Models of codend size selection

Abstract

Different models relating codend selectivity to net variables such as mesh size, extension length and codend diameter often fit experimental data almost equally well. The choice of an appropriate model must then be determined by other considerations. This is particularly important when, for management reasons, it is necessary to extrapolate outside the range of the experimental data. Two families of codend selectivity models were considered. One family was developed using empirical fits to data. The other was developed using arguments about the physical and biological mechanisms which underlie selectivity: this family provides plausible prior selectivity models which might be useful for extrapolation. The fit to experimental data and the consequences of extrapolating each model was investigated. Although an 'empirical' model often provided the best fit, this sometimes gave unrealistic predictions outside the range of experimental data and such a model should not be used for extrapolation. When a plausible prior model gives the best fit to the data, its use is generally recommended. Otherwise, procedures are suggested for balancing the conceptual advantages of a plausible prior model with the empirical evidence for an alternative model.