Evaluation of the New Zealand marine environment classification for shallow coastal rocky reef fish communities

Abstract

The Marine Environment Classification (MEC) is a hierarchical spatial framework that classifies a large area of ocean surrounding New Zealand into classes of similar environmental character. This thesis uses multivariate statistical methods to assess the utility of this environmental classification as a management tool for coastal ecosystems, by examining its correspondence to biological patterns among sites in an existing dataset of shallow coastal rocky reef fish communities. Overall, the relevance of the MEC to coastal biological patterns was limited and varied widely among its hierarchical levels of resolution. Three broad-scale biogeographic patterns in reef fish communities were identified in the biological data, of which the MEC reproduced only one. Direct, hierarchical classifications of sites in the biological dataset were built, and these were able to explain up to twice as much variation in fish communities than the MEC, which explained fifty percent of variation at maximum resolution. The inability of the MEC to adequately correspond to coastal biological patterns was attributed to the persistence of a large generic coastal class and its failure to make partitions along a strong biological latitudinal gradient. A likely explanation is that while the variables used may be appropriate for offshore oceanic areas, some of the variables used are inaccurate and inappropriate for classification of coastal regions. Options for the further development of spatial models for coastal management in New Zealand are discussed. Two primary objectives for such models are (1) to provide a basis for the design of a network of Marine Protected Areas (MPAs), and (2) to delineate spatial units for management and environmental reporting. I argue that spatial variation in coastal reef fish communities at scales relevant to national management is largely continuous rather than discrete. If this is true, a categorical model or classification provides a limited and potentially inaccurate reflection of real biological patterns. Two implications of this were highlighted. First, continuous spatial models, as opposed to categorical or classification models, should be developed and used for MPA design and biodiversity management more generally. Second, any system of boundaries for categorical spatial management is likely to be largely arbitrary.