Geostatistical tools supporting improved management practices for the New Zealand surfclam fishery

Abstract

This thesis comprises two bodies of work that are presented in the form of unpublished articles. These lay the groundwork and begin the process of proactively exploring different management options for New Zealand’s surfclam stocks. The first body work presents an R package named spatialSim, which was developed to carry out simulation studies for comparing the performance of different fisheries management strategies. Explicitly, the package provides users with an R interface to a C++ based multi-species spatiotemporal size-structured operating model for sedentary species. Four key features of the model include (1) the efficient simulation of systems at high spatial resolution, (2) an efficient harvest algorithm that supports the simulation of realistic commercial fisher dynamics and catch data through the specification of site selection constraints (e.g., economic) and different levels of preferential targeting and fishing intensity across space, (3) support for simulating scientific abundance surveys with any user specified design (e.g., a stratified survey of the biomass or number of individuals at size), and (4) support for specifying different levels of density dependence in recruitment. Overall, the spatialSim operating model supports more complexity and realism than a lot of the currently available alternatives, yet retains enough simplicity to allow the model and its parameters to be easily understood by users. The second body of work describes two bespoke geostatistical models that were developed for application to New Zealand’s surfclam stocks. Additionally, it reports the results of a spatialSim based simulation study that assessed the geostatistical model’s ability to estimate surfclam biomass using simulated scenarios representing New Zealand’s Kapiti Coast S. aequilatera and M. murchisoni surfclam fishery. This study compared the relative performance of the geostatistical models to the random stratified survey estimator that is currently used for assessing these stocks. Overall, the geostatistical models performed similarly, with both demonstrating potential to produce less biased and markedly more accurate estimates of each stock’s biomass than the survey-estimator. Moreover, they were able to achieve these superior levels of performance while using less data. The results demonstrated that the newly presented models could be very effective tools for supporting the sustainable management of New Zealand’s surfclam stocks, and they highlighted different aspects of each model that could benefit from further improvement.