Library Open Repository

Describing and predicting the spatial distribution of benthic biodiversity in the sub-Antarctic and Antarctic


Downloads per month over past year

Hibberd, T (2016) Describing and predicting the spatial distribution of benthic biodiversity in the sub-Antarctic and Antarctic. PhD thesis, University of Tasmania.

PDF (Whole thesis)
Hibberd_whole_t...pdf | Download (32MB)
Available under University of Tasmania Standard License.

| Preview


The global imperative to sustainably manage deep-sea bottom fisheries and mitigate impacts to benthic habitats is constrained by the limited existing biological data available to inform decisionmaking. Physical surrogacy, where benthic biodiversity is characterised based on its relationship with environmental parameters, was explored as a means of extrapolating the distribution and biomass of benthic species from sample sites to seascapes of the deep-sea. The evaluation of surrogates focused on those benthic species considered most susceptible to disturbance from bottom fishing gears including sponges and corals (termed ‘vulnerable taxa’) and data from the sub-Antarctic Heard and McDonalds Islands (HIMI) region. HIMI hosts an established bottom fishery with protection for biodiversity afforded through a large marine protected area (MPA). However whether the MPA meets CAR principles (comprehensive, adequate and representative) in the context of vulnerable taxa remains largely unknown due to a limited understanding of the HIMI benthic habitats.
To readdress the paucity of basic information and provide empirical data with which to develop predictive models, quantitative benthic samples were collected from 104 stations in depths of 200 to 1000 meters and analysed to document benthic biodiversity and community structure across HIMI. Data from HIMI were then used to develop surrogacy methods that were applied to other regions in the deep sea.
A total of 312 taxa were recorded in the deep-sea at HIMI. Diversity was dominated by sessile suspension-feeders, including numerous undescribed and possibly endemic taxa, and was similar to other sub-Antarctic islands but lower than rich areas on the continental shelves of Antarctica and Australia. Analyses of assemblage structure using taxa biomass records and the clustering method 'Partitioning Around Medoids' revealed a clear zonation between HIMI’s eastern and western banks, the central plateau, south-facing slopes and waters deeper than 500 m, which was driven mostly by changes in seafloor current speed, temperature and the concentration of particulate organic carbon. Disturbance from bottom fishing was not identified as an important proxy for biodiversity despite extensive trawling for more than 10 years, and instead suggests a strong link between benthos and environmental parameters, highlighting the vulnerability of these communities to changing environmental conditions. Similarly, the restricted distributions of many taxa and levels of endemicity in some groups highlight the uniqueness and vulnerability of the HIMI benthic habitat and importance for conservation. Nonetheless, it was acknowledged that the study failed to sample the most heavily trawled areas at HIMI and that further taxonomic scrutiny (e.g. bryozoans are largely unsorted at this stage) might impact the study conclusions.
From empirical data at HIMI, ten vulnerable taxa were selected for which there were sufficient observations for model training (n >50). Four modelling approaches were contrasted to determine an appropriate method to model and predict vulnerable taxa across HIMI using physical surrogates: generalized linear models (GLM), generalized additive models (GAM), boosted regression trees (BRT) and random forests (RF). For each method, two sequential models were constructed; one to predict the occurrence probability of each vulnerable taxa (termed ‘occurrence model’) and one to predict the biomass of that vulnerable taxa given their presence in an area (termed ‘biomass model’). To contrast model performance, data were split into training and test datasets (cross-validation) and predictions evaluated using a series of performance indices relating to accuracy, calibration and bias between observed and predicted values. RF was identified as the preferred method to further explore and predict vulnerable taxa across HIMI due to consistent good performance (i.e. good accuracy, good calibration and low bias between observed and predicted values) and hence predictions were made using this approach. The predictions of occurrence and biomass of vulnerable taxa across the HIMI seascape indicated a higher frequency and biomass in shallow depths (<500 m), and on complex seascape features (e.g. HIMI's banks and craggy slopes), compared with the deeper areas of the plateau. Analysis of predictions using the conservation planning software Zonation highlighted HIMI's banks and numerous areas across the central plateau and continental slope as priority areas for conservation, many of which are currently protected by the MPA.
To test the broader applicability of the RF framework, models of vulnerable taxa were subsequently constructed for the continental shelf of East Antarctica (30°E – 150°E) where the Commission for the Conservation of Antarctic Marine Living Resources (CCAMLR) is considering the adoption of a Representative System of MPAs (RSMPA). Like HIMI, the benthos here are poorly described and marine conservation planning may benefit from model estimates to assist decision making. Benthic samples from the region and collocated environmental variables (including sea ice properties) were used to parameterise RF models for eight vulnerable taxa that were classified using the same principles as for HIMI. Both the occurrence and biomass models returned high accuracy according to the indices used, suggesting a high level of confidence in predictions across East Antarctica, and highlighting the transferability of the RF framework to other seascapes. Model estimates revealed a number of hotspots, namely the Prydz Bay region, but also Gunnerus Ridge, west of both Enderby Land and Casey Station and patches between Adelie Land and George V Land, the majority of which are encompassed within the proposed RSMPA.
Importantly the model estimates presented in this study suggest that CAR principles have been achieved for vulnerable benthos in the HIMI and (proposed) East Antarctic MPAs. In developing these MPAs, the distribution and hence representativeness of protection for benthic habitats and their biota within the MPA was inferred largely from physical variables as the empirical data required to characterise these habitats were sparse or not available at the time. My predictive modeling results that do incorporate empirical data and have produced similar recommendations for biodiversity conservation at HIMI and in East Antarctica suggest that the use of physical surrogates were an adequate tool for marine planning in the absence of biological data in these systems. More broadly, the results suggest that management or mitigation measures for benthos based on physical parameters may provide adequate precautionary management in other marine ecosystems where the empirical data necessary to evaluate the benthic habitat are lacking.
The accuracy of predictions and transferability of the RF framework means that methods developed here might be readily applied to other seascapes where decision-making may benefit from predictions. Sample size, model extent and data resolution were all potential sources of uncertainty which would best be addressed through targeted field sampling and surveys. However given the immediacy of the issue of managing bottom fishing to prevent significant adverse interactions with vulnerable ecosystems, and the practical difficulties associated with obtaining empirical data, surrogate-based management is the only practical means to make reasoned decisions about high seas resource management and for the establishment of a CAR system of MPAs throughout the Southern Ocean.

Item Type: Thesis (PhD)
Keywords: benthic invertebrates, deep-sea, surrogacy, modelling, Heard Island, Antarctica, MPA
Copyright Information:

Copyright 2016 the Author

Date Deposited: 11 Apr 2017 01:59
Last Modified: 01 Aug 2017 00:43
Item Statistics: View statistics for this item

Actions (login required)

Item Control Page Item Control Page