Modelling Choices Do Not Lead to Overestimation

Middleton et al. (2007) argue that some of the modelling choices in my analysis of Hector’s dolphin Cephalorhynchus hectori bycatch (Slooten 2007) lead to an overestimation of the impact of fishing on this species and of the potential for population recovery if fishing mortality were reduced to zero. This reply explains the rationale behind the model choices in more detail and shows that they do not influence the results in the way that Middleton et al. (2007) claim. One of the clearest demonstrations of this comes from a model (Davies et al. 2007) that became available after Slooten (2007) was published. Davies et al. (2007) is coauthored by 2 of the authors of Middleton et al. (2007) and it follows virtually all of Middleton et al.’s (2007) suggestions regarding catch rate, analysis of fishing effort and other aspects of model structure. I estimated that by the year 2050 populations would decline to 5475 ind. (coefficient of variation, CV = 0.20) if current management continues, and recover to 15 411 ind. (CV = 0.16) if fisheries mortalities are reduced to zero. The corresponding estimates from Davies et al. (2007) are 5688 and 14 799 ind. Putting Middleton et al.’s (2007) comments into perspective, we are essentially arguing over differences of a few hundred individuals over a 50 yr projection. Nevertheless, I will respond to their comments in detail below. Middleton et al.’s (2007) criticism of the way that fishing effort was dealt with is surprising, as the New Zealand Seafood Industry Council (SeaFIC) were consulted on this issue. After seeking advice from SeaFIC and the New Zealand Ministry of Fisheries (Mfish), I deliberately modelled fishing effort in a simple, straightforward manner to match the scale and quality of the data available. Effort data for gillnet fisheries are collected only to the scale of the 16 areas used in the model. Fishers are not required to provide data on distance offshore, water depth or detailed location information. More detailed treatment of fishing effort (spatially and temporally) would have been a highly subjective exercise involving many assumptions and arbitrary decisions based on anecdotal information from fishermen, voluntary logbook programmes etc. For example, Middleton et al.’s (2007) suggestions about inferring gillnet effort from the reported catch of 2 species (rig Mustelus lenticulatus and school shark Galeorhinus galeus) depend crucially on effort for those 2 species being representative of the fishery as a whole and on catch per unit effort remaining constant over time. In addition, such treatment ignores the fact that the rig fishery started as a trawl fishery (see Francis & Smith 1988). Gillnetting increased as trawling decreased, and both of these fisheries have an impact on Hector’s dolphin. A more complex approach was taken by Davies et al. (2007), filtering fishing effort data to exclude effort targeting species usually caught in shallow water (e.g. flatfish, mullet) and relatively deep water (e.g. hapuku, bass). For example, just under half the gillnet effort in Area 12 was removed, as these species were deemed not to overlap with Hector’s dolphin. This is plainly inconsistent with the fact that Hector’s dolphin eat flatfish and mullet (Slooten & Dawson 1994, Anonymous 2007a) and have been caught in gillnets set for several of the excluded species. Regardless of the likely bias introduced by excluding some fish species, such arbitrary decisions go well beyond the data available, relying heavily on anecdotal information and guesswork. Middleton et al. (2007) are incorrect in assuming that the way I have dealt with fishing effort would lead to overestimation of historical bycatch and carrying capacity (K). This would only be the case if there were substantially more overlap between gillnets and dolphins in Canterbury (where the observer programme was carried out) than in other areas. This is unlikely, for 2 reasons. Firstly, commercial gillnetting has been