Untangling the confusion surrounding the estimation of gillnet selectivity

Abstract

Three alternative methodologies for estimating selection curves for gill nets have recently appeared in the fisheries literature (Hovg rd 1996; Helser et al. 1998; Millar and Fryer 1999). This comment shows that these three ap-proaches have a common underlying structure. Identification of this common structure permits the differences between the approaches to be clarified, which in turn enables the rela-tive merits of these methodologies to be compared using well-known statistical results. The methodologies considered here are for indirect esti-mation of gillnet selection curves. Typically, gill nets of varying mesh sizes are fished simultaneously with equal ef-fort and in such a manner as to avoid unnecessary biases in the catch. The data are the catches of differing size-classes of fish in the various gill nets. In contrast, direct estimation is possible if the size distribution of the population (or a tagged subpopulation) being fished is known. Direct meth-ods are not considered here. The analysis of data from indirect gillnet selectivity stud-ies has a rather confusing history of competing methodolo-gies and assumptions, which is partly due to the often inadequate specification of "selectivity." Here, the selection curve is denoted r(l) and is defined to be the relative proba-bility that a fish of length l is captured given that it contacts the gear. To distinguish between the varying mesh sizes, r j (l) will be used to denote the selection curve for the jth mesh size. To keep things simple, this discussion will avoid consider-ation of length-dependent avoidance behaviour and availabil-ity, and the comparison of methodologies will be made for the case when all mesh sizes are fished with equal effort and have equal fishing power. These issues are extremely rele-vant to gillnet selectivity studies and are discussed in Millar and Fryer (1999) where the notions of contact selection, available selection, population selection, and fishing inten- is a (relative) contact selection curve (i.e., a retention curve), and Millar and Fryer (1999) explained that r(l) can be as-sumed to have a maximum height of unity without loss of generality. It is not possible to estimate available selection or population selection curves from indirect gillnet studies without additional information. Note that the above definition uses length as the measure of fish size. It is likely that girth would be a more relevant measure of size for the purposes of estimating gillnet selec-tivity, and this methodology applies equally well if length is replaced by girth. In practice, it is usually length that is mea-sured, and this is assumed here. Henceforth, l will be used to denote the length-class having length l as its midpoint. For example, the 30-cm length-class might be all fish with length between 29.5 and 30.5 cm.