Efficiency of automatic analyses of fish passages detected by an acoustic camera using Sonar5-Pro

Abstract

The acoustic camera is a non-intrusive method increasingly used to monitor fish populations. Acoustic camera data are video-like, providing information on fish behaviour and morphology helpful to discriminate fish species. However, acoustic cameras used in long-term monitoring studies generate a large amount of data, making one of the technical limitations the time spent analysing data, especially for multi-species fish communities. The specific analysis software provided for DIDSON acoustic cameras is problematic to use for large datasets. Sonar5-Pro, a popular software in freshwater studies offers several advantages due to its automatic tracking tool that follows targets moving into the detection beam and distinguishes fish from other targets. This study aims to assess the effectiveness of Sonar5-Pro for detecting and describing fish passages in a high fish diversity river in low flow conditions. The tool's accuracy was assessed by comparing Sonar5-Pro outputs with a complete manual analysis using morphological and behavioural descriptors. Ninety-eight percent of the fish moving into the detection beam were successfully detected by the software. The fish swimming direction estimation was 90% efficient. Sonar5-Pro and its automatic tracking tool have great potential as a database pre-filtering process and decrease the overall time spent on data analysis but some limits were also identified. Multi-counting issues almost doubled the true fish abundance, requiring manual operator validation. Furthermore, fish length of each tracked fish needed to be manually measured with another software (SMC). In conclusion, a combination of Sonar5-Pro and SMC software can provide reliable results with a significant reduction of manpower needed for the analysis of a long-term monitoring DIDSON dataset.