Tsitsikamma National Park (TNP) possesses the oldest (established 1954), and one of the largest (350 km2) ‘no-take’ marine protected areas (MPA) in South Africa. A long-term monitoring (LTM) programme to observe the subtidal reef fishes in the TNP MPA was established in 2007. To date, 243 angling replicates have been completed, and a total of 2,751 fish belonging to 41 different species have been caught and released. In an era of unprecedented global biodiversity loss, data that can be used to monitor ecosystems and gauge changes in biodiversity through time are essential. This thesis aims to improve the methodological and statistical processes currently available for LTM of subtidal reef fish by providing an evaluation of the TNP MPA LTM programme. Angling data revealed definitive spatial structuring, in the form of spatial autocorrelation, and a shift in viewing spatial dependency as a statistical obstacle to a source of ecological information created a new avenue of data inference. Species-specific distribution maps identified localized habitat as the main predictor variable for species abundance, emphasizing the need for accurate a priori bathymetric information for subtidal monitoring. ‘Random forest’ analyses confirmed spatial variables are more important than temporal variables in predicting species abundance. The effectiveness of Generalized Linear Mixed Models (GAMMs) to account for spatial autocorrelation was highlighted, and evidence that disregarding spatial dependencies in temporal analyses can produce erroneous results was illustrated in the case of dageraad (Chrysoblephus cristiceps). Correlograms indicated that the current sampling strategy produced spatially redundant data and the sampling unit size (150 m2) could be doubled to optimize sampling. Temporal analyses demonstrated that after 50 years of ‘no take’ protection the TNP MPA ichthyofauna exhibits a high level of stability. Species-specific size structure was also found to be highly stable. Dageraad was the only species to exhibit a definitive temporal trend in their size structure, which was attributed to recruitment variation and the possibility that large individuals may migrate out of the study area. The inadequacy of angling as a method for monitoring a broad spectrum of the fish species was highlighted, particularly due to its selectivity towards large predators. As a result, a new sampling technique known as Stereo Baited Remote Underwater Videos (stereo-BRUVs) was introduced to the LTM programme in 2013. Stereo-BRUVs enabled sampling of 2640 fish belonging to 52 different species, from 57 samples collected in less than two years. A comparison of the sampling methods concluded that, compared to angling, stereo-BRUVs provide a superior technique that can survey a significantly larger proportion of the ichthyofauna with minimal length-selectivity biases. In addition, stereo-BRUVs possess a higher statistical power to detect changes in population abundance. However, a potential bias in the form of ‘hyperstability’ in sites with unusually high fish densities was identified as a possible flaw when using stereo-BRUVs. In an attempt to provide a more rigorous method evaluation, simulation testing was employed to assess the ability of angling and stereo-BRUVs to accurately describe a decreasing population. The advantage of this approach is that the simulated population abundances are known, so that each sampling method can be tested in terms of how well it tracks known abundance trends. The study established that stereo- BRUVs provided more accurate data when describing a distinct population decline of roman (Chrysoblephus laticeps) over 10- and 20-year periods. In addition, spawner-biomass was found to be a more accurate population estimate than relative abundance estimates (CPUE and MaxN) due to the inclusion of population size structure information, highlighting the importance of length-frequency data. The study illustrated that an evaluation framework that utilizes simulation testing has the potential to optimize LTM sampling procedures by addressing a number of methodological questions. This includes developing a procedure that aligns data collected from different sampling methods by applying correction factors, thus ensuring LTM programmes are able to adapt sampling strategies without losing data continuity.
| Identifer | oai:union.ndltd.org:netd.ac.za/oai:union.ndltd.org:rhodes/vital:5385 |
| Date | January 2016 |
| Creators | Parker, Denham |
| Publisher | Rhodes University, Faculty of Science, Ichthyology and Fisheries Science |
| Source Sets | South African National ETD Portal |
| Language | English |
| Detected Language | English |
| Type | Thesis, Doctoral, PhD |
| Format | 178 leaves, pdf |
| Rights | Parker, Denham |
Page generated in 0.0021 seconds