Videographic analysis of the Coelacanth, Latimeria Chalumnae, and associated habitats in the Isimangaliso Wetland Park, KwaZulu-Natal, South Africa

• Main Author: Thornycroft, Rosanne Elizabeth
• Format: Others
• Language: English
• Published: Rhodes University 2012
• Subjects: Isimangaliso Wetland Park; Coelacanth; Coelacanth > South Africa > KwaZulu-Natal; Habitat (Ecology); Coelacanth > Behavior > South Africa > KwaZulu-Natal; Coelacanth > Habitat > South Africa > KwaZulu-Natal
• Online Access: http://hdl.handle.net/10962/d1005179

Videography is a valuable tool in biological and ecological studies. Using video footage obtained during previous coelacanth surveys, this thesis investigated coelacanths and their associated habitats in the submarine canyons of iSimangaliso Wetland Park, South Africa. This thesis aimed to (1) describe the biological habitats within the submarine canyons, (2) determine coelacanth distribution within these habitats, and (3) assess the use of computer-aided identification to successfully identify individual coelacanths. Seven different habitat types were noted with the most distinctive being the canyon margins that consisted of dense agglomerations of gorgonians, wire and whip corals, and sponges. Results suggested that although substratum type has a great influence on invertebrate community structure in the canyons, depth is the principal factor. Coelacanths were associated with cave habitats within the steep rocky canyon walls. Habitat analyses allowed predictive classification tree models to be constructed. Depth, underlying percentage of rock, and percentage cover of gorgonians and sponges were the most important variables for determining coelacanth presence and absence. The overall correct classification rate for the model was estimated at 96.6%, correctly predicting coelacanth absence (> 99%) better than presence (60%). Because coelacanths have a unique spot pattern it was possible to quickly and accurately identify specific individuals photographically using computer-aided identification software. Without any manual intervention by an operator the software accurately identified between 56 and 92% of the individuals. Identification success increased to 100% if the operator could also manually select from other potential matching photographs. It was also shown that fish exhibiting a yaw angle not exceeding 60° could be accurately identified in photographs. Each of the sections presented in this thesis represent a possible step towards analysing coelacanth-related habitats, locating and then analysing new habitats. Steps include first locating a population and then performing a habitat analysis. Coelacanth location within the different habitats can then be determined allowing the development of predictive models to potentially identify possible locations of new populations. The final step is to identify individual fish within the population for assessing demographic parameters and population monitoring.