Figure 1.
Map of the Great Barrier Reef Marine Park (Australia) showing the location of all baited remote underwater video stations sampling sites and the distribution of sightings for the most common sharks.
Figure 2.
Shark species richness (mean ±SD) by (a) the cumulative number of baited remote underwater video stations and (b) the cumulative number of sites surveyed.
Maps show the distribution of shark species richness (c), and patterns (contours and colour shading) of variation of location along (d) and across (e) the Great Barrier Reef (GBR) continental shelf (rotated view), using an interpolation with a smooth spline with barriers technique. Distance along the shelf ranged from 0 on the southern edge of the GBR to 1 on the northern edge. Distance across was set to 0 on the coast and 1on the outermost edge of the continental shelf (80 m isobath).
Table 1.
Predictors used in the aggregated boosted regression tree and multivariate regression tree analyses.
Table 2.
Summary of shark sightings, abundance (MaxN; % MaxN) and the proportion of adults recorded on baited remote underwater video stations.
Figure 3.
Multivariate regression tree analysis of the occurrence of shark species explained by 12 environmental/habitat predictors (Cross-Validated Error: 0.90±0.05 SE).
The bold numbers at each node show the predictors that were most influential in predicting different shark assemblages. Histograms on the “leaves” show the frequency of occurrence of each species and the number of sites (n) with the node names and node numbers. The Dufrêne-Legendre species indicators (DLI) characterising each branch and terminal node (leaf) of the tree were included. Shark species at node 5: Sphyrna mokarran; node 6: Loxodon macrorhinus; node 7: Carcharhinus amblyrhynchos, Galeocerdo cuvier, Triaenodon obesus; node 15: C. albimarginatus.
Figure 4.
Summary of the relative contributions (%) of the top eleven predictors used in aggregated boosted regression trees (ABT).
Models were developed with cross-validation on data from 364 sites using tree complexity of 5 and learning rate of 0.001. Shark species richness and the occurrence (presence-absence data) from the indicator species of shark assemblages (see Fig. 4) were used in the ABT.
Figure 5.
Partial dependency plots from the aggregated boosted regression tree analysis of the occurrence and richness of shark species observed on baited remote underwater video stations.
The effects of the four most influential environmental/habitat predictors on the occurrence of Carcharhinus amblyrhynchos, C. albimarginatus, Galeocerdo cuvier and Triaenodon obesus. The bottom panel shows the effect of environmental predictors on species richness. For individual shark species, the y-axis represents the mean probability of occurrence centered at zero across all sites. Grey lines indicate ±2 SE for the predicted values, estimated from predictions made from 500 trees fitted in 5-fold cross validation at the site level.
Figure 6.
Effect of zoning on shark abundance, Great Barrier Reef of Australia.
The predicted abundance for (a, b) all shark species pooled, Carcharhinus amblyrhynchos (c, d, e), C. albimarginatus (f, g, h), Galeocerdo cuvier (i, j, k), and Triaenodon obesus (l, m, n) was examined across the range of hard coral cover (%), days since the new zoning (effective since July 2004) and nearest distance to reef (km). Areas closed (black lines) and open (red lines) to fishing and 95% confidence intervals are shown.
Table 3.
Summary results of Poisson (P) and negative binomial (NB) regression models used to examine the effect of zoning (areas closed/open to fishing) on the relative abundance of sharks (2004–2010).