Combined video–hydroacoustic survey of nearshore semi-pelagic rockfish in untrawlable habitats

Abstract

New survey technologies are needed to survey untrawlable habitats in a cost-effective and nonlethal manner with minimal impacts on habitat and nontarget species. Here, we test the efficacy of integrating data from a suspended underwater camera with acoustic data to generate population estimates for nearshore Black (Sebastes melanops), Blue (Sebastes mystinus), and Deacon Rockfish (Sebastes diaconus). We surveyed Seal Rock Reef near Newport, Oregon, and compared our results to population estimates derived from a mark–recapture study conducted at the same reef. We compared fish density estimates from video deployments to those calculated from applying published target strength to length regression models to our acoustics data. Densities derived from the acoustics, using a generalized physoclist target strength to length model, were significantly different from densities derived from video; conversely, a rockfish-specific target strength to length model generated densities that were not statistically different from video densities. To assess whether, and how, fish behaviour was influenced by the presence of an underwater camera, we deployed our camera under the acoustic transducer. No statistical difference was observed in the acoustic density of fish before, during, or after camera deployment. Our work suggests that combining acoustic and stereo video data provided a similar population estimate to historic survey results, but an accurate acoustic density estimate was dependent on using the proper acoustic target–strength model. We contend that combining camera data with hydroacoustic data is effective for surveying rockfish in untrawlable habitats.

Estimating abundance of fish associated with structured habitats by combining acoustics and optics

“…trawls, traps, line fishing) are not amenable (e.g. Boldt et al, 2018; Rasmuson et al, 2022). It combines the benefits of rapid acoustic sampling, which allows for the enumeration of schooling fish as well as other fish, with the benefits of optics to determine species composition.…”

“…The use of RUVs in acoustic–optical surveys presents challenges including time‐consuming image analysis, visibility, identification of small or distant fish and limited sampling. To reduce annotation time and the likelihood of observing fish not detected by acoustics, we analysed the first 120 s of video footage, consistent with Rasmuson et al (2022). To increase sample size, acoustic–optical surveys should determine appropriate sample sizes for characterizing species proportions, and consider compact drop cameras as an alternative to traditional RUV platforms (e.g.…”

“…To increase sample size, acoustic–optical surveys should determine appropriate sample sizes for characterizing species proportions, and consider compact drop cameras as an alternative to traditional RUV platforms (e.g. Boldt et al, 2018; Fernandes et al, 2016; Rasmuson et al, 2022). The efficiency of extracting data from video or drop camera samples may also be improved by adopting deep learning methods of species identification and counting, as has been demonstrated for C. auratus in Shark Bay (Connolly et al, 2021; Marrable et al, 2022).…”

“…The first 2 minutes of video camera recordings (Rasmuson et al, 2022), from the time the RUVs hit the bottom, were used to partition the acoustic data, minimizing temporal and spatial mismatch. SeaGIS EventMeasure (SeaGIS, 2021) was used to analyse video footage from the left‐side camera.…”