Workshop on methods for merging metiers for fishery based sampling (WKMERGE)

WKMERGE was established to assist EU Member States to evaluate the appropriateness of their national sampling schemes for collecting metier-based biological data as required under the EU Data Collection Framework (DCF). Meeting held in Copenhagen, 19-22 January 2010.WKMERGE was established to assist EU Member States to evaluate the appropriate-ness of their national sampling schemes for collecting metier-based biological data as required under the EU Data Collection Framework (DCF). The DCF specifies collec-tion of data on discards and length/age composition of catches by fleet metier at the level of gear, target assemblage, mesh band and selectivity devices (DCF Level 6) in defined fishing grounds, with precision estimates at the stock level. Following from recommendations in ICES WKPRECISE (ICES CM 2009/ACOM:40), the Version 2009 of the DCF Guidelines for the submission of National Programme Proposals recog-nises that highly resolved and temporally dynamic metier definitions are an ineffi-cient basis for stratification schemes. This can lead to over-stratification and problems of under-sampling or non-sampling of strata, and poor control over sampling prob-abilities. Rather, Member States should specify sampling frames and sample selection schemes with temporally stable strata that are capable of providing sufficient data for the required metiers and fishing grounds. The metiers are treated as domains of in-terest rather than strata, unless the metier is sufficiently stable over time to act as a stratum with controlled sampling probabilities. Provided the population of vessels or fishing trips in each stratum is sampled representatively at each stage in the sample selection process, estimates by metier may be obtained using post stratification or ratio estimators, provided the number of sampled trips within metiers is sufficient. In this case, the sampled trips within metiers are re-weighted based on information about the distribution of all trips in the population. Hence, the use of post-stratification requires information on all trips in the fishery to adjust the sample weights. When trips cover more than one metier, knowledge on the metier of all hauls of sets within trips would be required for the post-stratification by metier. When planning the sampling programme, the number of trips to sample per metier becomes an expectation based on previous years’ fishing activities, not a target, and may alter if the fleets ‘activities change over time.

An important role of WKMERGE was to provide training on the design of robust sampling schemes for at-sea and on-shore sampling of fishing vessels to provide data on metier based biological variables. The workshop covered all aspects of sampling design including defining objectives; identifying the population to be sampled and suitable frames for accessing primary sampling units; stratification schemes; sample selection schemes including equal and unequal probability methods, and associated estimation procedures. The use and data-needs of model-based estimators were dis-cussed, including the pros and cons of “quota” sampling for model based and design based estimators. Examples of applying vessel list frames for at-sea sampling and area (access point) frames for on-shore sampling were covered in detail, and methods of combining data from both types of frames are included in the WKMERGE report. Workshop participants provided initial descriptions of their national sampling pro-grammes using a supplied pro-forma, and then reviewed these based on the outcomes of the workshop. A slightly revised version of the pro-forma, with associated guide-lines, is provided in the WKMERGE report to help Member States provide descrip-tions of their sampling schemes in their DCF National Programme submissions for 2011–2013. WKMERGE dealt primarily with the selection of vessels and fishing trips to sample for metier-related variables and did not address the selection of individual fish to sample for length or age from each trip. This is covered by ICES WKPRECISE.

Methods of optimising sampling schemes to meet multiple objectives were consid-ered, using an example given in a Working Document by France. Statistical procedures for identifying metiers as homogeneous groups of fishing operations (in terms of species and size compositions) were also discussed in the context of “merging of metiers”, using an example of multivariate analysis of data from observer trips given in another Working Document by France. Methods for comparing length composi-tions of species catches as a basis for merging metiers were described in a Working Document from Finland. WKMERGE recommends further development and agree-ment on appropriate statistical methods e.g. multivariate methods for identifying homogeneous metiers that are stable over time, and are at a resolution relevant to fishery management and capable of being sampled adequately with feasible sampling resources.

A primary focus of WKMERGE was the design of sampling schemes that avoid prob-lems of under-sampled and non-sampled strata or domains requiring imputation of missing data. When imputation is required, it should be done at the analysis stage using expert knowledge of the fisheries. Automated procedures for filling missing entries in databases with data “borrowed” from neighbouring samples or strata should be avoided. A major problem is non-accessibility of vessels for sampling at sea or on shore, as the vessels not available for sampling may have a different catch com-position and size frequencies than the accessible vessels. Characteristics of the non-accessible vessels should be recorded to allow retrieval of any auxiliary variables shown to be correlated with discarding or size compositions in the sampled vessels (e.g. gear; mesh; area; trip duration etc.).