A broad range of corals are protected in New Zealand seas by legislation that prohibits intentional damage and removal, and several protected areas have been designated in which bottom-contact fishing methods such as bottom trawling are prohibited. To understand the degree to which the distributions of these corals overlap with bottom trawling, and therefore to begin to assess the effectiveness of the current network of protected areas, it is necessary to have reliable maps of species distributions for these corals and identify key areas of high abundance across all taxa.
This report presents maps of abundance within the New Zealand region for eleven coral taxa, based on species distribution modelling using abundance values measured at 949 sample sites from image data collected by towed camera or remotely operated vehicle (ROV) systems. Most of these sites were sampled using NIWAs Deep-towed Imaging System (DTIS) and abundance values were based on archived analyses of video data. Further analysis of video data from three surveys was undertaken during this study to provide abundance values for selected locations not covered by archived data.
The species distribution models were constructed as an ensemble of predictions using two tree-based methods, Random Forests and Boosted Regression Trees. For each method abundance was estimated using a hurdle model – the combination of a presence-absence model predicting probability of presence and a regression model predicting abundance at locations of species presence. Uncertainty in the predictions was estimated using a bootstrap resampling technique to measure variability in predictions among 100 models built from random subsets of the sample data for each taxon.
The environmental predictor variables offered to the models represented a combination of seafloor characteristics, water chemistry, and productivity, tailored slightly for each taxon according to the chemical composition of their skeleton. In addition, and for the first time in any New Zealand species distribution modelling study, a date-specific trawl-contact variable representing bottom-trawling on the seafloor prior to the sample date was also included, based on compilation of recorded fishing effort in the region since 1990.
Models were successfully constructed for 11 of the 15 taxa considered, with observations of the remaining 4 taxa proving too rare in the data for models to be produced. Model performance metrics indicated good fits to the input data and a high level of agreement between observed and predicted values; these were backed up by metrics based on comparison of predicted values with values from data withheld from the model in each bootstrap sample.
A map of overall abundance for the set of protected corals modelled was made by adding abundance estimates across all taxa. Hotspots of coral abundance, nominally defined as areas where overall abundance was predicted to be over 40 individuals per 1000 m2, represented about 2% of the modelled area (the area of the seafloor between 50 m and 3000 m). These hotspots were spread widely around the region, most notably on the Norfolk Ridges, Challenger Plateau, Macquarie Ridge, Eastern Chatham Rise, and Kermadec Ridge.
The influence of the fishing impact variable was low in models for most taxa. For two taxa where it was important, Radicipes spp. and Goniocorella dumosa, maps were constructed to estimate pre-fishing distributions. However, these proved to differ very little from those predicting current distributions, probably because of insufficient temporal spread in the coral sample data for models to detect negative effects from fishing disturbance (i.e., most historical fishing effort and impacts precede collection of coral data in certain areas).
Anderson, O., Schnabel, K., Bowden, D., Davey, N., Hart, A. 2023 Identification of protected coral hotspots using species distribution modelling. POP2021-02 final report prepared by NIWA for Conservation Services Programme, Department of Conservation. 63 p.