In Alaska, sonar estimates of inriver passage often provide the basis for estimating spawning escapement and for regulating sport fish and commercial harvests of important salmon stocks. The Alaska Department of Fish and Game has been evaluating the potential for DIDSON data to provide improved discrimination of larger Chinook from smaller but more abundant sockeye salmon based on length measurements taken directly from the DIDSON images. The results have been promising. However, making the measurements manually is very labor intensive and fatigue can become a serious issue.
DIDSON movie clip of one large Chinook salmon followed by a group of three sockeye-sized salmon.
The goal of this project has been to obtain fish length estimates from DIDSON data with a method that is faster and requires fewer user interactions than manual measurements but at the same time produces comparable estimates of Chinook salmon abundance in a mixture model. Our approach has been to automate the process as much as possible while allowing for enough human supervision to ensure a high accuracy of the results.
The left window shows a DIDSON image of one Chinook and one sockeye-sized salmon. In the right window the same image is overlaid with the outline of the automatically detected fish.
To make the most of existing software the semi-automated analysis uses a combination of two software packages: DIDSON Control and Display Software (Sound Metrics) and Echoview (Myriax). The DIDSON Control and Display Software executes a pre-processing step that removes the image background and empty frames (frames without fish) from the ddf file. Echoview is used to detect and track the fish images and estimate fish length. Part of the Echoview process has been automated with the Kenai EV Auto Processor which uses a template that has been tailored to Kenai River data.
The method has been developed and tested with DIDSON data collected with a long-range system equipped with a high resolution lens. We derived auto length estimates for a total of nearly 2,000 fish in 84 x 20-minute periods and 72 x 10-minute periods. The resulting point estimates of Chinook salmon abundance, as well as the precision of these estimates, agreed very well with estimates that were derived from manual length measurements made with the Sound Metrics DIDSON Control and Display Software.
Length echogram of DIDSON data. The color of the echoes represents the length of the detected fish on a given frame. Fish whose echo traces are shown in warmer colors, for example, orange and red, are larger than fish whose echo traces are blue or green. The vertical dashed line indicates the DIDSON frame shown above. Note that the large fish shows up as a red trace while the smaller fish appears blue and green. The consistency of the colors provides fast feedback on the accuracy of the fish detection and length determination process.
Angle echogram of DIDSON data. The color of the echoes represents the angle at which the fish has been detected. Fish that swim upstream leave echo traces that run from blue to red. Fish that move downstream leave echo traces that run from red to blue. The vertical dashed line indicates the current DIDSON frame shown above. Note that the smaller fish on the current frame is moving downstream while the large fish is swimming upstream. The progression of the angle colors provides fast feedback on the accuracy of the fish tracking process.
References
Semi-automated DIDSON data analysis for the assessment of Kenai River Chinook salmon.
Field manual prepared for the Alaska Department of Fish and Game. Download pdf (1MB)
Video 1 - Echoview Basics and Routine Processing Workspace.
Narrated video tutorial prepared for the Alaska Department of Fish and Game. Download avi (47MB zipped)
Video 2 - Echogram Interpretation.
Narrated video tutorial prepared for the Alaska Department of Fish and Game. Download avi (48MB zipped)
Video 3 - Fish Tracking - Basics.
Narrated video tutorial prepared for the Alaska Department of Fish and Game. Download avi (88MB zipped)
Burwen, D.L., S.J. Fleischman, and J.D, Miller. 2010. Accuracy and precision of salmon length estimates taken from DIDSON sonar images. Transactions of the American Fisheries Society. In Press.
Burwen, D.L., S.J. Fleischman, and J.D, Miller. 2007. Evaluation of a dual-frequency imaging sonar for detecting and estimating the size of migrating salmon. Alaska Department of Fish and Game, Fishery Data Series 07-44, Anchorage.
Fleischman, S. J. and D. L. Burwen. 2003. Mixture models for the species apportionment of hydroacoustic data, with echo-envelope length as the discriminatory variable. ICES Journal of Marine Science. Volume 60, Issue 3, Pages: 592-598.