A Semiautomated Approach to Estimating Fish Size, Abundance, and Behavior from Dual-Frequency Identification Sonar (DIDSON) Data KEVIN M. BOSWELL* Department of Oceanography and Coastal Sciences, School of the Coast and Environment, Louisiana State University, Baton Rouge, Louisiana 70803, USA MATTHEW P. WILSON SonarData Pty., Ltd., General Post Office Box 1387, Hobart Tasmania 7001, Australia JAMES H. COWAN,JR. Department of Oceanography and Coastal Sciences, School of the Coast and Environment, Louisiana State University, Baton Rouge, Louisiana 70803, USA Abstract.—We present a semiautomated analytical ap- proach incorporating both image and acoustic processing techniques to apply to dual-frequency identification sonar (DIDSON) data. Our objectives were (1) to develop a standardized analysis pathway in order to reduce the effort associated with counting, measuring, and tracking fish targets; and (2) to empirically obtain estimates of basic target information (e.g., size, abundance, speed, and direction of travel). Analyses were conducted on DIDSON data collected at three different locations (the Kenai River, Alaska; Mobile River, Alabama; and Port Fourchon, Louisiana) with different equipment and deployment configurations. We developed an efficient postprocessing approach that can be applied to a variety of data sets, independent of user and deployment method. For two of the three data sets analyzed, the estimates of fish abundance derived from DIDSON analyses were not significantly different from the manual counts of DIDSON files. The analyses produced estimates of mean fish length, direction and speed of travel, and target surface area for all targets within each data set. A consistent analysis platform increases the acceptance and reliability of the DIDSON as a tool for fisheries surveys and further demonstrates the usefulness of DIDSON technology in fisheries applications. Recent developments in sonar imaging have provid- ed a means to obtain near-video-quality imaging of fish in dark or turbid waters (Moursund et al. 2003; Tiffan et al. 2004; Mueller et al. 2006). Originally developed for naval surveillance, the dual-frequency identification sonar (DIDSON; Sound Metrics Corp.) has been adopted by fishery scientists to obtain both size and abundance estimates of fish (Moursund et al. 2003; Holmes et al. 2006; Mueller et al. 2006; Burwen et al., 2007) and to image fish habitats (Tiffan et al. 2004). In addition, the behavior of fish relative to habitat and other stimuli can be observed regardless of ambient light levels and turbidity, providing an important advantage over traditional video census techniques (Willis et al. 2000; Stoner 2004). A unique feature of the DIDSON is the capacity to simultaneously image both substrate and other habitats and ensonified fish within the same transmitted pulse, yielding data that are more straightforward and interpretable than those obtained by other methods. Although traditional acoustic techniques (e.g., sin- gle-beam and split-beam techniques) are often used in fishery assessments, the interpretation and classifica- tion of data are often challenging and require extensive experience and effort (Jech and Michaels 2006). These acoustic systems are more susceptible to boundary effects, turbulence, and background noise than is the DIDSON, particularly when attempting to enumerate or identify fish near scattering boundaries (Holmes et al. 2006; Boswell et al. 2007a). Furthermore, when accounting for noisy data, postprocessing of data obtained from an echo sounder can be very compli- cated and time consuming, often necessitating complex analyses (Simmonds and MacLennan 2005; Holmes et al. 2006; Boswell et al. 2007b). Current capabilities for both handling and processing DIDSON data are limited and lack the functionality needed to adequately support the growing number of DIDSON users. To date, no methods have been devised that allow combined acoustic and image processing techniques to analyze and provide quanti- tative estimates of fish abundance, size and behavior. The integration of a defined pathway by which to postprocess and classify acoustic data from the DIDSON will both standardize and objectify the outcome, reducing the need to transfer expertise and biases in interpretation (Jech and Michaels 2006). *Corresponding author: kboswe1@lsu.edu Received June 21, 2007; accepted November 29, 2007 Published online May 8, 2008 799 North American Journal of Fisheries Management 28:799–807, 2008 Ó Copyright by the American Fisheries Society 2008 DOI: 10.1577/M07-116.1 [Management Brief]