Towards informing a data-driven approach to marine bioregionalization in South Africa: a case study using benthic epifaunal datasets from the southern Benguela shelf

Wozniak, Donia

Towards informing a data-driven approach to marine bioregionalization in South Africa: a case study using benthic epifaunal datasets from the southern Benguela shelf

Thesis / Dissertation

2023

Abstract

Marine bioregional, ecosystem and habitat classifications and maps are important for understanding and managing the marine environment. Benthic epifaunal assemblages often inform marine ecosystem classifications and maps, being recognized as good surrogates for broad benthic biodiversity patterns. In South Africa, ecosystem classification and mapping follow an expert-derived data-informed hierarchical approach. A move towards employing data-driven approaches to bioregionalization using quantitative biological and environmental datasets is underway. However, quantitative datasets collected with different sampling methods cannot easily be combined in analyses. It is also unknown whether available biological and environmental datasets can sufficiently define bioregions using existing datadriven approaches. As a case study, this research focuses on the southern Benguela shelf, located on the western margin of South Africa, where research trawl and towed camera sampling methods regularly collect quantitative data on epifaunal assemblages. This research therefore aims to 1) quantify congruency of epifaunal abundance patterns detected by a research trawl and a towed camera, so that their datasets can be appropriately weighted or prioritised in data-driven approaches and to 2) classify and predict epifaunal bioregions by applying a data-driven approach to bioregionalization using abundance data collected by research trawling. Various univariate and multivariate analyses were used to compare differences in species composition, diversity and assemblage structure at 18 sites (50–700 m) collected between 2017 and 2020 by towed camera and research trawl sampling methods. To quantify congruency in multivariate assemblage patterns between sampling methods, a symmetric cocorrespondence analysis (Co-CA) was used on the log+1 transformed abundance matrices. Univariate patterns of diversity were not significantly correlated between sampling methods, which detected mostly different subsets of epifaunal assemblages. The towed camera detected small and patchily distributed epifauna (e.g. the small brittle star Ophiura trimeni) and Anthozoans better than the trawl, while the trawl captured patterns of larger, highly motile Decapoda (e.g. the hermit crab Sympagurus dimorphous) and burrowing Asteroidea better than the towed camera. Though broad similarities in assemblage structure were evident between sampling methods, with high correlations found between important Co-CA axes (r = 0.93, 0.93, 0.79, 0.80), patterns were not significantly similar (p > 0.05). To statistically determine epifaunal bioregions across the study area, Regions of Common Profile (RCP) models were applied, using abundance data collected by research trawling between 2017 and 2020 from 325 sites. An RCP modelling approach was selected as a potential data-driven method for marine bioregionalization in South Africa, since classification and prediction are performed simultaneously, thereby quantifying uncertainty in estimated bioregions. Research trawl datasets were used due to their systematic sampling design which covers a greater spatial and temporal extent than other sampling methods across the study area. Rare species and collinear predictors were removed prior to modelling, resulting in 46 species and three environmental predictors (bottom temperature, dissolved oxygen and slope) used in the final model. Five bioregions were identified, based on lowest Bayesian Information Criterion (BIC). Low values of dissolved oxygen (< 0.3 ml/l) and low bottom temperature (3.76– 5.24 ˚C) were important predictors for bioregions which aligned with the inner shelf (RCP 5, < 150 m) and upper slope (RCP 1, > 500 m) respectively. These bioregions were associated with the highest confidence in spatial predictions. The highest uncertainty was attributed to bioregions across intermediate depths (RCP 2, 3 and 4, ~150–500 m) where species richness was highest. This study recommends the use of both towed camera and research trawl datasets to describe epifaunal assemblages holistically, though consideration should be given to the strengths and limitations of each dataset for specific applications. A low sample size of sites available for comparison between sampling methods may have influenced findings to be inconclusive, and further comparisons are recommended. However, findings suggest the strength of congruence between sampling methods is dependent on the species, habitat and spatial scale (resolution) of interest. Bioregions defined in this study aligned with broad depth breaks and known biogeographical patterns, though further effort is required to source and test relevant environmental predictors for species distributed across intermediate depth ranges (~150– 500 m). Data-driven approaches to bioregionalization which can quantify uncertainty in spatial predictions of bioregions and utilise quantitative datasets provide more information for management applications. As such, RCP models informed by research trawl datasets could be a viable option when delineating marine bioregions for South Africa, though validation with independent datasets is strongly recommended. This study highlights the importance of context, method and spatial resolution when detecting ecological patterns. When collecting data, different sampling methods may detect patterns with varying degrees of congruence depending on location, species sampled or the spatial scale of interest. When analysing data, the type and quantity of environmental predictors and species used to inform data-driven approaches likely influence bioregional patterns produced. The importance of long-term monitoring using a variety of sampling methods is emphasised to reliably compare and quantify bioregional patterns. Rigorous comparisons between datasets and analytical approaches are encouraged to improve understanding of their advantages and limitations. This study contributes methodological advances towards informing a data-driven approach to marine bioregionalization in South Africa.

Keywords

Biological Sciences

Reference:

Collections

Masters

Full item page