Starr et al.: Submersible-survey and acoustic-survey estimates of fish density 



121 



however, documented high variability in trawl sur- 

 vey estimates of rockfish abundance and implied that 

 trawl surveys may not adequately estimate rockfish 

 abundance. Krieger ( 1993) compared trawl and sub- 

 mersible estimates of rockfish abundance on flat bot- 

 toms and came to the same conclusion. He suggested 

 that on flat terrain, trawl surveys overestimate rock- 

 fish abundance because bridles and otter doors herd 

 rockfish into a trawl and bias density estimates. 

 Adams et al. ( 1995) found remotely operated vehicle 

 (ROV) surveys to be better than trawl surveys for 

 benthic species on flat bottoms. They reported that 

 ROV estimates of fish density were higher and had 

 lower coefficients of variation than did trawl esti- 

 mates. For rockfishes swimming off the bottom, how- 

 ever, the reverse was true. Trawls yielded better es- 

 timates of rockfish density and densities of other 

 species with patchy distributions and off-bottom be- 

 havior. Despite the improvement in estimates, Adams 

 et al. (1995) suggested that neither ROV nor trawl 

 methods adequately assessed off-bottom rockfishes. 

 Both Krieger ( 1993) and Adams et al. ( 1995) acknowl- 

 edged that trawls are poor tools for assessing rock- 

 fishes on high-relief terrain. Kulbicki and Wantiez 

 (1990) compared trawl surveys with diver observa- 

 tions and determined that trawl surveys and direct 

 observations each have biases that are dependent 

 upon habitat usage by different fishes. In their study, 

 species size, shape, coloration, and swimming hab- 

 its greatly influenced the ratio of diver to trawl density 

 estimate. Thus, a combination of survey methodologies 

 is probably needed to estimate adequately the abun- 

 dance of many fish species. A similar conclusion was 

 reached by Uzmann et al. ( 1977 ) in comparing submers- 

 ible, camera sled, and otter trawl techniques. 



Acoustic target strength analysis 



The target strength of a fish is dependent upon a 

 variety of factors, including the size of the fish, its 

 orientation to the acoustic signal, and its swim- 

 bladder characteristics (Ehrenberg and Lytle, 1977). 

 Of particular importance is the orientation of the fish 

 to the acoustic signal. Small changes in the tilt angle 

 of a fish caused by differences in fish behavior can 

 cause large changes in target strength. Traynor and 

 Williamson (1983), for example, estimated a 3-dB 

 difference in target strength of fishes due to day-night 

 differences in behavior. Dual-beam echo processing 

 methods can resolve many of the problems associ- 

 ated with differences in fish orientation (Traynor and 

 Williamson, 1983) but require the resolution of indi- 

 vidual targets. Dual-beam methods cannot estimate 

 target strength in the case of overlapping echoes, 

 such as those produced by schooling fish. 



We chose to use the mean length of fish observed 

 (Love's equation) instead of dual-beam methods to 

 scale echo integrator output for two reasons. First, 

 the number of individual, nonoverlapping targets 

 needed for dual-beam analysis was relatively small 

 in several of the acoustic transects (Table 2). Sec- 

 ond, Traynor et al. (1990) suggested that target 

 strength estimates of schooling fish may not accu- 

 rately reflect the actual size offish insonified because 

 the equipment measures fish on the periphery of the 

 schools. Fish on the periphery of schools may not be 

 the same size as fish in the center of the school or 

 may be exhibiting different behavior (orientation). 



Although Love's equation was the primary means 

 of scaling echo integrator data, we calculated dual- 

 beam target strength as well. Target strength esti- 

 mated from Love's equation included schooling fish. 

 Dual-beam analysis, however, included only nonover- 

 lapping echoes, i.e. those fish not in a school. At sta- 

 tions with few schools, the back-scattering cross sec- 

 tions obtained from dual-beam analysis were almost 

 identical to the back-scattering cross sections derived 

 from the mean length of observed fishes (Table 4). At 

 station 2, where large schools of juvenile rockfishes 

 occurred, the back-scattering cross section estimated 

 from dual-beam methods was 4.8 times higher than 

 estimates from Love's equation. When the mean 

 length of only nonschooling fishes was used in Love's 

 equation at station 2, the dual-beam estimate and 

 Love's equation estimate of target strength were more 

 similar. 



In this study, a combination of two survey meth- 

 ods provided a better estimate of the distribution and 

 relative abundance of rockfishes than did either 

 method alone. Submersible surveys yielded estimates 

 of fish density near the bottom as well as informa- 

 tion used to provide ground truth for acoustic sur- 

 veys and to scale echo integrator values. Acoustic 

 equipment enabled portions of the water column not 

 observed on submersible transects to be surveyed and 

 provided additional information on the vertical and 

 horizontal distribution of fishes. 



Acknowledgments 



Funding for submersible work was provided by 

 NOAA's West Coast National Undersea Research 

 Center, Fairbanks, Alaska. We thank the crew of the 

 DSV Delta and the crew of its support ship for their 

 work at sea. Bob Hannah, Mary Yoklavich, and two 

 anonymous reviewers provided helpful reviews of the 

 manuscript. This paper was funded in part by a grant 

 from the National Sea Grant College Program, Na- 

 tional Oceanic and Atmospheric Administration, U.S. 



