11. Biological Regionalisation: Analytical Strategies

A range of regionalisation strategies were examined for this report to determine the robustness of derived regionalisation attributes. As discussed above, two different types of regional structures were examined: core provincial regions and zootones representing a distributed spatial boundary within which terminations or starts for two or more core provinces take place. The strategies used to produce the bioregions were:

1. An informal delphic regionalisation based on fish distribution information supplied by the expert taxonomists at the BioTax '96 Workshop. The list of participants in this regionalisation is as listed in Section 2. The delphic regionalisation was conducted only for the Australian mainland proper and did not consider a regionalisation for Tasmania.
2. An objective analysis of boundary coincidence using all the refined species from the BioTax workshop. A subset of the refined species with high information content and high reliability and accuracy was also analysed.
3. An analysis of the dissimilarity in species distribution as represented by a variation of the Jaccard index (Jaccard, 1912) and its extension to the two-dimensional domain for Bass Strait, as developed in this project.
4. For the oceanographic analysis, a multivariate pattern recognition analysis of selected attributes was performed.

For the purposes of the analysis in 2 and 3, the polygonal distributions were collapsed to a string by projecting the 2-dimensional distribution to the coast so that each polygon was reduced to a string with a start and end location along the coast. This analysis was designed to identify locations along the coast and shelf requiring further high-resolution analysis to refine boundary or zootone demarcations. An index ranging from 0 to approximately 283 units was used to identify locations along the mainland Australian coast. As the string attempts to follow the coastline the string distance will be function of the smoothness, or ruggedness, (i.e. fractal nature) of the coast. A plot of the string route (Figure 11-1) shows the disparity in the interval locations which are relatively close together for the rugged north-west region compared to, for example, the smoother southern half of the WA coastline. The effect of this non-uniformity in string distance is to bias those analyses which rely upon some measure of spatial closeness. Whilst it would have been desirable to enable the analyses to take account of such non-linearities, this was not feasible within the limited time-frame of this project.

Figure 11-1 Plot of indices (every 10th) along the "string" route beginning at an origin (0) just west of Cape York and proceeding clockwise around a coastline of Australia. The length of the string was approximately 283 units long representing a distance for the coastline used of some 37,000km.

For Tasmania, a separate string analysis, not including the offshore islands, was conducted (Figure 11-2). This was generated in a similar way as for the mainland string route.

Figure 11-2 Plot of indices (every 2nd) along the "string" route beginning at an origin (0) at about the Tamar River region and proceeding anti-clockwise around Tasmania.

Whilst the string analysis can provide a good first-order cut of the regions it is unable to deal with such two-dimensional areas as Bass Strait and to some extent, the Gulf of Carpentaria. Bass Strait in particular is bounded by open seas on two sides and by land on the other two together with infringing islands making it a somewhat complex environment to regionalise. We resorted to a two-dimensional analysis for this region by extending the boundary recognition schemes used for the string analysis.

Next Chapter: 12. Biological Regionalisation: String Analysis