We used pattern recognition algorithms and image processing to identify individual Chinook salmon Oncorhynchus tshawytscha. Using melanophore spot patterns located on the dorsal head region, algorithms ranked all database images against each other. We coupled this technology with a graphical user interface to visually confirm or reject top-ranked algorithm results and tested this process on 295 juvenile Chinook salmon in seven photo sessions over a 251-d period. Juveniles began developing spots, identifiable in photo images, between 167 and 197 d after conception (52.7-mm fork length [FL]). Unique spot patterns appeared 197-232 d from conception, beginning at approximately 104-mm FL. Of 254 fish surviving the experimental period, 106 (42%) demonstrated identifiable patterns, 102 (40%) developed spots but patterns were insufficient for identification, and 46 (18%) exhibited a complete lack of spots. Spot patterns continued developing on individual fish by study end. On average, fish that developed recognizable spot patterns did so at approximately 140-mm FL. Once they did, reidentification was 100% correct in up to four subsequent trials. Patterns remained identifiable even after a 25-32% size increase over a 55-d period and as juveniles went through smoltification. Although patterns occurred at sizes typically larger than salmon observed at some California Central Valley monitoring locations, this technique provides a potentially valuable, noninvasive method of identifying individual salmon during emigration. Improved image collection techniques and use of body areas exhibiting identifiable patterns at earlier developmental stages may increase fish available for pattern identification. These results demonstrate the indexing of a large database using pattern recognition algorithms for Chinook salmon. The utility of such an approach may be valuable for addressing specific biological questions associated with massproduced (wild and hatchery), migratory salmonids, especially as individuals develop, grow, and move through the various habitats available to them.

A sonar system was designed and built that could transmit and receive a broadband signal. This transmission takes the form of a chirp, ranging in frequency from 50–150 kHz. Four species of fish represented by ten size classes were used in a laboratory environment as targets. An acoustic signature was developed for each of the species and each of the size classes within each species. Pattern analysis techniques were then employed to test unknowns against these signatures. The system as a whole was capable of 95% accuracy in both species and size class discrimination. (Those fish misclassified to specie were considered to be misclassified to size irrespective of the relative sizes of the signature and unknown.) Size-class accuracy is approximately ± 10% of the fork length of the fish. Hydroacoustic and pattern recognition techniques are presented.