High-Resolution Acoustic Cameras Provide Direct and Efficient Assessments of Large Demersal Fish Populations in Extremely Turbid Waters

Abstract

Monitoring fish species populations in very turbid environments is challenging. Acoustic cameras allow work in very poor visibility but are often deployed as a fixed observation point, limiting the scope of the survey. A BlueView P900-130 acoustic camera was deployed in rocky marine habitats off the coast of French Guiana in order to assess the total abundance, size structure and spatial distribution of a demersal fish population. The relevancy of using an acoustic camera to achieve these three objectives was evaluated by comparing acoustic data to those obtained from fishing surveys. The detection and identification of large demersal fish species were possible with the shape and size of the acoustic signal and acoustic shadow silhouette as well as swimming behavior. Mobile surveys combined with stationary surveys increased the probability of distinguishing individuals from inanimate objects. Estimated total length based on the acoustic signal underestimated the actual length of fish measured on deck, but the data showed the same trends in spatial and temporal variation. Acoustic cameras overcame the extreme lack of visibility by increasing knowledge of fish use of habitat, therefore providing much more efficiency in the effort, more accurate data on the abundance, size structure and spatial distribution than the fishing method. Thus, despite few limitations, acoustic camera surveys are far superior to fishing surveys in evaluating large demersal fish stock status.

Monitoring fish using imaging sonar: Capacity, challenges and future perspective

“…High counting accuracy is often achieved due to the high‐quality data obtained and the single swimming direction of migrating fishes. In addition, MFLS is also widely used in surveys of habitats, such as artificial reefs (Guo et al, 2018; Plumlee et al, 2020), estuary (Becker et al, 2016; Becker et al, 2017; Lankowicz et al, 2020), river (Hayes et al, 2015; Kerschbaumer et al, 2020; Mora et al, 2015), lake (Jing, Han, Wang, et al, 2018a), estuarine shoreline (Smith et al, 2021), offshore habitat (Artero et al, 2021; Tassetti et al, 2020; Van Hal et al, 2017), reservoir (Huang & Gong, 2020; Mo et al, 2015; Shen et al, 2018), etc., to study the fish stock and interaction of fishes with the habitats. The fish density is typically calculated to estimate abundance and studied for the habitat population distribution (Plumlee et al, 2020; Zhou et al, 2014).…”

“…The ability of imaging sonar to monitor fish in turbid water has been highlighted by various researchers (Artero et al, 2021; Becker, Whitfield, et al, 2011b; Faulkner & Maxwell, 2020). Artero et al (2021) utilized BlueView sonar (900 kHz) to successfully assess fish abundance, length and spatial distribution in one of the world’s most turbid region French Guiana. Becker, Whitfield, et al (2011b) found the data quality was not appreciably different when the turbidity ranged from 5 to 28 nephelometric turbidity units (NTU).…”

“…Grothues et al (2016) and Shahrestani et al (2017) employed high‐frequency MLFS to observe the abundance around piers where the traditional methods are hard to apply, guiding the pier design. The ability of imaging sonar to monitor fish in turbid water has been highlighted by various researchers (Artero et al, 2021; Becker, Whitfield, et al, 2011b; Faulkner & Maxwell, 2020). Artero et al (2021) utilized BlueView sonar (900 kHz) to successfully assess fish abundance, length and spatial distribution in one of the world’s most turbid region French Guiana.…”

“…The camera‐based and hydroacoustic systems are the two major nonintrusive approaches for monitoring fish. As the light attenuates rapidly in water, the camera‐based system can only sample a limited water area and is restricted by the water visibility (Artero et al, 2021). Hydroacoustic systems address this limitation by detecting fishes with sounds that propagate a much longer distance in the water than light (Martignac et al, 2015).…”

Fish Species Identification on Low Resolution - A Study with Enhanced Super Resolution Generative Adversarial Network (ESRGAN), YOLO and VGG-16

“…Many researchers take photos in a controlled laboratory environment using digital cameras or cell phones, then construct databases [26]. The quality of the obtained pictures determines the effectiveness of a fish classification system; however, water turbidity has been a key issue impacting the quality of the acquired images [27]. When the turbidity of the water affects the vision of the fish, several researchers have proven experimentally that frontal lighting with backlight images can produce rather acceptable results [28].…”