February 1st to 5th 2016
Olhão, Portugal
Tweet #MaresConference2016

Species composition and mortality rates of sea turtles interacting with the Portuguese pelagic longline fishery targeting swordfish in the Atlantic

Oral Presentation
Natural Resources
Wednesday, February 3, 2016 -
14:45 to 15:00

Coelho, R. 1 Rosa, D. 2 Lino, P.G. 3 Santos, M.N. 4

11. Portuguese Institute for the Ocean and Atmosphere (IPMA) 2. Centre of Marine Sciences (CCMAR), University of Algarve, Campus de Gambelas, 8005-139 Faro, Portugal
2Portuguese Institute for the Ocean and Atmosphere (IPMA)
3Portuguese Institute for the Ocean and Atmosphere (IPMA)
4Portuguese Institute for the Ocean and Atmosphere (IPMA)

Sea turtles occasionally interact and are captured as bycatch in pelagic longline fisheries. In oceanic waters of the eastern and tropical north Atlantic, the main species interactions are with leatherbacks (Dermochelys coriacea), loggerheads (Caretta caretta), olive ridley (Lepidochelys olivacea) and Kemp's ridley (Lepidochelys kempii). However, a relatively large proportion of the interactions are only reported as not-identified sea turtles, making species-specific assessments difficult to conduct. In this work we used statistical models to estimate species composition and fate of sea turtles captured in a pelagic longline fishery, specifically the Portuguese pelagic longline fishery targeting swordfish in the equatorial and tropical north Atlantic. Multinomial models were used to predict the species composition and binomial models to predict the fate of the captured specimens, with both models including spatial and seasonal effects. Both models showed good goodness-of-fit. In the multinomial models the estimated coefficient of determination (R2) was 0.562 and the 10-fold cross-validation procedure resulted in a classification error rate of 46.8%. In the binomial models the estimated R2 was 0.293, the Area Under the Curve (AUC) was estimated to be 0.805 with a sensitivity of 74.5% and a specificity of 75.5%, and the 10-fold cross validation procedure resulted in a prediction error rate of 21.1%. The models produced can be used to reconstruct the species composition and estimate the fate of sea turtles that interact with this pelagic longline fishery in the equatorial and tropical north Atlantic, providing a better understanding of the impacts of this fishery in sea turtle populations.
keywords: 
Catch reconstruction, mortality rates, pelagic longline fisheries, sea turtles, species composition.

Log in to upload your presentation