Estimates of harvest rates for fisheries management are commonly used in a weight of evidence approach to provide an annual index capable of detecting variability in the thresholds set in harvest strategies. However, when using harvest rates to assess recreational fisheries, it is difficult to account for biological variation, changes in management or environmental conditions, and fisher behaviour. The use of nominal harvest rates can mislead stock assessments and subsequent management decisions. This study used a roving creel survey to determine the factors (survey year and month, targeting, fishing platform, fishers’ avidity, time of day and day type) that influence variability in harvest rates of finfish species caught by shore-based recreational fishers in the Perth metropolitan region. The species of interest were Australian herring Arripis georgianus, School whiting Sillago spp and Garfish Hemiramphidae spp. Generalised linear models, including the zero-altered gamma (ZAG) and Tweedie models, were assessed for their suitability via a five-fold cross-validation to compare their performance. Results showed that the significant variables and performance of models varied among species with ZAG and Tweedie models having the best performance overall. For Australian herring and Garfish there was a decrease in the harvest rate from 2010 to 2016, while for School whiting there was a slight increase from 2010. Estimates of harvest rate were significantly higher for fishers targeting School whiting and Garfish, and for these species, targeting was the only influential variable. In contrast, for Australian herring, the type of fishing platform and time of day were also key contributors in addition to targeting. This variation in significant variables across species shows that a blanket approach to choosing a model is not appropriate and that the model choice, as well as explanatory variables, is species dependent.
Â