Research, monitoring and management of Hawaiian coral reef fishes remains heavily reliant on data collected using underwater visual censuses by SCUBA divers. However, diver-based visual sampling is subject to limitations, including: 1.) Biases caused by different responses of fishes to divers at fished versus remote locations. For example increased wariness of target fishes at heavily fished locations could lead to substantial undercounts. Equally, the attraction of large bodied roving predatory species, such as sharks and jacks, to divers at remote locations has likely led to inflated predator biomass estimates; and 2.) Domains limited to ‘shallow’ depths that can be readily surveyed by divers on open-circuit SCUBA – i.e. ~30m or less. As many ‘depth generalist’ coral reef fish species are distributed across a broad range of depths, and in some cases more abundant in deeper ‘mesophotic’ waters > 30 m, underwater visual censuses do not assess large portions of coral reef fish populations, and potentially overlook important conservation or management information (e.g. mesophotic depths may act as refuges from fishing). Conversely, ‘depth specialist’ species may be constrained to shallow or deeper depth strata, affecting overall community structure and limiting refuge potential.
Baited Remote Underwater Stereo-Video systems (stereo-BRUVs, herein denoted as ‘BRUVS’) represent one widely-used, alternative approach to bridge these knowledge gaps. However, this method has not been adopted as a standard tool surveying and monitoring reef fish and large-bodied roving predators in the Hawaiian Archipelago. While BRUVS may be capable of generating valuable information in data-deficient areas, concerns over additional biases relating to use of baited video surveys (e.g. attraction of predators, herbivore avoidance) have yet to be addressed in Hawaii. There is a need to increase the use of remote survey methods for surveying and monitoring reef fish and large-bodied roving predators and to critically assess the strengths, advantages, weaknesses, limitations and biases of BRUVS in comparison with other underwater visual census approaches. The interpretation of data collected from different methods, which have disparate biases and depth limitations, may result in very dissimilar estimates of coral reef fish populations leading to conflicting management actions.
This thesis combines several novel methodological and ecosystem-based comparisons for the Hawaiian Archipelago. Chapter 2 compares and contrasts the functional group composition of coral reef fish communities and a subset of predatory target species over a range of soak time intervals (i.e. 0 – 20, 0 – 40, and 0 – 60 minute sampling intervals) for both BRUVS and its unbaited analogue (RUVS). I also investigate the time of first arrival (‘TOFA’) and the time to maximum abundance (‘MaxNT’). I conclude that while shorter RUVS or BRUVS soak times of 20 minutes are sufficient to capture overall assemblage group structure and to survey resident or “fast reacting” species, longer BRUVS soak times (i.e. 60 minutes) improve herbivore assessments, and are better at capturing predatory sessile macropiscivores (eels belonging to Muraenidae and Ophichthidae) and generalist macropiscivores (large-bodied, roving sharks, jacks, and snapper), particularly targeted jack species that are considered depleted around human population centers.
In Chapter 3, I use BRUVS to compare and contrast reef fish communities between shallow water and mesophotic depths in the Main Hawaiian Islands (MHI), with continued emphases on
fishery-targeted species, and expanded to include Hawaiian endemics and linkages between assemblage counts and environmental features. A combination of multivariate PERMANOVA, pair-wise permutational ANOVA, and Canonical Analysis of Principal Components (CAP) analyses showed that while significant community shifts were detected when transitioning from shallow water to mesophotic depths, relative abundance and species richness were highest in < 30 m habitats. Two functional groups (mobile invertivores, generalist macropiscivores) were notable exceptions, having higher abundance and richness values in mesophotic depths. Multivariate regression trees (MRT) and distance-based linear modeling (DistLM) indicated that depth, habitat complexity, unconsolidated sediment (sand), and macroalgae act as structural reef fish assemblage drivers, with indicator species assigned to specific environmental node breaks. There was also evidence of depth-based shifts in the composition and abundance of endemic species, and mesophotic refugia for target species belonging to several functional groups (using length-based kernel density estimates). Finally, I discovered a mesophotic ground fish interface in Halimeda beds that are utilized by juvenile Pristipomoides filamentosis.
Chapter 4 focuses exclusively on reef-associated generalist macropiscivores (roving predators) from shallow to mesophotic depths, and includes BRUVS assessments from both the (populated) MHI and the (unpopulated) NWHI. Through a combination of tests (Bootstrapping, PERMANOVA, Hierarchical Clustering, SIMPER, non-parametric quantile regression splines), I found clear shifts in roving predator community composition, and differences in relative abundance between regions and depth zones. For example, I observed that there was up to an order of magnitude more jacks and five times more sharks sampled in the NWHI compared to the MHI. In addition, differences in target species length-distributions between depth zones and regions were examined using non-parametric Kolmogorov-Smirnov tests, which provided evidence for;
1.) Depth-based predator refuges around populated areas in the MHI – e.g. that target fishes tended to be larger in deeper waters. This assertion was complemented by stark differences between shallow and mesophotic predator abundances in the MHI (e.g. Caranx ignobilis solely encountered in mesophotic zones); and
2.) The prevalence of smaller Carcharhinus galapagensis and Caranx ignobilis in < 30 m depths in the NWHI suggests possible body size and depth segregation, or potential avoidance of intra- or inter-specific predation pressures in deeper waters.
In Chapter 5, I investigate the scope for diver-based underwater visual censuses to properly assess the abundance of large-bodied, roving predators across the Hawaiian Archipelago. Specifically, there are recurring, long-standing concerns that (i) divers are typically limited to only a narrow part of the depth range used by these species; and (ii) acquired behavioral differences of roving predatory species – such as avoidance of divers at fished locations – seriously and substantially bias estimates of relative abundance between populated and remote locations. Here, I compared roving predator abundance estimates gathered by shallow water point-count and towed-diver surveys in < 30 m against data collected using RUVS and BRUVS across the Hawaiian Archipelago, including comparisons among datasets collected in shallow depths exclusively, as well as with data from RUV-BRUV spanning 0 – 100 m. The ‘boot’ and ‘boot.ci’ functions were used from the boot package in R, Version 3.3.0, with 10,000 iterations to produce adjusted 95% confidence intervals (type = “bca”) for each diver census and video assessment method used. Although RUVS and BRUVS data showed significantly higher roving predator densities in the remote NWHI, they also generated substantially lower NWHI:MHI ratios than were generated from diver data – i.e. significantly decreased scales of difference. Comparative example include (but are not limited to) the snapper Aprion virescens (SPC NWHI:MHI ratio was 62:1, towed-diver 24:1, RUVS 5:1, BRUVS 3:1), reef sharks (SPC 142:1, towed-diver 76:1, RUVS 20:1, BRUVS 11:1), and jacks (e.g. trevallies: SPC 5:1, towed-diver 17:1, RUVS 2:1, BRUVS 2:1). Clearly, these results corroborate concerns about the limitations of data that can be readily collected by divers for those groups of fishes, and emphasizes the need to further evaluate diver-based predator estimates in order to foster effective management policies for predatory species.
In conclusion, I demonstrate that BRUVS are an appropriate alternative and/or complement to diver-based visual censuses that provide a means to gain a more holistic representation of reef fish communities and higher-level predator groups. As a relatively new approach to US coral reef fish assessment strategies, the use of BRUVS could be further expanded to include comparatively understudied mesophotic coral reefs and other associated habitats in the Hawaiian Archipelago, along with other poorly studied deep-water regions in the US Pacific Territories (Pacific Remote Islands, American Samoa, Guam and the Commonwealth of the Northern Marianas). Lastly, the extension of BRUVS surveys to other areas hosting large, sustained roving predator populations would allow for a more nuanced assessment of the plausibility of ‘inverted biomass pyramids’ as part of future monitoring procedures and management approaches.