committee that “no meristic or morphometric data have been or are being regularly collected that would enable separation of the historical landings in potential stock components.” However, any fishery that spans a large range might include multiple stocks. Actions Needed Testing the validity of the unit stock assumption requires an analysis of growth rates, fecundity, mortality, and other population factors among sub-areas. NMFS should conduct separate assessments of the stock if there are indications of different population factors (e.g., a difference in size-at-age between the northern and southern ends of the range), using whatever data are available specific to sub-areas of the range. Some relevant data may be available from the states, but a lack of standardization of their methods and differences in the timing and scope of their surveys may make such data inappropriate for reliable comparisons. The decision about implementing a split stock assessment depends on the benefit of doing so and the ability of the councils to use advice based on such assessments. What natural mortality rate is appropriate to use in summer flounder assessment models? Natural mortality rate (M) is the rate at which fish are removed from the population for reasons other than fishing activity, for example, predation, disease, and permanent emigration. In most cases, M is assumed to be the difference between total mortality and fishing mortality. For many temperate fish stocks, including summer flounder, stock assessment scientists assume that M is constant over time and at different ages of fish after they are large enough to be caught in the fishery and unaffected by fish population density.1 In the case of summer flounder, M is set at a level of 0.2 per year on the fished ages of the stock; this level is commonly, though perhaps too pervasively, used for many fisheries worldwide. The meaning of this M value is that if a year-class of the population were not fished, its numbers would decline exponentially so that year-class number N(t) is given by the formula N(t) = N(t0)e(−0.2*(t-t0)), where t is the current age in years and t0 the initial age. For example, this formula predicts that after one year has elapsed, 18 percent of the year class would have died, after 5 years, 63 percent would have died, and that after 10 years, 86 percent would have died of causes other than fishing. For fished populations, fishing mortality rate (F) and natural mortality rate (M) are generally viewed as additive. Both F and M have the units of year−1, but can also be expressed as instantaneous rates. In practice, fishing mortality is often the dominant cause of death at ages after fish have become large enough to be caught in the fishery. Errors in the assumption about the level of natural mortality lead to some well-understood effects on the resulting assessment. For example, errors in the estimate of M lead to approximately equivalent and opposite errors in the estimates of F. Thus, if M is underestimated by 0.1, F will be overestimated by approximately 0.1, and F targets for future catches will be lower than they should be and harder to achieve. Thus, an incorrect choice of the natural mortality rate moves the estimates of current and target fishing levels in a direction opposite to the error in M. A lower natural mortality rate causes the estimated yield- 1 Mortality of young-of-the-year fish, prior to their recruitment, is not included in the M used for management.
OCR for page 26
IMPROVING THE COLLECTION, MANAGEMENT, AND USE OF MARINE FISHERIES DATA FIGURE 2-2 Summer flounder yield per recruit for three levels of natural mortality. F and M have the units of year−1. per-recruit curve2 to steepen and peak at a lower level of fishing mortality rate and a higher M causes the curve to flatten and peak at higher levels of fishing mortality rate (Figure 2-2). An error in M may cause an error in the opposite direction for calculated biological reference points such as Fmax and F0.1 (see Box 1-2 for definitions), which are often targets for management. However, under certain circumstances, an error in M may cause an error in the same direction for biological reference points, such as Fmed (Sissenwine and Shepherd, 1987; Jakobsen, 1992). Thus, Fmed may lead to management advice that moves F in the correct direction even if natural mortality is estimated incorrectly. Because it is managed to an Fmax limit reference point, summer flounder management is affected by the choice of M. The preceding discussion assumes that natural mortality rate is constant with fish age and through time. If M changes with age, the estimate of F also is affected differentially by age. If M changes through time for fish of specific size, perhaps as a result of changes in the abundance of or susceptibility to predators, the effects may be variable and subtle. One effect of a period of unaccounted-for increased natural mortality is to underestimate the number of young fish expected to have entered the fishery. This is particularly the case if natural mortality increases on fish of post-recruit ages. Such effects are sometimes investigated using multispecies assessment models by accounting for predation mortality. Research on fisheries in other areas of the world suggests that predation mortality may be quite high and variable on the younger and smaller sizes of fish (Sogard, 1997). However, predation mortality estimates for North Sea flounder species are generally low (ICES, 1997). Multispecies models, in theory, allow estimation of natural mortality for prey species directly because they quantify M by observing predation deaths through stomach content data from predators. Multispecies models of the Georges Bank ecosystem have been constructed in the past, but have not been used to estimate predation 2 Such curves describe the expected yield from an individual fish (a single recruit) over its lifetime.
OCR for page 27
IMPROVING THE COLLECTION, MANAGEMENT, AND USE OF MARINE FISHERIES DATA mortality of summer flounder.3 Although a number of methods exist for estimating natural mortality rate indirectly, few are reliable at this time because estimates of M are confounded with estimates of the proportion of fish vulnerable to fishing at each age. Tagging experiments in which the initial mortality and tag loss rates on fish of catchable ages are well known and low provide another, more direct means of estimating M. The combination of knowing how many fish were released and the full time series of annual recaptures allows estimation of natural mortality rate (Hamre, 1980). Tagging data can also be used to estimate the dependence of M on fish age and size (Hampton, 1991). In fisheries characterized by a substantial mixed use (e.g., recreational and commercial harvest of summer flounder), each sector is likely to have a different reporting rate (typically recreational anglers are more likely to return tags) and each must be estimated properly. When tagging studies are combined with port sampling and intercept surveys wherein tags are obtained by the surveyor, both natural and fishing mortality can be estimated even in mixed-use fisheries (Brooks et al., 1998). Committee Investigations Because natural mortality rate is one of the more uncertain parameters used in a stock assessment, the committee investigated the effect of using a standard Laurec-Shepherd tuning of the virtual population analysis (VPA) of 1996 data (which gives results similar to the 25 th Stock Assessment Workshop assessment for the same data [NEFSC, 1997]). The biological reference points and future TACs and spawning stock biomass (SSB) estimates were then calculated on the assumption that the 1997 catch was 7,162 metric tons. Calculations were also made under the assumption that fishing mortality remained at the status quo level (the 1996 level) in 1997. These factors were then recalculated assuming a natural mortality rate of 0.1 (instead of M = 0.2) at all steps of the calculation (Table 2-1). Points to note from this analysis are that Fmax, F0.1, and F20% are lower for M = 0.1 than for M = 0.2. However, Fmed and the estimate of the current level of fishing mortality F1996 are estimated to be higher with M = 0.1 than M = 0.2. These specific results for summer flounder show how the various biological reference points change in response to a change in M, in line with the general theoretical expectations discussed earlier in this section. With respect to the predictions, the level of fishing mortality needed to take the TAC in 1997 was very similar in both cases, but represented a somewhat larger relative reduction in fishing in the M = 0.1 case (0.45 to 0.14), because the current level of fishing mortality appears higher. For M = 0.1, the equivalent catch in 1998 at Fmax is much smaller. This reflects the need to reduce fishing mortality to a lower target level. Spawning stock biomass is similar in both runs in 1997 but is higher for M = 0.1 in 1998. This is due to the larger reduction in fishing mortality rate. Catches and spawning stock biomass achieved in 1997, applying the same mortality rate as in 1996 (the status quo level), have similar results for both M values. Thus, these specific results for summer flounder also support the general conclusion that natural mortality estimates do not affect estimates of TACs much when M is held constant from one year to the next. It is clear that the correct value of natural mortality rate does influence the biological reference points and the likely target catch rates. A choice of a natural mortality of 0.1 would lead to more restrictive management decisions than one of 0.2, while one of 0.3 (not shown) would require lower reductions. M is a parameter that should be esti- 3 The Northeast Stock Assessment Workshop in Fall 1995 (SAW 20) examined methods of Hoenig (1983), Pauly (1980), and Fmax = 3/M (e.g., Anthony, 1982) and chose to continue to use M = 0.2 for all ages for summer flounder. In stomach content analyses, young summer flounder do not seem to be preyed on heavily by other species, with the exception that young summer flounder are occasionally found in the stomachs of striped bass and dogfish (M. Terceiro, NMFS, personal communication, 1999).
OCR for page 28
IMPROVING THE COLLECTION, MANAGEMENT, AND USE OF MARINE FISHERIES DATA TABLE 2-1 Effects of Changes in Assumed Natural Mortality Rate on Biological Reference Points, Estimates of Fishing Mortality, and Predicted Future Catch and Spawning Stock Biomass (SSB) M = 0.2 M = 0.1 Biological Reference Points (1996) F0.1 0.15 0.09 Fmax 0.24 0.14 F20% 0.28 0.17 Fmed 1.48 1.77 Estimated fishing mortality in 1996a F1996 0.95 1.03 Predictions for 1997 under status quob F1997 0.95 1.03 Catch1997 (mt) 12,248 12,585 SSB1997(mt) 12,263 11,403 Predictions for 1997 under TAC of 7,162 mt F1997 0.44 0.45 Catch1997 (mt) 7,162 7,162 SSB1997 (mt) 16,419 16,077 Predictions given 1997 TAC of 7,162 mt and F=Fmax in 1998 F1998 0.24 0.14 Catch1998 (mt) 4,541 2,898 SSB1998 (mt) 32,677 37,204 NOTE: SSB = spawning stock biomass; TAC = total allowable catch. See Box 1-2 for other definitions. The assumed value of M not only affects estimates of fishing mortality, survivorship, and biomass, it also affects biological reference point estimates. Thus, the consequences of assuming an alternative M must be assessed in the context of all the estimates influenced by M and should be examined on a time scale appropriate to the life span of the fish. a F1996 = the estimated fishing mortality for fully recruited summer flounder in 1996. This is the arithmetic mean of the fishing mortality over ages two, three, and four. b The status quo condition assumes that fishing mortality stays at the 1996 level (F = 0.95). mated correctly unless a management method less sensitive to its value (e.g., using Fmed) is used. What is the correct level of natural mortality for summer flounder? As with other species, the natural mortality rates of flounder species are difficult to estimate and often have been assigned values below 0.2. However, the M value for summer flounder has been estimated using a number of indirect approaches in previous stock assessment workshop (SAW) assessments (NEFSC, 1997) and the SAW concluded that M = 0.2 is a reasonable working assumption. Actions Needed Given that Fmax is a basis for management plans, and that it depends on the value assumed for M, it is essential to obtain an independent estimate of M for summer flounder. A well designed and carefully executed tagging program is an excellent method for estimating M. Such a tagging program has the added advantage of also providing estimates of F and of movement rates between different parts of the range of the population. Tagging programs are time consuming and
OCR for page 29
IMPROVING THE COLLECTION, MANAGEMENT, AND USE OF MARINE FISHERIES DATA difficult to conduct and require full cooperation of all stakeholders in the fishery to be effective. Emigration of fish from the surveyed area can significantly confound tagging studies, so estimates of emigration should be obtained. Relatively large-scale attempts to estimate emigration for summer flounder of the Chesapeake Bay region are described in the unpublished dissertation by Desfosse (1995). The benefits of such an experiment (unless results from a suitable previous experiment exist) would take a number of years to become available. In addition, the cost associated with a large-scale tagging study should be balanced with the benefit of obtaining information on this component of population dynamics relative to other, more pressing data needs, such as the vulnerability of older fish to fishing pressure across the management area and over time. Multispecies models that estimate predation mortality from stomach content analyses is an approach to estimating the importance of predation as a component of natural mortality. However, unless these kinds of data are already available, a new intensive (and expensive) sampling program would have to be implemented to sample the stomach contents of animals that prey on summer flounder. Such a sampling program may not be cost effective. It might also be possible to estimate M internally in the model, using Bayesian or other methods, as has been done in previous assessments for other species. Are there differences between the growth and mortality of male and female summer flounder (sexual dimorphism) and, if so, how do the differences affect the assessment? Fish size generally increases with age up to some upper biological limit for the species. This limit may vary with environmental conditions, differs from species to species, and may differ within a species, for example, between sexes. For a number of flatfish species, males grow at a faster rate than females, but often to a smaller maximum size. For such species, the largest fish are almost always females; for example, female North Sea plaice typically grow to a greater length and weight than do males (Rijnsdorp and Ibelings, 1989). Spring and fall surveys show a 20 percent larger size-at-age among female summer flounder at five years (NOAA, 1992). Eldridge (1962), cited in NMFS (1981, p. 15) also notes a 20 percent greater size in females than males; this difference increased to 30 percent by age 8. Differences in growth between sexes may also lead to different mortality rates at age between sexes if mortality is related to size. This difference may make it difficult to ascertain the full impact on the stock from data for both sexes combined. The sex of summer flounder can be determined by port samplers for commercial catch and by survey scientists from survey catches, and otoliths and scales can be collected to determine the ages for the same fish. Incorporating sex differences in growth and selectivity at age may be possible. If information collected from surveys can be used to identify sex-specific age and size characteristics that can be applied to commercial and recreational landing data, the assessment may be able to account for differences resulting from the sexual dimorphism. Committee Investigations The committee did not have suitable data to calculate the effect of ignoring sexual differences in growth. However, mortality at age should be greater for females than for males, until both are equally well retained by trawl nets in commercial fisheries and hooks in recreational fisheries. In turn, this greater mortality for females would lower the estimated spawning potential of the stock, decrease the effective spawning stock biomass, and increase the effective fishing mortality. Actions Needed As discussed in the next section, differences in catchability at different ages can affect stock assessments and the resulting estimates of fishing mortality and biomass. This is analogous to the multi-stock problem discussed in the previ-
OCR for page 30
IMPROVING THE COLLECTION, MANAGEMENT, AND USE OF MARINE FISHERIES DATA ous section, where one stock is male and the other female. If catchability differs with size (and thus between sexes), this will influence the evaluation of stock status. NMFS should determine whether there is enough difference in size at age and/or catchability at age between the sexes to warrant concern about the spawning potential of the stock. Currently, sex is not determined for catch samples from the summer flounder commercial and recreational fisheries. As a result, these data alone do not provide sufficient information to account for differences in growth between males and females in the assessment. It is possible to determine the sex of some fish species in catch samples when fish are gutted at sampling time, or when identifiable external morphological characteristics are available. Apparently this is not the case for summer flounder. It may be possible, however, to use survey information to distinguish gender-related differences since sex, age, and length are determined for survey samples. Use of this information could improve assessment results that are influenced by sexual dimorphism. If such data become available, NMFS should conduct catch-at-age or catch-at-length analyses that explicitly account for sex differences in size-at-age and selectivity. Such data could be applied retrospectively, assuming constant size-at-age by sex for years when observations are not available. QUESTIONS RELATED TO SUMMER FLOUNDER SAMPLING A common concern voiced by fishermen is that the gear used for sampling fish is antiquated and thus survey results are biased. This section addresses some of the concerns about the gear and methods used in the fishery-independent surveys. What are the appropriate survey and commercial catchabilities of summer flounder? Commercial fishermen believe that NEFSC surveys fail to fish appropriately in the survey strata areas and depth where the industry believes large summer flounder are found. Presentations to the committee by commercial fishermen suggested that large numbers of large summer flounder used to be taken on the portion of Georges Bank near the U.S.-Canada border (the Hague Line). This distribution of larger summer flounder is supported by some data from the American Littoral Society tagging program, the dissertation by Desfosse (1995), and Rountree (1994). Fishermen claim that even within the strata fished, the randomized stations tend to miss the sites where aggregations of large summer flounder are found. They also argue that the research gear (roller-rigged)4 and the method for using it (including 30-minute tows) are unsuited to catching the larger, faster swimming summer flounder. Fishermen believe that the groundfish surveys, particularly the fall survey, do not adjust for flounder migrations, so the surveys do not adequately sample the population. Thus, fishermen conclude that the age distribution observed is unrepresentative of the population. The tow duration issue is important to resolve, particularly because NMFS has considered reducing tow durations to 15 minutes in an attempt to increase the number of tows and thereby the precision of estimates based on the surveys. NMFS has not tested the effect of tow durations on catchability-at-age of summer flounder and other flatfish. However, the catch of flounders larger than 45 cm in length, in both the commercial fishery and surveys, has recently reached its highest point since 1982, according to NMFS. In considering these criticisms, it is important to understand how commercial and recreational catch-at-age data and survey catch per unit effort (CPUE)-at-age data are used in assessment models to estimate population size and structure. If both were used directly as estimates of population structure, rather than interpreted in the con- 4 Roller gear is a type of footrope (the rope attached to the bottom front of a trawl net) made of round rubber or steel bobbins interlaced with rubber discs and rigged to roll as the net is pulled across the bottom.
OCR for page 48
IMPROVING THE COLLECTION, MANAGEMENT, AND USE OF MARINE FISHERIES DATA These approaches can be combined, with various degrees of reliance on data and models. The first approach requires access to the elements of the sampling data and a clear knowledge of how the samples are extrapolated to the overall figures. Although this is usually straightforward for survey data, sampling of commercial fisheries often contains gaps that have to be filled with default assumptions (e.g., the catch of a fleet in a given month was not sampled but was assumed to be like that of the same fleet in a different month). Such decisions often have to be based on professional judgement, although cross-validation techniques can be used to calibrate the likely accuracy of gap-filling decisions. These assumptions can be difficult to track in retrospect unless they are very well documented, although some information can be salvaged by interviewing the staff that handled the data. An example of the second approach has been provided by Shepherd and Nicholson (1991), who developed a simple linear modeling approach that can be applied directly to survey catch rate-at-age data, weight-at-age data, and catch-at-age data. They noted that much of the systematic changes in the catch rate from a survey should be explained by the product of an age factor (to account for differential catchability across ages and the progressive decline of numbers with age due to fishing and natural mortality) and a year-class factor (to account for differences in the sizes of cohorts). This model is readily applied using an analysis of variance approach that fits age and year-class factors to a logarithmic transformation of the survey data. The lack of fit between the model and the data is ascribed to sampling error that is usually taken to have a log-normal distribution. However, a range of alternative sampling distributions can be fitted if a generalized linear model is used. Weight-at-age data also can be investigated with similar models, again with age factors (to account for growth), year-class factors (to account for systematic changes in growth between year classes), and possibly year factors (to account for changes in growth between years) fitted to log-transformed data. Again, any lack of fit is ascribed to sampling error. The same approach also can be adopted for catch-at-age data using age and year-class factors to explain sources of systematic variation seen in the survey catch rate data and a year factor (to account for annual changes in fishing mortality rate as well as systematic changes through time). The second approach is easier to apply in retrospect, but it suffers from the problem that the simple models adopted may not capture all the variation that results from systematic changes (e.g., changes in fishing patterns in catch-at-age data). Such variation may thus be ascribed incorrectly to sampling variation when it is a real signal in the data, inflating statistical variation and giving an impression of greater uncertainty than is the case. Most obviously, catch-at-age data may be affected systematically by changes in the selectivity of the fishing gear or the fishing practice of the commercial and recreational fishermen, or by the changing balance of catch between these users. Some of these changes can be examined by adding terms to the simple models described above, but others may be wrongly ascribed to variation in the data due to sampling variation. This can increase sampling variation and this possibility needs to be considered in the interpretation of the results. With this caveat the approach provides a simple means to examine the quality of the several data sets taken individually. Committee Investigations The analysis of design efficiency presented in Appendix C is an example of the detailed approach described above. In this analysis the stratified random designs used in the three NEFSC seasonal surveys in 1995 and the winter survey in 1996 were evaluated. In the first part
OCR for page 49
IMPROVING THE COLLECTION, MANAGEMENT, AND USE OF MARINE FISHERIES DATA of the analysis, the winter, spring, and fall surveys for 1995 were evaluated with respect to the application of the stratified design. A number of the strata in these surveys had only one station allocated to them and hence did not contribute to the variance estimates of the survey indices. In fact, sampling intensity is quite low for all of these surveys given the area covered, especially in the case of the 1995 fall survey, in which 48 out of a total of 56 strata had only one or two tows assigned to them. Design efficiency is a measure of how much the survey design has contributed to increasing the precision of the survey estimates. In the case of the stratified random design, two components contribute: the strata boundaries and the number of tows allocated to each stratum. The efficiency of the design for the 1995 surveys could not be evaluated because of the occurrence of strata with only one tow. Instead, the winter survey in 1996 was evaluated for each of three major flounder species caught in the survey—summer flounder, winter flounder, and yellowtail flounder. The contribution of the strata component to the imprecision was substantial for all three species, indicating that the current strata boundaries are related to differences in the distribution of the flounder species. However, the allocation of tows to strata was not optimal with respect to precision and, in fact, worked against the advantages contributed by the strata component. Our investigations indicated that the precision of the survey estimates could be improved by basing the tow allocation scheme for the current year's survey on a combination of the previous year 's survey results and the grouping of strata based on species distributions. The committee did not have the time to extend its analysis to the other years of the survey, but believes that such analyses would be very useful. The analysis of variance model approach was used to examine suitable survey data sets (some only provided data for one age, which is not sufficient for fitting the model), the weight-at-age data, and the overall catch-at-age data used in the summer flounder assessments. Table 2-6 shows the estimate of the standard deviation of the log(e)-transformed data, by age and overall, for the various survey series and for the weight-at-age and catch-at-age data of each data set both by age and overall. The standard deviation of the log-transformed data approximates to the coefficient of variation of the untransformed data series when it has small values, so these figures (if converted to percentages) give a first approximation of the coefficients of variation of the various data sets. Table 2-6 suggests that the individual surveys are rather variable. In practice they are used collectively to tune the virtual population analysis and indices used are all assigned the same weight.9 Hence, it is some average of their values rather than their individual values that affects the assessment outcome. Table 2-6 shows an attempt to estimate their combined logarithmic standard deviation (calculated as the square root of the average variance from all the surveys). Note that individually, the NEFSC fall and winter surveys are more variable than the combined survey estimate. The weight-at-age data seem to have lower standard deviations of log-transformed data, which is adequate for most assessments. Note, in particular, on a survey-by-survey basis, that the NEFSC fall and winter survey estimates are less variable than those of the state surveys. However, the combined survey estimates sometimes have half the standard deviations of those given for the NEFSC survey estimates alone. Clearly, some judgment has to be made about the precision, quality, and consistency of the survey estimates before they are averaged into a combined estimate for the assessment. By their nature, these data are susceptible to systematic variations that the simple model used here cannot capture. These variations are particularly likely to affect the catch numbers of 9 Because of differing precision at different ages, some indices might be more informative at some ages than are other indices. Assessment methods are available to estimate unique weightings among surveys for each age group.
OCR for page 50
IMPROVING THE COLLECTION, MANAGEMENT, AND USE OF MARINE FISHERIES DATA TABLE 2-6 Comparison of Federal and State Survey Precision in Terms of Standard Deviation of Log(e)-Transformed Data Data Set Age 0 1 2 3 4 Overall Survey catch rate-at-age data NEFSC fall survey n/a n/a 0.42 0.54 0.46 0.46 MADMF fall survey n/a n/a n/a 0.77 0.77 0.75 NEFSC winter survey n/a 0.58 0.55 0.50 n/a 0.51 NJDF survey n/a 0.82 0.82 n/a n/a 0.79 CTDEP fall survey n/a n/a 0.85 0.60 0.60 0.68 CTDEP spring survey n/a n/a 0.67 0.58 0.53 0.58 NEFSC spring survey n/a 0.73 0.62 0.64 0.65 0.65 MADMF spring survey n/a n/a 0.76 0.76 n/a 0.75 Combined survey estimate n/a 0.41 0.26 0.24 0.27 0.23 Commercial data Weight-at-age data 0.19 0.11 0.11 0.13 0.14 0.14 Catch numbers-at-age data 0.40 0.30 0.24 0.29 0.39 0.32 Catch numbers-at-age dataa (omitting age 0) n/a 0.30 0.21 0.25 0.34 0.27 NOTE: Cells with “n/a” are for those ages with no available data. CTDEP = Connecticut Department of Environmental Protection; MADMF = Massachusetts Department of Marine Fisheries; NEFSC = Northeast Fisheries Science Center; NJDF = New Jersey Department of Fisheries. a Omitting the age-0 catch numbers affects the values for older ages because of the way the ANOVA model used in this analysis fits factors for age and year-class to the data. Numbers at age 0 are highly variable and influence the overall estimates of variance for each combination of age and year-class. the youngest ages. It is noticeable that the apparent coefficient of variation of these data improves if the age-0 fish are omitted from the analysis. It is thus possible that the estimates in the table present a pessimistic view of data quality. It would be worthwhile for NMFS to investigate further the quality of these data. Actions Needed The efficiency analysis conducted by the committee was based on only two years of surveys. A similar analysis over a longer time series needs to be done to understand the persistence of the distribution patterns for summer flounder, as well as other species routinely caught by the survey. This information could be used to develop compromise allocation schemes to produce near-optimal results similar to those presented in Appendix C. If spatial patterns are not predictably persistent over time, the application of other methods, such as adaptive allocation techniques (Thompson and Seber, 1996), need to be explored. Simple tests of the quality of the assessment data suggest that the precision of the weight-at-age data are adequate but that the survey and to a somewhat lesser extent the catch-at-age data could be made more precise. As an initial step, the committee recommends that NMFS routinely calculate the variance associated with its routine catch-at-age data sampling. The committee recommends that constructions of estimated variance for commercial samples be included in the NMFS computer programs used to estimate quantities from these samples and that assessments of the statistical quality of results be given with the estimates. This will help to ensure that sampling schemes are performing appropriately and that the sampling design is relevant and provides the best value for the sampling effort. Moreover, it produces a value that is helpful in monitoring data
OCR for page 51
IMPROVING THE COLLECTION, MANAGEMENT, AND USE OF MARINE FISHERIES DATA quality, and if necessary, defending appropriate sampling levels and methods. The problem of missing data is particularly acute in the VPA method, but some other methods are better able to deal with missing data, highlighting the usefulness of analyzing data with multiple models. QUESTIONS RELATED TO THE INFORMATION CONTENT OF THE MODEL AND MODEL ASSUMPTIONS CURRENTLY IN USE What information does each model structure require and how do these requirements relate to information in the data? The structure and assumptions used in formulating a population model influence the modeling results in ways that are not always documented by stock assessment scientists and recognized by fishery managers and factored into their management decisions. The choice of a model and assumptions leads to some level of modeling uncertainty, as illustrated through the committee's re-analysis of summer flounder data. Much of the model structure and assumptions are derived from basic beliefs and principles, such as the notion that fish in a certain area are members of a closed population and that they die due to natural causes in proportion to the number present. However, some of the “information” or structure is imposed to provide estimates that cannot be derived from the data or to overcome deficiencies in knowledge about the system. Such imposed structure often includes assumptions that selectivity or catchability is constant through time, that natural mortality is constant and known, and that the same fishing mortality applies over the entire region. Although we know that such assumptions are unlikely to be completely valid, we cannot avoid making assumptions of some kind. We can, nevertheless, explore the sensitivity of our results to the choice of model and model assumptions by exploring the sensitivity of outputs to alternative models, alternative assumptions, and reasonable variations in the input data. Some of the assumptions have been examined earlier in this section. Here we consider how much difference in model formulations (with their typical assumptions) influence the outcome of assessment, including three related questions: (1) What weight does the model give to the different pieces of information that come into it? (2) How is uncertainty expressed in the assessment output provided to managers? and (3) Under what level of precision can managers expect to manage with data currently available? Committee Investigations Three alternative assessments of summer flounder—using VPA, ADAPT, and CAGEAN models—were carried out independently by different analysts and were compared with the NMFS 1999 assessment, which was done with an ADAPT model (see Appendix D). The same data were used in these alternative assessments, but the outputs from each of the models differed in response to the assumptions and structure of the model chosen (see Figure D-4 and Figure D-5). The effects of a number of alternative assumptions were explored, including different weightings of the data components and whether the selectivity curve fell or remained constant for older ages. The most significant difference among model results occurred as a function of the assumed consistency in fishing mortality across years. In the tuned VPA-type models, variations in F were controlled by the degree of shrinkage10 of estimated Fs to the terminal F. In the CAGEANtype model, the variation of fishing mortality was influenced by the assumption of constant selectivity. The ADAPT model apparently does not use shrinkage and could be emulated by the LaurecShepherd VPA model (by eliminating weighting towards recent years and by not using any shrinkage). However, the CAGEAN-type model in its pure form cannot reproduce the NMFS results, because of its different treatment of F, in that it interprets catch-at-age data in terms of F values that depend on an age-varying selectivity (constant through time) and a year-varying fishing intensity (constant over ages)— the so-called separability assumption. The VPA-based Laurec-Shepherd and ADAPT models, by contrast, allow fishing mortality rate to vary more flexibly but the price is that catch-at-age data have to be treated as though they are exact, which they are not (since F is based on a sampling process). 10 Shrinkage is the term for a procedure that gives some weight to the assumption that fishing mortality rate (and in some models also population numbers) are unchanging through the recent past. Thus, the values at age for the most recent year may be estimated from the values at age of past years, which are known with more confidence due to convergent properties of the VPA.
OCR for page 52
IMPROVING THE COLLECTION, MANAGEMENT, AND USE OF MARINE FISHERIES DATA Model results also differ as a function of each model's assumptions about how the population estimates should be interpreted given the data. For data collected after 1994, a marked decline is seen in the commercial catch numbers of age-0 fish and a smaller decrease in the catch numbers of age-1 fish. In the survey data from years after 1994, the numbers of age-0 and age-1 fish are equal to or greater than the number in preceding years. The models may interpret these changes in one of two ways, depending on the model assumptions. In one sense, the relative number of young fish may be decreasing, as evidenced by the commercial catch-at-age data. This is the interpretation resulting when selectivity and fishing mortality at age are assumed to remain constant (as in the CAGEAN model). Conversely, there may be a shift to a lower selectivity in the fishery for the younger age classes. Such a shift to catching less young (small) fish might be expected from the increase in mesh size and minimum size regulations and seems to be corroborated by the survey's data regarding relative abundance data. The resulting estimates, if there is lower selectivity for young fish in the fishery, would be more optimistic and the overall population size could be interpreted as being larger than if fishing mortality and selectivity at age had not changed. Under the latter assumption, the population is estimated to have held steady or decreased. The actual situation probably falls somewhere between the two extremes, because neither assumption is entirely valid. The risks for the fishery are very different under the two hypotheses, so the consequences of both should be explored. Some investigation is possible using the wellknown properties of VPA estimates. Fishing mortality rates estimated by this approach for younger ages in earlier years can be regarded as being relatively unbiased by assumptions. Consequently, graphing the ratio of the fishing mortality at each age to the average of the fishing mortality rate on 2- and 3-year-olds can be used to track changes in age-specific catchability. Summer flounder mortality estimates seem to indicate that the catchability of younger fish has decreased since 1982 and to have dropped sharply since 1994 (Figure 2-6). The results for 1995 and 1996 are likely to be fairly robust to assumptions made in 1998, but the 1997-1998 results are still strongly influenced by the assumptions made in that year (as indicated in the discussion above). A second approach uses analysis of variance of catch-at-age data to look for evidence of changing catchability (extending the approach of Pope and Shepherd ), but in the case of summer flounder this analysis was inconclusive. A final suggested alternative would be to divide catch into the survey catch-per-unit-effort index to derive a measure of effort that should be linearly related to annual fishing mortality, selectivity, and survey catchability, and analyze the variance of these values. In all three approaches, departures from simple linear models can indicate trends that might not otherwise be noted in the nonlinear assessment model results. A quick example of this method using the total catch-at-age data by year (C[a,y]) divided by the NEFSC spring catch-at-age data for ages 1-3 and their winter survey for the 0 groups by year (U[a,y]) is presented in Table 2-7. In applying this last method to the summer flounder data, the total catch-at-age data from the fleet (C(a,t) is divided by the NEFSC spring catch-at-age data (U(a,t)) for ages 1-3 and the winter survey for the age-0 groups. The quotient
OCR for page 53
IMPROVING THE COLLECTION, MANAGEMENT, AND USE OF MARINE FISHERIES DATA FIGURE 2-6 Summer flounder age-based selection relative to fishing mortality of age-2 and age-3 fish. C(a,t)/U(a,t) is approximately s(a)f(t)/q(a), where s(a) and f(t) are the selectivity at age a and full recruitment fishing mortality at year t for the fleet and q(a) is the catchability at age of the survey. Consequently, an ANOVA applied to the log ratio should be able to distinguish a year effect (log(f(t))), and an age effect (log(s(a)/q(a)). As a broad check for interactions, we divide the log ratio matrix into four quadrants (young ages-early years, old ages-early years, young ages-late years, old ages-late years) such that the sum of quadrant 1 and 4 minus the sum of quadrants 2 and 3 equal zero. From this an estimate is derived with one degree of freedom that indicates how selectivity at age in the commercial catch may be changing in earlier versus later years relative to the survey catchability at age, which we assume to be constant (Figure 2-6). The significant quadrant effect shown in Table 2-7 indicates that there is a systematic shift in s(a)/q(a) through time. Assuming that q(a) in the survey is constant, we must conclude that s(a) is changing. This would indicate that the constant selection hypothesis must be rejected. If the constant selectivity assumption of the CAGEAN model is relaxed to allow selectivity to change progressively through time, the NMFS results are better approximated. This echoes what the ANOVA above indicates—the consistency of selectivity is a key assumption and certainly critical in determining biomass and fishing mortality estimates for the most recent years. TABLE 2-7 Analysis of Variance Table for Survey Data Source of Variation Degrees of Freedom Sums of Squares Mean Squares F-statistic Value Probability Value Age (a) 3 59.30 19.77 35.20 0.0000 Year (y) 15 28.08 1.87 3.33 0.0010 Quadrant 1 7.94 7.94 14.14 0.0005 Residual 43 24.14 0.56 Total 62 119.46 1.93
OCR for page 54
IMPROVING THE COLLECTION, MANAGEMENT, AND USE OF MARINE FISHERIES DATA BOX 2-3 Committee's Conclusions About Assumptions Used in Summer Flounder Assessments and Recommendations for Studies Needed Using the virtual population analysis model runs, the committee concludes that: the common assumption that the natural mortality rate M = 0.2 is reasonable, given knowledge of natural mortality rates for other flounder species and data from the summer flounder stock. As long as M is not changed from year to year its effects on total allowable catch and spawning stock biomass are small. if older summer flounder are substantially less susceptible to trawl gear used in fisheryindependent surveys than younger fish (NMFS assumes constant catchability over all ages), the fishing mortality rate would be substantially lower than estimated by NMFS (but would still not attain the target for summer flounder) and the total allowable catch and spawning stock biomass would be substantially higher. if the NMFS values of summer flounder ages are correct and the North Carolina values are incorrect, the spawning stock biomass and TAC would be lower than if both are assumed to be correct. Conversely, if the North Carolina values are correct and the NMFS values are incorrect, spawning stock biomass would be higher. the statistical precision of the winter flatfish surveys (including summer flounder) could be increased by allocating sampling effort in a more efficient manner. Through other analysis and committee discussions, the committee concludes that: data available as part of the stock assessment were not adequate to determine whether the summer flounder population should be managed as a single unit. females are 20 percent larger than males at 5 years of age. This probably would not result in a major difference in the assessments now, because few flounder survive to age 5. However, if the fishery recovers to a broader age distribution, this sexual dimorphism may need to be considered in summer flounder stock assessments. the level of observer coverage in the fishery is probably too low to estimate discards accurately. Accurate estimation of discard levels is important because discard mortality is so high (assumed to be 80% for the commercial fishery) that a discarded fish has almost the same effect on the fishing mortality level as a fish that is landed. it is not apparent whether state surveys are timed and located to coincide with annual movements of summer flounder; it is therefore impossible to determine whether these surveys should be standardized. some components of recreational catch, particularly for the charter and party boat components, are relatively imprecise. The committee recommends actions in each case that NMFS could take to improve data and/or assumptions used. Priorities for changes in the summer flounder stock assessments and for research related to this species are discussed in the following points (not listed in priority order). One of the major disagreements between NMFS and commercial fishermen regards whether the methods and locations of NEFSC trawl surveys undersample large summer flounder. Because this is a major area of contention and the committee's simulations show that incorrect assumptions could substantially affect the actual values for fishing mortality and spawning stock biomass, it is important that NMFS and industry work together to resolve this issue. The following steps should be considered: Use data for fish up to 7 years of age, rather than 4 years, to constrain (tune) the model. Use tagging studies to determine whether the average life expectancy of fish is greater than the average age of fish caught in surveys, indicating the existence of a greater number of older fish in the population than observed in the surveys. Conduct sampling surveys of flounder eggs to determine whether the total biomass of spawning females is greater than assumed. Study whether the ability of survey trawl gear to catch large summer flounder is overestimated, thus underestimating the biomass of these fish. Conduct catch-at-age or catch-at-length analyses that explicitly account for sex differences in size at age and selectivity at age. Find a way to include commercial effort data in the stock assessments and use commercial vessels to assist in conducting summer flounder surveys, as exemplified by the Canadian sentinel surveys and adaptive sampling exercises. Increase observer coverage to improve the accuracy and precision of estimates of bycatch, discard rates, and landings of summer flounder, as well as to assist in tagging programs. Determine how state surveys are designed, timed, and conducted in relation to summer flounder migrations. Use the findings of this exercise to determine how best to use the state survey data in the summer flounder assessment. Increase the precision of recreational catch and effort estimates, perhaps by finding ways to better identify participants for inclusion in sampling surveys. Maximize the precision of trawl surveys relative to sampling objectives, as demonstrated in Appendix C. Work with fishermen and processors to ensure that larger (sushi market) flounder are properly sampled. Box 2-3 contains a summary of the committee's findings and recommendations related to analyses of the summer flounder assessments and data. POSSIBLE IMPROVEMENTS TO THE SUMMER FLOUNDER DATA SETS Based on the answers to the questions in Box 2-1, a number of actions could be taken to improve data collection. Actions could include
OCR for page 55
IMPROVING THE COLLECTION, MANAGEMENT, AND USE OF MARINE FISHERIES DATA using tournament data, conducting social and economic studies of commercial and recreational fishermen, employing commercial fishermen as collectors, and improving observer program data. Any changes in the sampling intensity should be judged in the context of the contribution of specific data to the overall assessment and management of the species and the value of improving management of summer flounder in relation to the management of other species.
OCR for page 56
IMPROVING THE COLLECTION, MANAGEMENT, AND USE OF MARINE FISHERIES DATA FIGURE 2-7 Virginia saltwater fishing tournament (summer flounder citations). Numbers indicate the minimum weight in pounds for a citation to be awarded.
OCR for page 57
IMPROVING THE COLLECTION, MANAGEMENT, AND USE OF MARINE FISHERIES DATA Tracking and Management of Data Precision In stock assessment, it is usually a mistake to measure one input (e.g., commercial or survey CPUE) with great precision, while only measuring other inputs approximately. The precision of all inputs, therefore, should be reviewed. Matching precision of inputs can be difficult because inputs often are sampled by different federal and state bodies drawing on different budgets and using different methods. Moreover, sampling for summer flounder is often carried out in conjunction with sampling for other species, with the result that costs are shared and optimization of sampling needs is viewed in terms of the wider program. To achieve the greatest possible precision of outputs, we suggest that precision targets be set for all sampling schemes and that performance be monitored annually in terms of achieving these targets. If this is done at the same time that data are analyzed, the requirement need not increase the time needed to conduct assessments. Such tracking of precision would assist year-to-year adjustments of data sources to keep them in appropriate balance. A useful exercise would be to conduct a thorough sensitivity analysis to evaluate how a given percentage change in the precision of each of the different inputs is translated by the assessment model into changed precision in the assessment outputs. Coefficients of variation (CVs) between 10 percent and 20 percent for composite estimates (e.g., total catch at age over significant ages, or combined survey indices of abundance) would be appropriate for most important fish stocks, at present, though changing needs and problems could require greater precision. Precision of the subcomponents of such measures could also be set to achieve this target range of CVs. For example, in the case of components of catch-at-age data, this would require a CV that is approximately equal to the overall target CV/(the square root of the catch share) (Pope, 1983). For example, if the target were a 10 percent CV, a fleet sector that caught 25 percent of the catch should require a CV of 20 percent, that is, half the precision. A Role for Commercial and Recreational Fishermen in Data Collection The summer flounder fishery clearly demonstrates the need for better cooperation among recreational anglers, commercial fishermen, state agencies, and federal agencies in data collection so that data quality is improved and its credibility is enhanced. There is room for improvement from all sectors. Strengthening the active support of commercial fishermen is particularly important. Chapter 3 and Chapter 4 include discussions of and suggestions about different ways that commercial fishermen could be enlisted to help with surveys and improve the quality of commercial fishery-dependent data. Summer flounder fishermen related several concerns about data collection methods currently in use. In other fisheries, mistrust between fishermen and scientists has been greatly reduced by their joint participation in data collection efforts: fishermen provide boats and advise on appropriate gear types and deployment procedures and scientists gain “indigenous” familiarity with the fishery and get to explain quality control limitations with which they must be concerned. Summer flounder fishery managers should examine these partnership approaches and test a working model in the summer flounder fishery. New Sources of Recreational Fishery Data Actions also could be taken to increase the participation of anglers in collection of summer flounder data. Anglers, who are particularly important in summer flounder population dynamics, participate in at least two types of programs that reward the catch of large fish and keep track of either exact fish weights (tournaments) or weights above a certain pre-set value (citations). Ancilliary data exist in state citation programs, and data from tournaments could provide qualitative trends in fish abundance. Many of the At -
OCR for page 58
IMPROVING THE COLLECTION, MANAGEMENT, AND USE OF MARINE FISHERIES DATA lantic Coast states sponsor citation programs in which anglers are awarded citations for fish that are at or above specific weights. The Virginia program is the best example of this type of data (Figure 2-7). The Virginia Saltwater Fishing Tournament records the number of citations awarded each year to anglers catching trophysize fish of specified weights. Most other state programs are much smaller. Citation data must be used with care because the effort of obtaining these trophy fish is not included and the weight that defines a trophy fish has changed over the years. Nonetheless, these data do reveal general trends in the abundance of large fish. For summer flounder there was a general increase in the number of trophy citations from 1958 to 1977, even with an increase in minimum citation weight. From 1978 to 1992, the number of citations remained low, despite a lowering of the minimum weight for a citation fish. Since the early 1990s, the number of citations awarded has increased dramatically. The trends in citations reflect, to some extent, the availability of large fish in the population. Although commercial fishermen expressed concern that older fish were potentially missed by NEFSC trawl surveys, some of these fish are apparently distributed near shore, where they are captured in recreational fisheries, primarily in the summer. A comparison of such data with commercial catch data from the same area could be used to crosscheck data sources and assumptions. Tournaments provide another auxiliary source of recreational data, particularly snapshots of the abundance of legal-size fish at particular times and locations.
Representative terms from entire chapter: