BRUV as a method for assessing community baselines in Adriatic fish

The following results are a sumary excerpt from:

Dubravko Pejdo*, Claudia Kruschel, Stewart Schultz, Danijel Kanski, Ivana Zubak, Martina Markov, Petra Peleš. 2015. Fish monitoring in Kornati National Park: Baited, remote, underwater video (BRUV) versus trammel net sampling. VI. SAVJETOVANJE O MORSKOJ TEHNOLOGIJI in memoriam akademiku Zlatku Winkleru 20. i 21. studenog 2015, Hrvatska akademija znanosti i umjetnosti – Znanstveno vijeće za pomorstvo, Tehnički fakultet Sveučilišta u Rijeci.

We evaluated (1) the suitability of two alternative methods for fish monitoring: trammel net sampling and BRUV (Baited Remote Underwater Video), and (2) the potential to cross-calibrate the methods based on a set of shared species with high catch probabilities. A statistical power analysis concluded that BRUV can be conducted with sufficient sample size to perceive small changes in fish populations with high power, and therefore can be used as a sentinel monitoring method. We found that fish species detected by both methods amounted to almost a third of the number of species in each method's catch, and that 90% of these species are candidates for cross-calibration. 74% of the species at BRUV and 50% at trammel had occurrence probabilities above 10%, a reasonable threshold allowing stock assessment of these species. The sampled and predicted total species richness, extrapolated from the species accumulation curves, were almost identical across methods. We conclude that cross-calibrating the two methods and eventual replacement the trammel method with non-destructive BRUV is feasible. The most effective areas of improvement are increased BRUV night-sampling effort and increased total sampling size to increase the statistical power of BRUV as a monitoring tool.

We deployed 74 BRUV units and 16 trammel netsets in Kornati National Park, at islands of Klobučar, Kurba, Mana, and Šilo Vela. We then performed statistical power analyses for detection of difference in abundance of each species, with abundance definied as the maximum number of individuals observed in any video frame of each species. We also performed statistical resampling species accumulation curves to estimate the total number of species in the community detectable by each method.

We calculated statistical power to detect differences in fish abundance inside and outside MPAs assuming a coefficient of variation of approximately 0.5. Sample size was calculated based on catchability of each species, equal to the probability of detection of that species in a given BRUV deployment or trammel netset. We found that the minimum number of deployments necessary to detect a population loss of 50% in a species with catchability of 0.1, is 270 BRUVs, which corresponds to 68 trammel netsets, to equal the field effort in the two methods. This means that for a given difference in abundance of a fish species, the necessary sampling effort of trammel net is four times the sampling effort of BRUVs.

The following table indicates the minimum number of independent replicate samples needed to statistically detect the indicated propoportional decline in a fish population, for the indicated catchability of a fish species, at a significance (alpha) level of 0.05, and a statistical power (1 – beta) level of 0.95. This indicates that sample size decreases with catchability, with reasonable sample sizes in BRUVs capable of detecting losses of 40 to 50% for a catchability of 0.1. These values assume a binomial variable (presence/absence) and powers could be higher if abundance is used. Mobile and far ranging chase predators were far more likely to be caught by BRUV, such as Dentex dentex, Sparus aurata, Diplodus sargus, D. puntazzo, and Sphyreana sphyreana. Nocturnal ambush predators were more likely to be caught by the trammel method, primarily Scorpaena scrofa, and S. porcus; more mobile nocturnal predators were caught less often, such as Zeus faber and Pagellus erythrinus. Overall, the BRUV method was shown capable of monitoring approximately 5 more species than the trammel net method, regardless of the species' catchability, for any given sample size.

Mobile and far ranging chase predators were far more likely to be caught by BRUV, such as Dentex dentex, Sparus aurata, Diplodus sargus, D. puntazzo, and Sphyreana sphyreana. Nocturnal ambush predators were more likely to be caught by the trammel method, primarily Scorpaena scrofa, and S. porcus; more mobile nocturnal predators were caught less often, such as Zeus faber and Pagellus erythrinus. Overall, the BRUV method was shown capable of monitoring approximately 5 more species than the trammel net method, regardless of the species' catchability, for any given sample size.

The two methods detected approximately equal numbers of species overall, approximately 40, and species accumulation curves indicated similar extrapolated species richness of approximately 50 to 60 species in methods with low standard errors.

Overall, BRUV was capable of detecting larger and more mobile species more likely to be targeted by harvest, and therefore is better suited to monitoring such species than trammel netsets, including in MPAs designed in part to provide a refuge for such species. Trammel netsets, however, are better able to detect mostly sedentary mesopredators. The two methods have a species overlap at present of only about 30%, and cross-calibration procedures are now within reach for these species. However, more research is necessary to increase this overlap, especially for nocturnal ambush predators that have thus far mostly eluded capture by BRUV. In conclusion, BRUV is a valuable non-destructive, fisheries independent method for assessing abundance of a minimum of approximately 40 species, including large crusing predatory species with commercial and conservation importance.