New quantitative approaches for classifying and predicting local-scale habitats in estuaries

This study has developed quantitative approaches for firstly classifying local-scale nearshore habitats in an estuary and then predicting the habitat of any nearshore site in that system. Both approaches employ measurements for a suite of enduring environmental criteria that are biologically relevant and can be easily derived from readily available maps. While the approaches were developed for south-western Australian estuaries, with a focus here on the Swan and Peel-Harvey, they can easily be tailored to any system. fication of the habitats in each of the above estuaries was achieved by subjecting to hierarchical agglomerative clustering (CLUSTER) and a Similarity Profiles test (SIMPROF), a Manhattan distance matrix constructed from measurements of a suite of enduring criteria recorded at numerous environmentally diverse sites. Groups of sites within the resultant dendogram that were shown by SIMPROF to not contain any significant internal differences, but differ significantly from all other groups in their enduring characteristics, were considered to represent habitat types. The enduring features of the 18 and 17 habitats identified among the 101 and 102 sites in the Swan and Peel-Harvey estuaries, respectively, are presented. The average measurements of the enduring characteristics at each habitat were then used in a novel application of the Linkage Tree (LINKTREE) and SIMPROF routines to produce a “decision tree” for predicting, on the basis of measurements for particular enduring variables, the habitat to which any further site in an estuary is best assigned. h estuaries, the pattern of relative differences among habitats, as defined by their enduring characteristics, was significantly correlated with that defined by their non-enduring water physico-chemical characteristics recorded seasonally in the field. However, those correlations were substantially higher for the Swan, particularly when salinity was the only water physico-chemical variable employed. The lower correlations obtained for the Peel-Harvey were due either to little or erratic variability in particular water physico-chemical variables and/or to the spatial differences in those variables not being well captured by those in the enduring data. inary studies in each estuary indicate that the pattern of relative differences among habitats in the compositions of their fish and benthic invertebrate assemblages are closely correlated with those in their enduring characteristics. Such complementation would allow reliable prediction of the species likely to occupy any nearshore site in an estuary, simply by using the current predictive approach to assign that site to its most appropriate habitat.