Annex C
Method

This annex provides:

• additional information on the structure of the discard database;

• support for assumptions regarding certain discard rates used in the database; and

• discussion on the problems of determining accurate estimates of discards, with particular reference to discard sampling and raising or extrapolating the sample estimate to the population (i.e. the fleet, species or fishery).

An accurate determination of bycatch and discard rates is important for fisheries management. If the rates assumed for a fishery are too low, then TACs are likely to be exceeded (Pacific Fishery Management Council, 2001). The long-term biological stability and yield of the fishery may be affected and rebuilding strategies for depleted stocks may not be successful. Chronic underestimation of fishing mortality places the future economic benefits to the industry at risk through further depletion of resources. If the bycatch and discard rates assumed for the fishery are too high, then total mortality is overestimated and TACs may be set too low, which unduly restricts fishing, resulting in less economic benefit to the industry.

C.1 DIAGRAMMATIC REPRESENTATION OF CATCH CONCEPTS

Figure 3 provides a diagrammatic representation of the catch concepts upon which the FAO Fishstat statistical system is based.

C.2 DISCARD SAMPLING

Accurate estimates of discards depend on discard sampling and the subsequent raising or extrapolation of the sample estimates to the entire population. Sampling discards and raising of sample values to the species, fleet or fishery level pose numerous technical difficulties that are briefly discussed below.

The complex nature of many discard investigations can make them costly in terms of financial and human resources. Costs can prove prohibitive to many fisheries administrations. A recent investigation into the discard and escape mortality of Nephrops and roundfish from demersal trawls required the concerted efforts of six different institutes and other parties from five different countries (Denmark, Norway, Sweden, the United Kingdom and the United States), at a cost of 1.9 million euros (Fisheries Technology Committee, 2000). Multivariate analysis of discard sampling data may also provide insights into the design of management measures (Murawski, 1996).

Clarity regarding the purpose of sampling^[136] is essential for design of an effective sampling protocol. If the discard estimates are to be used as an input to stock assessments, then detailed information on parameters such as sex, weight, age, length, maturity and fecundity may be required.

There are several approaches to estimating discarding in a commercial fishery:

• observers

• retention of discards by fishers

• questionnaires

• simulated commercial fishing

• modelling

• Delphi.

FIGURE 4: Diagrammatic representation of catch concepts (FAO)

C.2.1 Use of trained observers

Sampling of discards by observers is generally regarded as the most effective and accurate method (Punt, 1999), but not if discarding is illegal.^[137] Many national authorities and regional fisheries bodies (e.g. NAFO, ICES/EC, NEAFC, CCAMLR) make use of observers for sampling of discards.

Observer programmes encounter a range of difficulties. Discards cannot be assessed accurately where catches are slipped. Safety of observers at sea is a concern, particularly if the vessel crew are “hostile”. In situations where there are a limited numbers of observers and several fleets (vessel strata), there are problems in selecting the vessels on which to place observers, as there is a need to cover all strata and in particular the strata with the greatest inherent discard variability (Cotter et al., 2002). Vessels also change gear on trips. Discard variability tends to be higher between trips than between hauls or sets, requiring more trips rather than more fishing operations to be sampled. Vessels are the primary sampling units and days and trips may prove difficult to use for random sampling. In addition to the difficulties in designing discard sampling programmes to reflect the variability in fleet characteristics, fisher behaviour and the spatial and seasonal variation in fish distribution, some sampling may require consideration of the diurnal variation in fish behaviour, placing additional burdens on observers (van Beek, 1998).

Observer schemes are only useful for estimating total bycatch where there is also an adequate measure of total fleet activity. Furthermore, observer schemes can only provide a minimum estimate of bycatch (European Commission, 2002b). Even the most vigilant observer will miss some events. Animals that are trapped in fishing gear underwater, but then fall from the gear before it is hauled back to the boat, for example, will almost never be counted. Observers must also be able to see the net or other gear as it reaches the boat and access the catch as it is sorted. During the hours of darkness this ability may be compromised, depending on lighting conditions, and this can also lead to underestimation. If automatic sorters are used, the observer may never get a full view of either the catch or the discards. Moreover, the presence of the observer may alter the discard behaviour of the fishers.

A general impression obtained from the literature on discards is that insufficient attention is devoted to discards of non-commercial invertebrate species, such as echinoderms, tunicates, sponges and crabs. This may be because the primary focus of observer reports is on commercial species or because of the notion that such invertebrates are “rubbish”, “debris”, or of no interest. Even in regimes that prohibit discards, allowance is made for discarding of non-commercial species (such as in Iceland).

Real discards are always higher than visual estimates made by experienced observers, at times more than ten times higher (NAFO, 2000). However, observer reports are still the most reliable^[138] means of determining levels of discards and bycatch, even though the reports may reflect a minimum, rather than total level of discards. Where bycatch quotas are managed at the vessel level, managers and owners quickly recognize the potential consequences of biased sampling, and the sampling and estimation process may receive greater scrutiny.^[139]

The reply of NOAA/NMFS to Oceana

The reply of NOAA/NMFS to Oceana provides a valuable and comprehensive summary of the issues relating to the sampling of bycatch and, by extension, the sampling of discards (NOAA [Department of Commerce], 2003). The reply provides a balanced discussion on the objectives, needs, priorities, coverage and costs of bycatch sampling, with particular regard to the use of observers and the legal obligations on the administration to provide accurate estimates of bycatch.

C.2.2 Logbooks and retention by fishers

Requiring or requesting fishers to record discards in logbooks can provide a valuable source of discard information. For example, NAFO (NAFO, 2002) and NEAFC have rules on logbook discard data.

Comparison^[140] of discard information recorded by observers and in vessel logbooks may enable the correction of vessel logbooks to provide improved estimates of discards. While individual trip logbooks may not provide a high level of accuracy, discard rates based on logbook data averaged across trips, tows or fleet may be closely correlated with observer-based discard rates. If such a correlation can be shown, a correction factor may be applied to the logbook data, which are generally shown to underestimate discards.

Fishers may be asked (or paid) to collect, preserve and hold samples of discards from their own catches. Scientific staff then process the samples when the vessel returns to port. This approach may place a heavy reliance on the fisher to do the sampling or respect the sampling protocol but may be more cost effective than observer programmes (Lart, 2002).

C.2.3 Questionnaires and interviews

Responses to questionnaires may be inaccurate, or those fishers willing to complete the questionnaires may bias the results. Interviews must be confidential, they require a knowledgeable interviewer and a relaxed atmosphere, and they can be expensive. The Delphi method is a subjective method using the accumulated experience of recognized experts in the field. All approaches have mixed results.

C.2.4 Simulating commercial fishing

A research vessel or a chartered commercial fishing vessel is chartered and deployed with gear similar to that used commercially. The level of discarding can be estimated from the length distributions found in the catches by comparison with the length distributions in landed commercial catches (Medley, 2001). It is necessary to assume that fishing techniques successfully simulate those of commercial vessels and that fishing is geographically and temporally representative of how the fleet fishes. Similar inferences can be made from the composition of landings or even the size grades of exports of different fleets fishing the same fishery (The Irish Skipper, 2003).

C.2.5 Modelling

Discards may be estimated using data for total landings by the fleet, knowledge of the size selectivity of commercial fishing gear, and knowledge of the length distributions of the fish population (Casey, 1996). The latter may come from a research vessel survey using a small mesh trawl. This method may be helpful when no direct measures of discarding can be obtained. Although many assumptions are inherent in the method, no raising problems arise because modelling is applied to the total landings data. The approach may not be suitable for the estimate of total discards, i.e. fish that are not commercial and not length sampled, or for factory vessels. Models can also be of considerable assistance in designing a more effective sampling protocol; in testing discard reduction measures; and in replacing average discard rates with a more robust interpretation of the variability (Helser, Methot and Hastie, 2002).

C.2.6 Sampling design, sampling strata and data storage

If the primary objective of discard sampling is to improve the catch estimates used for stock assessments, it is advantageous to use the same sampling strata and to ensure compatibility between landings and discard databases. Raised discards can then simply be added to landings to give the total catch^[141] for that stratum.

However, in practice, use of these strata for sampling discards is not necessarily efficient. First, quantities of fish discarded are not necessarily proportional to quantities landed, depending also, for example, on the size selectivity of the fishing gear and the availability of quota to land a species. Second, it is possible to sample the landings of several vessels during one visit to a port, but one discard sampling trip may take two to three weeks. Third, the variability in discards is unlikely to be similar to the variability in landings. Therefore, a statistically valid sampling scheme (Tamsett et al., 1999; Allen et al., 2001) is likely to require different sampling strata and, to obtain accurate estimates of discards, sampling effort must be designed to concentrate on the levels of largest variability (Rochet et al., 2000; Allen et al., 2002). Furthermore, large numbers of sampling strata are not practical for discard sampling with small numbers^[142] of observers, as in most European countries. In many sampling periods it simply will not be possible for observers to sample trips from each of many strata. The result is that unbiased estimation is made very difficult. Several other factors constrain establishing standardized, at-sea sampling schemes for discards:

• high diversity of fleets in different countries;

• different levels of information available about these fleets (e.g. vessel lists, total effort);

• different trip lengths (e.g. one day or 30 days);

• different observer resources (from two covering a long coastline, to all trips observed, as in Canada); and

• to some extent, differences of opinion regarding the statistical framework and theory suitable for discard surveys.

While many countries are embarking on new discard programmes there may not be an effective system for the storage, retrieval and analysis of discard data. Logbook information on discards often remains unanalysed. Considerable additional efforts are required to establish statistically sound, cost-effective discard sampling protocols.

C.2.7 Transboundary stocks and fisheries

Transboundary stocks may require different approaches to sampling, particularly if international quota stocks are involved. Among ICES^[143] members, the country of origin generally makes arrangements to sample discards and associated landings. Observer programmes in international fisheries pose particular problems regarding responsibility for observers.

The country of landing

Observers in the country of landing often have difficulty in finding out when and where a foreign vessel will be landing. This makes scheduling of the trips with their main sampling programme for national vessels very difficult. Having successfully boarded a vessel in the observer’s own home country, the observer may be disembarked in another distant country and be faced with an expensive return journey. The observer may have language difficulties on board. There may subsequently be difficulties obtaining data from the foreign country to permit raising of results from the sampled trip to the appropriate fleet. Alternatively, results may be sent to the foreign country for their own use. Teams of observers tend to be so busy with sampling their own national vessels that they are reluctant to sample for other countries. This could be altered by making contractual arrangements, but the scheduling problem remains.

The country of ownership

Observers are likely to have to travel both to and from the country of ownership. The country of ownership may not have comprehensive lists of all the vessels owned (but only those registered with them). Inclusion of all vessels in a sampling scheme may therefore be difficult. The country of ownership is unlikely to have the information necessary to raise trip results to the appropriate fleet level. Observers from the country of ownership are likely to speak the same language as the crew.

The country of registration (flag country)

The flag country will have all available information about the vessel and will therefore be in a good position to schedule sampling and raise trip results to the appropriate fleet. Transportation and language problems are likely to arise for observers. However, contracting the country of landing to undertake the sampling and to send back the data for the trip may be practical, provided that sufficient notice is given.

International discard database

Baltic countries have established a common database (BALTCOM) which demonstrates (ICES, 2002) the opportunities for holding discard data regionally. However, it is necessary to assess carefully how such a database could be applied to regions other than the Baltic. Possible problems are:

• difficulties in drawing boundaries between seas;

• fishers may disapprove of the level of disaggregation of the data, which may reveal the identity of a vessel to fishing competitors or to enforcement agencies, even though the name of the vessel itself is not stored. Fishers might withdraw their cooperation with sampling activities as a result;

• national legislation on data privacy may constrain sharing of “raw” data;

• national sampling authorities may claim intellectual property rights over some data or fail to contribute required information for other reasons, e.g. concern that national quotas may be reduced;

• different types of fishing vessels, discard sample surveys and data raising methods in each region may cause significant computational problems; and

• linking of discards and landings data in one database, if thought desirable, could cause various technical difficulties in each region.

C.3 RAISING OF DISCARD ESTIMATES

Raising factors are required to convert sample results for individual catches, trips or vessels to estimates of discarding by the total fleet or fishery over a given sampling period, e.g. one year. The design of sampling protocol is highly relevant to the raising of discard data. Raising of results for a trip to an estimate for a fleet may utilize the number of vessels in the fleet, the total effort or the total landings. Whichever factor is selected, reliable data must be available to fit the chosen sampling strata. For example, a fishery with a defined gear type (or mesh size) as a sampling stratum should have matching landings (catch) declarations.

One common way to establish the fleet estimate is to raise the quantity of discards in a sample of a catch to an estimate of the discards in the total catch, then to extrapolate from the individual catch to the trip, from the trip to the vessel, and finally from the vessel to the fleet. Quantities may be recorded as weights or numbers of fish. Estimates of discards based on the mean of individual tows appear to overestimate discards. Using the sum of the discard weights and landing weights from all sampled hauls has been shown to give a more accurate estimate of discard rates (Allain, Biseau and Kergoat, 2003). Common raising methods^[144] include those:

• by landings (or catch)

• by number of vessels

• by number of trips

• by trips per vessel and number of vessels (two step)

• by other measures of effort, e.g. hauls, 1 000 hooks

• by probability of sampling;

• by strata

• in accordance with a model.

Each approach has its advantages and disadvantages and each is based on assumptions regarding the relationship between discards and parameters recorded during sampling. In most cases the assumption is that this relationship is linear. This is the assumption that has been used in raising discards to fishery level in the discard database. However, in some fisheries it is suggested (Rochet, Péronnet and Trenkel, 2002) that there is no auxiliary variable upon which discards can be accurately projected, i.e. sampling is essential to determine discards. The absence of a coherent theory upon which discards can be forecast suggests that no one method can currently be judged as the superior for all fisheries (Trenkel and Rochet, 2001).

Problems with estimates

A comparison between raised retained quantities and officially reported landings may reveal substantial anomalies, which may have consequences at a political level. Autoconsumption and illegal landings may account for substantial unreported catches. The use of unreliable landings records to raise discard estimates may render these estimates inaccurate. As many national fisheries statistics are not readily available by fleet, by gear or by fishery, discard information may be difficult to raise to the fleet or fishery level. There may be significant differences between effort-based and landings- based discard estimates, particularly with regard to the incidence and quantity of discards of shoaling species in trawl fisheries (Walmsley, Leslie and Sauer, 2003; Trenkel and Rochet, 2001).

C.4 DISCARD DATABASE FILE STRUCTURE

The database file structure is given in Table 33. Access to and use of the discard database will be determined by FAO.

C.5 ASSUMPTIONS REGARDING DISCARD RATES

Most small-scale and artisanal fisheries in developing countries have been allocated a discard rate of either 0.5 or 1 percent, as a zero discard rate was considered to be unrealistic for some artisanal fisheries. The rate assigned to a particular fishery is based on additional information from the country, the region or fisheries assumed to be similar.

Certain highly selective fisheries have been assigned a zero discard rate based on the available literature, contacts with experts and the author’s own experience. These fisheries include diver fisheries for abalone, sea urchins, lobster; squid jig fisheries and artisanal troll fisheries. Some of the supporting evidence is presented in Table 35.

TABLE 33
Description of the discard database fields

TABLE 34
Number of records by country or area

Note: not all records contain discard information. The number of records is an indication of the number of fisheries recorded.

TABLE 35
Supporting evidence for low or negligible discard rates in certain fisheries

Annex D
Summary of the reasons for discards

D.1 CAUSES OF DISCARDS

A clear understanding of the reasons for discarding is necessary in order to change discarding practices and help design sampling and raising protocols. A basic classification of the catch may be made into: (i) fish (species/sizes/sex) always retained; (ii) fish always discarded; and (iii) fish partially/occasionally discarded. The reasons for the discards are likely to vary by type of fish and discard reduction efforts may be most effectively focused on fish that are partially discarded.

D.2 CAUSAL DIAGRAM AND DECISION FRAMEWORK
D.2.1 Evaluation of bycatch

FIGURE 5: Evaluation of bycatch

Adapted from Horsten and Kirkegaard, 2002.

Figure 5 presents the causes and consequences of discarding.

TABLE 36
A classification of causes of discards

FIGURE 6: Causal diagram of discards