Acoustic identification of female Steller sea lions(Eumetopias jubatus)a)Gregory S. Campbellb)Cetacean Behavior Laboratory, College of Sciences, San Diego State University, 5500 Campanile Drive,San Diego, California 92182-4611Robert C. GisinerCode 342, Office of Naval Research, 800 North Quincy Street, Arlington, Virginia 22217David A. HelwegCode D3501E, Space and Naval Warfare Systems Center San Diego, 53560 Hull Street, San Diego,California 92152-5000Linda L. MiletteMarine Mammal Research Unit, Fisheries Center, University of British Columbia, 2204 Main Mall,Vancouver, British Columbia V6T 1Z4, Canada共Received 30 May 2001; revised 28 January 2002; accepted 27 February 2002兲Steller sea lion 共Eumetopias jubatus兲 mothers and pups establish and maintain contact withindividually distinctive vocalizations. Our objective was to develop a robust neural network toclassify females based on their mother-pup contact calls. We catalogued 573 contact calls from 25females in 1998 and 1323 calls from 46 females in 1999. From this database, a subset of 26 femaleswith sufficient samples of calls was selected for further study. Each female was identified visuallyby marking patterns, which provided the verification for acoustic identification. Average logarithmicspectra were extracted for each call, and standardized training and generalization datasets created forthe neural network classifier. A family of backpropagation networks was generated to assess relativecontribution of spectral input bandwidth, frequency resolution, and network architectural variablesto classification accuracy. The network with best overall generalization accuracy 共71%兲 used aninput representation of 0–3 kHz of bandwidth at 10.77 Hz/bin frequency resolution, and a 2:1hidden:output layer neural ratio. The network was analyzed to reveal which portions of the callspectra were most influential for identification of each female. Acoustical identification ofdistinctive female acoustic signatures has several potentially important conservation applications forthis endangered species, such as rapid survey of females present on a rookery. © 2002 AcousticalSociety of America. 关DOI: 10.1121/1.1474443兴PACS numbers: 43.80.Ka, 43.80.Lb, 43.64.Tk 关WA兴I. INTRODUCTIONSteller sea lions were selected for this study becausebiological theory predicts that colonially breeding individu-als will possess refined means of individual identification inorder to invest finite resources in a manner most likely toenhance their reproductive success 共Trivers, 1972兲. Indi-vidual identification is potentially possible for any individualof any species; however, evolutionary theory predicts that theneed to reduce confusion over reproductive investment willresult in more obvious and consistent communication of in-dividual identity in colonially breeding species 共Beecher,1982兲. The task of developing a mathematical means of dis-criminating amongst individuals is thus more likely to suc-ceed with a colonially breeding species than with a specieswhere opportunities for mistaking identity and misplacingcompetitive, breeding, or parental resources are less frequentand therefore less consequential.Sea lions and fur seals 共Otariids兲 typically demonstrate apolygynous breeding system, with females gathering indense congregations to give birth and to copulate 共Riedman,1990兲. Mothers alternate between nursing their pups on landand feeding at sea, sometimes leaving the pup unattended forseveral days 共Riedman, 1990兲. Acoustic mother-pup callrecognition appears to be the primary means of initiatingcontact and orienting searching behavior after a period ofseparation 共Insley, 1989, 1992; Gisiner and Schusterman,1991; Schusterman et al., 1992a; Phillips and Stirling 2000,2001兲. Playback experiments have provided additional evi-dence to support the existence of mother-pup acoustic recog-nition systems. Studies on both the subantarctic fur seal 共Arc-tocephalus tropicalis兲 and Galapagos fur seal 共A.galapagoensis兲 have suggested that mothers and pups reactpositively to recordings of each others’ calls but not to thoseof strangers 共Trillmich, 1981; Roux and Jouventin, 1987兲.While individual call recognition among Otariids appears tobe common, few studies have used acoustical and statisticala兲Portions of this work were presented in ‘‘Acoustic identification of femaleSteller sea lions,’’ Proceedings of the 140th meeting of the ASA/NoiseCongress, Newport Beach, CA, November 2000, and ‘‘Neural networkclassification of individual female Steller sea lions 共Eumetopias jubatus兲,’’Proceedings of the 14th Biennial Conference on the Biology of MarineMammals, Vancouver, BC, Canada, November 2001.b兲Electronic mail: [email protected] J. Acoust. Soc. Am. 111 (6), June 2002 0001-4966/2002/111(6)/2920/9/$19.00 © 2002 Acoustical Society of Americaanalysis techniques to evaluate call stereotypy. Only studiesof the northern fur seal Callorhinus ursinus 共Insley, 1989,1992兲, South American sea lion 共Ferna´ndez-Juricic et al.,1999兲, and South American fur seal Arctocephalus australis共Phillips and Sterling, 2000兲 have classified known femalesbased on the acoustic structure of their calls.The Steller sea lion ranges along the North Pacific Rimfrom Japan to Central California, with centers of abundanceand distribution in the Aleutian Islands and the Gulf ofAlaska 共Loughlin et al., 1984; Riedman, 1990兲. In responseto population declines in Alaska since the 1970s, the Stellersea lion was listed in 1990 as threatened under the U.S.Endangered Species Act 共NMFS, 1992兲. Genetic and distri-bution information subsequently led to the classification oftwo separate breeding stocks of Steller sea lions: a ‘‘west-ern’’ stock, which included sea lions in the Gulf of Alaska,Aleutian Islands, and Bering Sea, and an ‘‘eastern’’ stockwhich included sea lions from Southeast Alaska to California共NMFS, 1995; Bickham et al., 1996; Loughlin, 1997兲.In1997, the western stock of Steller sea lions was relisted asendangered under the U.S. Endangered Species Act. Thisreclassification was in response to a dramatic population de-cline during the past 30 years, estimated at over 80%共Braham et al., 1980; Merrick et al., 1987; Loughlin et al.,1992; Trites and Larkin, 1996; Sease et al., 2001兲. Duringthis same time period, Steller sea lion numbers have re-mained stable or increased in Southeast