Transactions of the American Fisheries Society 132:746±758, 2003 ᭧ Copyright by the American Fisheries Society 2003

Paddle®sh as Potential Acoustic Targets for Abundance Estimates

R. SCOTT HALE* Ohio Department of Natural Resources, Division of Wildlife, 10517 Canal Road Southeast, Hebron, Ohio 43025, USA

JOHN K. HORNE University of Washington, School of Aquatic and Fishery Sciences, Box 355020, Seattle, Washington 98198-5020, USA

DONALD J. DEGAN Aquacoustics, Inc., 29824 Birdie Haven Court, Sterling, Alaska 99672-1473, USA

M. ELIZABETH CONNERS National Marine Fisheries Service, Alaska Fishery Science Center, 7600 Sand Point Way Northeast, Seattle, Washington 98115-6349, USA

Abstract.ÐUnderwater acoustics is a noninvasive sampling technique that potentially reduces expense and injury to target species, but this method may be underutilized for sampling large freshwater ®shes. We measured target strength (TS), developed anatomically based backscatter models, and conducted gill-net and acoustic surveys of paddle®sh Polyodon spathula to explore the potential use of acoustic surveys for estimating the abundance of large freshwater ®shes. Mean TS measured from two size-groups of paddle®sh at 200 kHz was Ϫ37.14 decibels (dB; SD ϭ Ϫ2.36) for age-0 ®sh (353±406 mm) and Ϫ27.25 dB (SD ϭϪ2.21) for adult ®sh (1,018±1,284 mm), indicating that TS could differentiate these size-groups. Backscatter models identi®ed strong contributions of the swim bladder to TS and revealed the sensitivity of acoustic backscatter to paddle®sh length, aspect, and acoustic carrier frequency. Model results were generally within one SD of measured means from individual ®sh of each size-group. Target strength results were used to count two populations of adult paddle®sh in mobile surveys using an echo sounder with a 200- kHz, 6Њ split-beam transducer. One population was stocked in 1.6-ha Hebron Pond, where no large ®sh were previously present. The other population resided in 28-ha Horseshoe Lake, an Ohio River backwater. Twenty-one paddle®sh stocked in Hebron Pond were accurately counted during the ®rst of six side-looking surveys, but subsequent surveys only counted between two and seven ®sh. Depletion gillnetting results in Horseshoe Lake provided an estimated baseline of 130 Ϯ 55 paddle®sh for comparison with abundance estimates from side-looking and down-looking acoustic surveys during day and night. Acoustic abundance estimates ranged from 187±313 ®sh (side- looking) to 3,464±13,489 ®sh (down-looking) depending on survey time (day or night) and the approach to analysis. Ratio estimates and cluster estimates provided similar results, and the co- ef®cient of variation of the mean (100´SE/mean) ranged from 20% to 50%. Acoustic estimates were either greater or more variable than those derived from depletion gillnetting, yet acoustic surveys required only 6 man-hours compared to 180 man-hours for the gillnetting estimate. Our study is the ®rst to indicate that TS can be used to count adult paddle®sh and that, upon re®nement of survey techniques, TS can be used to estimate paddle®sh abundance. The bene®ts of acoustic surveys may be realized sampling other large freshwater ®shes when the target species can be differentiated with TS and considerations are made for transducer selection.

Acoustics may improve assessment of large ditional insight into their ecology. Within North freshwater ®shes if acoustic target strength (TS) America, published acoustic studies of freshwater can be used to accurately identify and count large ®shes appear dominated by assessment of small ®sh in surveys. Further, acoustic surveys can de- pelagic ®shes (Burczynski et al. 1987; Brandt et termine distribution of large ®shes, providing ad- al. 1991; Degan and Wilson 1995; Schael et al. 1995; Fleischer et al. 1997; Warner et al. 2002) or salmonids (Thorne 1983; Mulligan and Kieser * Corresponding author: [email protected] 1986; Parkinson 1994; Beauchamp et al. 1997; Received April 8, 2002; accepted January 9, 2003 Yule 2000), with few results reported for larger

746 PADDLEFISH ACOUSTICS 747

freshwater ®shes common to inland rivers, lakes, rivers, paddle®sh populations are dif®cult to quan- or reservoirs. This may be the result of limited TS tify because they are very mobile and can be dif- data for larger ®shes or a lack of information re- ®cult to sample adequately. Estimation of paddle- garding how equipment factors or the biology of ®sh populations with mark±recapture methods is large ®shes in¯uence TS. hindered by violations of methodological assump- Target strength is derived from the logarithm of tions of both open and closed population models ␴ backscattering cross section ( bs), or acoustic re- and by a lack of suf®cient sample sizes (Runstrom ¯ectivity from a ®sh, and is strongly in¯uenced by et al. 2001), indicating a need for alternative meth- biological factors and transducer frequency. Fish ods of estimating abundance. In addition, interest length, ®sh behavior, and the presence of a swim in paddle®sh has increased since its protection in bladder are major biological factors in¯uencing 1994 through the Convention on International amplitude of acoustic backscatter, whereas TS gen- Trade in Endangered Species of Wild Fauna and erally increases with sonar carrier frequency and Flora (Appendix II listed species) (Jennings and ®sh size (MacLennan and Simmonds 1992; Horne Zigler 2000). Speci®c objectives of this study were and Clay 1998). The relations between biological to (1) measure TS of known-length paddle®sh, (2) and equipment factors provide insight regarding develop anatomically based acoustic backscatter TS, equipment selection, and use of acoustic sur- models for comparison with acoustic results, and vey data. (3) estimate paddle®sh abundance. Although pad- Differentiation of ®sh size by TS is complex dle®sh are most frequently associated with large because the swim bladder is the primary source of rivers and reservoirs, we conducted our study in acoustic backscatter in most ®shes, and the swim smaller environments to ensure that populations bladder's size, position, and angle in relation to were closed. the transducer in¯uence TS at any frequency (Mac- Lennan and Simmonds 1992; Brandt 1996). Gen- Study Area eralized equations to estimate ®sh lengths from TS Our study was conducted in a pond at the He- have been developed for the side, dorsal, and any bron State Fish Hatchery, Hebron, Ohio, and at aspects (Love 1969, 1971, 1977). These equations Horseshoe Lake, a small lake located along the are widely used, and have proven useful for a va- southern Indiana±Ohio border. Hebron Pond is 1.6 riety of species (Burczynski et al. 1987; Brandt et ha, is roughly circular, has a maximum depth of 4 al. 1991; Yule 2000; Warner et al. 2002). However, m, and contained no large ®sh prior to paddle®sh the empirical, species-speci®c TS±length relations introduction. Horseshoe Lake is a 28-ha excavated more closely estimate true ®sh size due to the com- quarry near the Ohio River. It is steep sided, has plexity of acoustic backscatter (Foote 1987; Mac- a maximum depth of 13 m, and is connected to Lennan and Simmonds 1992; Fleischer et al. the Great Miami River via a 40-m-wide inlet at a 1997). Empirical, species-speci®c TS±length re- point 2 km upstream from the con¯uence of the lations may be particularly important for large ®sh- Great Miami and Ohio rivers. Tagging studies in- es because TS becomes more sensitive to the swim dicate that Ohio River paddle®sh move to and from bladder position within a ®sh and to the ®sh angle Horseshoe Lake (Henley et al. 2001). to transducer as ®sh length increases, especially Hebron Pond and Horseshoe Lake were also at high frequencies (Clay and Horne 1994; Jech et used for abundance experiments. Hebron Pond al. 1995). provided a controlled environment that was easy We initiated this study to determine the potential to sample and had no additional large ®sh present of acoustics to estimate the abundance of a large that might be confused with paddle®sh in our freshwater ®sh species. We sought to determine acoustic surveys. Horseshoe Lake was selected be- whether large ®sh could be identi®ed via TS and, cause paddle®sh had been previously sampled if so, whether abundance could be estimated from there and because the lake inlet could be blocked counts of large ®sh derived from acoustic surveys. with a net to contain paddle®sh. Large ®shes that We used the paddle®sh Polyodon spathula as our might be confused with paddle®sh in Horseshoe study organism because paddle®sh were readily Lake acoustic surveys include bigmouth buffalo available and among the largest ®sh in our region. Ictiobus cyprinellus, smallmouth buffalo Ictiobus Paddle®sh are widely distributed throughout the bubalus, black buffalo Ictiobus niger, common Mississippi River basin and can attain lengths ex- carp Cyprinus carpio, channel cat®sh Ictalurus ceeding 2 m (Jennings and Zigler 2000). As is punctatus, ¯athead cat®sh Pylodictis olivaris, and characteristic of other large ®shes native to large striped bass Morone saxatilis. 748 HALE ET AL.

Methods placed ¯at on white freezer paper for dorsal and lateral tracing, moved to an acrylic cradle, and then Target strength measurements.ÐA 200-kHz radiographed in dorsal and lateral aspect with a BioSonics DT 6000 echo sounder containing a 6Њ Picker International model VTX 1050 X-ray ma- (3-decibel [dB], half-power points) split-beam chine (settings: 51±100 kVp [kilovotage peak], transducer with a source level of 214.1 dB//␮Pa 20±50 mAs [milliamp seconds]). Dorsal and lat- and receiver sensitivity ofϪ53 dB/␮Pa was used eral radiographs were converted to digital data ®les to collect acoustic data. Sample pulse duration was by tracing silhouettes of bodies (not including ®ns) set at 0.4 msec, the pulse rate was set at 10 or 15 and swim bladders, scanning traces to graphics pings/s, and a threshold of Ϫ57 dB was used to ®les, digitizing graphics ®les at 1-mm resolution, ®lter small targets. The system was calibrated at and scaling the images to their true sizes. The axis the BioSonics, Inc., facility in Seattle, Washington, horizontal to the transducer face was de®ned from and then ®eld calibrated with a tungsten carbide the tip of the snout to the middle of the caudal reference sphere following the procedures rec- peduncle (Olsen 1977). ommended by Foote and MacLennan (1984). Backscatter amplitudes of the 10 paddle®sh Target strength was measured on tethered hatch- were estimated with a Kirchhoff-ray-mode (KRM) ery and wild paddle®sh during October 4±6, 1999. backscatter model (Clay and Horne 1994). In cal- Five-month-old paddle®sh ranging in size from culations of backscatter, the ®sh body is repre- 353 to 406 mm standard length (SL) were obtained sented by a set of contiguous, ¯uid-®lled cylinders from the Kentucky State University Aquaculture surrounding a set of gas-®lled cylinders repre- Research Center, Frankfort, Kentucky. Wild ®sh senting the swim bladder. The KRM model com- ranging in size from 1,018 to 1,218 mm SL were putes complex backscattering lengths (Medwin collected from the Ohio River and Horseshoe Lake and Clay 1997) separately for the swim bladder with gill nets. These ®sh were held in ponds at (L ) and ®sh body (L ). Coherent backscatter for Hebron State Fish Hatchery until needed, whereas sb fb the whole ®sh (Lwf) is assumed, and Lfb and Lsb others were collected at Horseshoe Lake on Oc- ϭ ϩ sum as the complex function Lwf Lfb Lsb.A tober 5±6, 1999, and measured on-site. nondimensional measure of echo amplitude de- In preparation for TS measurements, each ®sh rived from backscattering length divided by ®sh was measured (SL) and anesthetized in a bath of length, the reduced scattering length (RSL), is cal- 60 mg/L clove oil (Anderson et al. 1997). Anes- culated from each ®sh as a function of standard thetized ®sh were placed in a gill-net sock and length (L; measured in meters), ®sh aspect relative lowered to 10 m beneath the transducer by use of to the transducer face (␪; measured in degrees), mono®lament line and small weights at the dorsal and acoustic wavelength (␭; measured in meters) aspect. We collected about 1,000-echo amplitude to compare individual components. The acoustic or TS measurements from each ®sh when the in- wavelength is a function of the speed of sound in dividual was near the center of the beam and move- water (c; measured in meters per second) and ment was minimal. Target strengths were estimat- transmitting frequency (¦; measured in hertz) (i.e., ϭ ϫ ␴ ed with the equation TS 10 log10( bs), where ␭ϭc/¦). Full details of the model can be found ␴ bs is the backscattering cross section (i.e., the in Clay and Horne (1994), Jech et al. (1995), or amplitude of echo returned from the ®sh). The null the appendix in Horne and Jech (1999). hypothesis that mean TS of age-0 and adult pad- Individual ®sh were modeled over a length range dle®sh were equal was tested with a pooled- of 300±1,300 mm, an aspect range of 70±110Њ, variance t-test (Snedecor and Cochran 1980) at an and an acoustic frequency range of 12±420 kHz. ␣ value of 0.05. Average and standard deviation backscatter values Acoustic backscatter modeling.ÐAcoustic back- at 200 kHz were calculated for age-0 and adult ϭ ϩ scatter was estimated via digitized dorsal and lat- paddle®sh, converted to TS ( 20´log10[RSL] eral radiographs and an anatomically based nu- 20´log10L), and plotted to examine the effect of meric model. Three age-0 (336±346 mm SL) and length on predicted backscatter. Field measures of seven adult (894±1,277 mm SL), surface-adapted TS were plotted in the same graph for comparison, (swim bladder allowed to adjust to normal surface but separate models were necessary for age-0 and pressure) paddle®sh were transported to the Ohio adult ®sh because data for intermediate-sized ®sh State University Equine and Bovine Veterinary were not available and differences in size were too Clinic for radiographing. Each ®sh was anesthe- great to model these ®sh collectively. tized as in tethering experiments, measured (SL), Abundance estimates.ÐThe potential use of PADDLEFISH ACOUSTICS 749 acoustics to estimate paddle®sh abundance was ex- for analysis of side-looking data (BioSonics 1999). plored in two small water bodies where paddle®sh Down-looking data were analyzed by selecting in- could be contained. The ®rst experiment was con- dividual tracks of ®sh and outputting echo results ducted to determine whether paddle®sh could be to ®les to provide estimated ®sh size, location, and counted with acoustic surveys; therefore, we depth. Side-looking data were similarly analyzed, stocked a known number of paddle®sh in Hebron but VTrack software was used to output tracked Pond and compared results of six consecutive sur- ®sh due to the larger number of ®sh in the dataset. veys. A pond survey was de®ned as one set of data VTrack could not be used for down-looking anal- collected with a side-looking transducer aimed ysis because it is unable to track and exclude bot- from the pond center toward the bank while we tom echoes. drove the boat in a circular pattern. The side- Counts of adult paddle®sh determined from looking transducer was mounted 1 m below the analysis of each survey were examined with a chi- water surface, and data were collected from points square contingency test (Snedecor and Cochran 1 to 20 m from the boat. On October 4, 1999, 21 1980). This test evaluated the null hypothesis that adult paddle®sh (912±1,364 mm SL) were stocked acoustic counts of paddle®sh in side-looking and in the pond. Acoustic sampling began 18 h after down-looking surveys were independent of time ®sh were stocked. Six pond surveys were con- of sampling (day versus night) at an ␣ value of ducted, and 15 min elapsed between the end of 0.05. each survey and the beginning of the next. Each survey result was used to generate ratio The second experiment was conducted in Horse- and cluster sampling abundance estimates. We pro- shoe Lake to explore the estimation of paddle®sh duced ratio estimates by dividing the total number abundance from counts of large ®sh, presumably of ®sh by the total area sampled to estimate density adult paddle®sh, as indicated by TS. A survey in (®sh/m2) and variance. To generate cluster esti- Horseshoe Lake was de®ned as a series of con- mates, we divided the total number of paddle®sh secutive acoustic transects conducted with either detected by the area sampled for each transect, and side-looking or down-looking data collections. then density estimates for survey transects were The population of paddle®sh was con®ned within averaged to produce a mean and variance. Results the lake and sampled with acoustics, and then from both methods were multiplied by the surface acoustic abundance estimates were compared to area of Horseshoe Lake to estimate total paddle®sh depletion gillnetting estimates. The inlet to Horse- abundance. shoe Lake was blocked off with a 25-mm-mesh Gillnetting conducted during October 6±8, net on October 5, 1999, to prevent paddle®sh from 1999, provided what we de®ned as a baseline es- moving into or out of the lake. Mobile acoustic timate of the Horseshoe Lake paddle®sh popula- surveys were conducted during day and night on tion for comparison to acoustic-based abundance October 5 with side-looking and down-looking estimates. Twelve 120-m-long, 4.5-m-deep, 127- sampling at survey speeds of 2.5 m/s. The top 2 mm-bar-mesh gill nets were ®shed for eight 2.5-h m of the water column were sampled by use of periods at ®xed locations. Our gear covered all side-looking surveys with the transducer mounted habitat types and depths due to the small size, steep 1 m below the water surface and with the maxi- banks, and shallow depth of Horseshoe Lake. mum range set to 20 m. Down-looking surveys Some nets were ¯oated at the surface, whereas were used to sample the entire water column from others were ®shed on the bottom. However, nets 2 m below the surface to approximately the bot- often spanned the water column from the surface tom, with the transducer mounted 0.5 m below the to the bottom and generally extended from the water surface and the start range set at 1.5 m. shoreline to one-third or one-half the distance from Targets greater than Ϫ29.46 dB were de®ned as shore to shore. Fishes caught in each net were potential adult paddle®sh based upon tethering ex- identi®ed and measured to the nearest millimeter. periment results. Latitude and longitude were re- The eye±fork length (EF) of paddle®sh was mea- corded during each survey to provide location of sured and converted to SL with the equation SL targets. ϭ EF(1.33) ϩ 73.71, which was derived from our Tracking TS of multiple consecutive pings was data. Total length (mm) was measured for nontar- used to identify and count paddle®sh targets. get species. All ®shes were marked with a hole- BioSonics Analyzer version 3.11 software was punch in the anal ®n. Paddle®sh were removed used for analysis of down-looking data, and from the lake and released in the Ohio River. Non- BioSonics VTrack version 0.98 software was used target ®shes were returned to the lake. The deple- 750 HALE ET AL.

FIGURE 1.ÐTarget strength distribution of measure- ments of age-0 and adult paddle®sh sampled at 200 kHz ϭ (N number of pings). FIGURE 2.ÐSwim bladder (uniformly dashed line), body (dotted and dashed line), and whole-®sh (solid line) components of reduced scattering length in an anatom- tion method was used to estimate the number of ically based acoustic backscatter model developed from paddle®sh in Horseshoe Lake (Ricker 1975). Gill- a 346-mm paddle®sh. This example represents results net bycatch indicated the percentage of large ®sh for 300±1,300-mm ®sh. that might be confused as paddle®sh in acoustic surveys. greater than Ϫ29.46 dB (TS ϩ 1 SD) to identify Results targets as adult paddle®sh in acoustic surveys. Target Strength Measurements This approach provided a conservative estimate of Target strengths were measured for 24 paddle- TS to identify adult paddle®sh and to avoid con- ®sh that we categorized as either age 0 or adult. fusing them with age-0 paddle®sh or other non- Intermediate-length paddle®sh were not available target species during surveys. Individual TS dis- because we could not capture them in the wild or tributions indicated that 98% of age-0 pings were obtain them by other means. Individual echo re- less than Ϫ29.46 dB, whereas 58% of adult pings turns from age-0 ®sh had a wider TS distribution were greater than Ϫ29.46 dB. We also used this than adult paddle®sh, despite minimal variation in approach because identi®cation of a large ®sh for lengths of tethered ®sh in either size-group. Large counting was based on tracking TS of multiple variation among individual echo returns resulted consecutive pings rather than the return from a in overlap of TS between age-0 and adult paddle- single ping. ®sh based on single pings, particularly between Ϫ40 and Ϫ30 dB (Figure 1). However, distribu- Backscatter Models tions of TS were distinctly different between the Anatomically based acoustic backscatter models two length-groups (Figure 1). of paddle®sh were developed for age-0 (336±346 The grand mean of TS from individual age-0 mm; n ϭ 3) and adult (894±1,277 mm; n ϭ 7) paddle®sh was signi®cantly lower than that of ®sh. Preliminary calculations of the ratio of ®sh adult paddle®sh, indicating that the two length- length to acoustic wavelength (L/␭) showed that groups could be differentiated despite some over- L/␭ values ranged from 2.69 (12 kHz) to 96.88 lap in TS distributions (P Ͻ 0.01, t ϭ 19.64, df ϭ (420 kHz) in age-0 paddle®sh and from 7.15 (12 22). Mean TS of four age-0 paddle®sh (353±406 kHz) to 357.56 (420 kHz) in adult paddle®sh. mm SL) ranged from Ϫ39.48 dB (upper SD ϭ These ratios became particularly large for adult Ϫ3.41) to Ϫ34.18 dB (upper SD ϭϪ4.05), with paddle®sh sampled with frequencies greater than a grand mean of Ϫ37.14 dB (SD ϭϪ2.36). Mean 70 kHz. TS of 20 adult paddle®sh (1,018±1,284 mm SL) Paddle®sh had a single, large swim bladder lo- ranged from Ϫ30.87 dB (upper SD ϭϪ3.85) to cated slightly posterior relative to the entire length Ϫ23.14 dB (upper SD ϭϪ3.46), with a grand of the ®sh when the rostrum was included. mean of Ϫ27.25 dB (SD ϭϪ2.21). Kirchhoff-ray-mode backscatter models indicated We assumed that most adult paddle®sh would that the swim bladder was the primary source of provide TS within one SD of the mean based on acoustic backscatter (Figure 2). Figure 2 provides tethering experiment results; therefore, we used TS model results developed from a 346-mm paddle- PADDLEFISH ACOUSTICS 751

®sh to indicate how components of a ®sh body clined precipitously after the ®rst of six pond sur- contribute to RSL for 300±1,300-mm paddle®sh. veys. Although we stocked 21 paddle®sh in the This model demonstrates that acoustic scattering pond, we counted 24 paddle®sh in the ®rst survey, by the swim bladder (Lsb; uniformly dashed line) indicating that some ®sh were counted more than was many times greater than that of the body (Lfb; once, as no other large ®sh were present. In the dotted and dashed line). When coherent backscat- second survey, we counted only seven paddle®sh, ter of the whole ®sh is assumed, components of and in the following four surveys we counted be- returned sound from the swim bladder and body tween two and ®ve paddle®sh. Paddle®sh behavior were summed (Lwf; solid line) to indicate the clearly in¯uenced our ability to count adult pad- strong in¯uence of the swim bladder on amplitude dle®sh with acoustic surveys in Hebron Pond. and of body shape on variability of returned sound. Horseshoe Lake abundance estimates.ÐAcous- Backscatter models for individual ®sh differed tic surveys of Horseshoe Lake identi®ed 13 pad- conspicuously between age-0 ®sh and adults. dle®sh targets during the day and 15 paddle®sh Backscatter amplitude was extremely sensitive to targets at night. The numbers of ®sh counted in ®sh tilt in the age-0 and adult models (Figure 3). side-looking and down-looking surveys were de- Sharp peaks in backscatter amplitude at 85Ϫ95Њ pendant upon time of the survey (P Ͻ 0.01, ␹2 ϭ aspects, which aligned the dorsal surface of the 5.07, df ϭ 1). We counted four ®sh in side-looking swim bladder perpendicular to the transducer (i.e., transects and nine ®sh in down-looking transects horizontal), quickly dissipated to low amplitudes during day surveys, and 11 ®sh in side-looking when ®sh tilted head up or head down. This effect transects and four ®sh in down-looking transects was most pronounced in adults. When ®sh length at night. Side-looking surveys sampled 2.14± was held constant along the L to ␭ axis, an increase 3.68% of the lake, and down-looking surveys sam- in acoustic carrier frequency corresponded to a pled 0.07±0.12% of lake. A larger percentage of greater L/␭ value. Backscatter amplitude increased lake area was sampled with side-looking surveys as L increased when carrier frequency was held than down-looking surveys because side-looking constant. Backscatter model results from adult ®sh surveys could consistently sample a full 20 m from returned stronger peak signals at all lengths than the boat, whereas area sampled in down-looking those derived from age-0 ®sh. Variation in peaks surveys depended upon depth. Maximum average at ®xed aspects re¯ected areas of constructive and beam width was 3.85 m for side-looking surveys destructive backscatter interference between the and 1.36 m (day) and 1.51 m (night) for down- swim bladder and ®sh body. Anatomically based looking surveys. Differences in beam width of models demonstrated that in¯uences of ®sh aspect down-looking surveys resulted from minor tran- on TS were greatest for large paddle®sh, but con- sect differences. However, mean transect lengths ®rmed that these effects could be reduced at lower were similar for each acoustic survey method, and L/␭ values for both length-groups of paddle®sh, the proportion of lake area sampled with side- for example by use of 38-kHz or 70-kHz frequency looking and down-looking methods was similar transducers. during day and night (Table 1). Backscatter model results were compared with Whole-lake abundance estimates differed be- ®eld measures at 200 kHz separately for age-0 and tween side-looking and down-looking surveys adult paddle®sh (Figure 4). Discontinuity of the both day and night, but ratio and cluster estimates solid line of Figure 4 occurred because two sep- produced similar results (Table 1). Day side- arate model results were necessary due to the large looking estimates (N Ϯ 95% con®dence interval) difference in the two size-groups of ®sh used to were 187 Ϯ 182 ®sh (ratio estimate) and 252 Ϯ model TS. Two of four measured results from age- 231 ®sh (cluster estimate), whereas day down- 0 ®sh were lower than model results, whereas 18 looking estimates were 12,831 Ϯ 8,208 ®sh (ratio of 20 measured results from adult ®sh were greater estimate) and 13,489 Ϯ 10,531 ®sh (cluster esti- than model results. However, means of measured mate). Night side-looking estimates were 299 Ϯ TS were generally within one SD of model results 176 ®sh (ratio estimate) and 313 Ϯ 121 ®sh (clus- for both size-groups, and the match between mea- ter estimate), whereas night down-looking esti- sured and modeled TS was good. mates were 3,464 Ϯ 3,373 ®sh (ratio estimate) and 3,556 Ϯ 2,992 ®sh (cluster estimate). Night side- Abundance Estimates looking estimates were the most precise, and the Acoustic counts in Hebron Pond.ÐThe acoustic coef®cient of variation of the mean (100´SE/mean) counts of paddle®sh stocked in Hebron Pond de- ranged from 20% to 50%. 752 HALE ET AL.

FIGURE 3.ÐPredicted reduced scattering length of (A) a 346-mm age-0 paddle®sh and (B) a 1,096-mm adult paddle®sh plotted as a function of ®sh aspect (␪) and the ratio of ®sh length (L) to acoustic wavelength (␭) and modeled over an aspect range of 70±110Њ and an acoustic frequency range of 12±420 kHz. PADDLEFISH ACOUSTICS 753

age-1±3 groups in Coeur d'Alene Lake, Idaho, al- though differentiating between ages 1±3 required trawl samples. By use of a 70-kHz, split-beam so- nar, Warner et al. (2002) identi®ed two or three size- classes of alewives Alosa pseudoharengus in sur- veys conducted in eight inland New York lakes from July through November. However, Buerkle (1987) cautioned against use of TS to differentiate length- groups by providing an example of TS reductions associated with increases in ®sh tilt rather than ®sh length in Atlantic cod Gadus morhua. Acoustic backscatter models con®rmed that paddle®sh TS is a complex function of both bio- logical and equipment factors and that adult pad- FIGURE 4.ÐTarget strength (solid line) and SDs (dot- ted lines) of age-0 and adult paddle®sh predicted from dle®sh TS is particularly sensitive to ®sh tilt, or anatomically based models, compared with the measured angle to transducer, and sonar frequency. Similar mean target strength (black dots; ϩ1 SD) at 200 kHz. sensitivity of TS to tilt and sonar frequency might be expected for other large freshwater ®shes be- cause models of thread®n shad Dorosoma pete- An estimated 130 Ϯ 55 paddle®sh were in Horse- nense (Jech et al. 1995) and Atlantic cod (Clay and shoe Lake based on depletion gillnetting results Horne 1994) identify similar interrelations of these used to develop the equation y ϭ 1.075±0.008x factors. Importance of the paddle®sh swim bladder (P ϭ 0.01, F ϭ 12.97, r2 ϭ 0.68), where y is gill- to TS is also consistent with ®ndings that the swim net catch per effort and x is cumulative catch. A bladder contributes 90±95% of maximum and av- total of 110 adult paddle®sh and 29 nontarget ®sh, erage dorsal-aspect acoustic backscatter for three including 11 common carp, 10 bigmouth buffalo, gadoid species (Foote 1980). Paddle®sh models 6 smallmouth buffalo, and 2 black buffalo, were identi®ed the importance of large swim bladders captured in 96 gill-net sets. Lengths of gillnetted to amplitude of acoustic backscatter, and also dem- paddle®sh overlapped with those used for mea- onstrated that when paddle®sh are tilted either surement of TS and development of backscatter head up or head down relative to the transducer, models, but few of the nontarget species, repre- the returned sound will greatly diminish. senting 21% of the total gill-net catch, overlapped The potential for diminished acoustic backscat- with lengths of paddle®sh sampled (Figure 5). ter for a given echo is more pronounced as L/␭ increases, but these concerns can be partially ad- Discussion dressed with selection of an appropriate sonar sys- Counting and estimating abundance of large ®sh tem. Clay and Horne (1994) suggest that sonar with acoustics is contingent upon meaningful TS frequencies providing L/␭ between 2 and 10 may information. Paddle®sh are physostomes, with a be most useful for ®sh identi®cation, yet in our large swim bladder and an irregular body shape study, L/␭ for adult paddle®sh was approximately compared with teleosts, and echo amplitudes from 154. Lower frequencies commonly used in acous- paddle®sh are large. We could differentiate age-0 tic surveys would have reduced L/␭ to 92 (120 and adult paddle®sh because differences in mean kHz), 54 (70 kHz), or 29 (38 kHz). Sonar fre- TS were large. Use of acoustic TS to differentiate quencies of 70 kHz or lower would improve ®sh large adult freshwater ®shes from age-0 ®sh of the identi®cation because the in¯uences of ®sh length same species or from smaller coexisting species and aspect on acoustic backscatter are reduced. should be possible because TS is generally pro- However, the transducer near ®eld also must be portional to ®sh size (Love 1969, 1971, 1977). Age- considered in selection of any sampling frequency, 0 and adult ®sh have been differentiated via TS for particularly in shallow habitats such as those we a number of smaller species. Burczynski et al. sampled for paddle®sh. Near ®eld is the area im- (1987) differentiated age-0 and older age-classes of mediately in front of the transducer face where rainbow smelt Osmerus mordax in Lake Oahe, transducer operation may not be linear, and is a South Dakota, with a dual-beam, 420-kHz sonar. function of transducer frequency and acoustic Parkinson et al. (1994) used a similar sonar to sep- beam angle. For example, a 200-kHz, 6Њ transducer arate kokanee Oncorhynchus nerka into age-0 and has a 0.8-m near ®eld, whereas a 38-kHz, 6Њ trans- 754 HALE ET AL.

TABLE 1.ÐAcoustic survey results and estimates of paddle®sh abundance in Horseshoe Lake, Ohio, October 1999. Means and 95% con®dence intervals are shown for transect length and abundance. The coef®cient of variation of the mean (100´SE/mean) is given in parentheses below each abundance estimate; N ϭ number of survey transects.

Abundance Transect Number length % of lake of ®sh Ratio Cluster Period Survey N (m) sampled counted estimate estimate Day Side looking 7 44 Ϯ 13 2.14 4 187 Ϯ 182 252 Ϯ 231 (50) (47) Down looking 6 48 Ϯ 15 0.07 9 12,831 Ϯ 8,208 13,489 Ϯ 10,531 (33) (40) Night Side looking 11 48 Ϯ 8 3.68 11 299 Ϯ 176 313 Ϯ 121 (30) (20) Down looking 10 44 Ϯ 11 0.12 4 3,464 Ϯ 3,373 3,556 Ϯ 2,992 (50) (43) ducer has a 4.1-m near ®eld; therefore, a 10Њ beam not believe that acoustic beam avoidance was an angle would be required to reduce the near ®eld issue in our mobile paddle®sh surveys or else we of a 38-kHz transducer to 1.5 m and maximize would have observed that response in the ®rst sur- sampling ef®ciency. Smaller beam angles at low vey. Given our results, we suspect that once we frequencies would miss a large portion of habitat created an initial disturbance by boat operation in in down-looking transects with mean depths of 5± Hebron Pond, paddle®sh moved to the bottom or 10 m, whereas large beam angles would increase center of the pond or swam at aspects that returned sampling area. weak echoes. Both types of behavior, avoidance As in other reported results, the measurement or orientation changes in response to an approach- of paddle®sh TS agreed well with TS predicted ing boat, are concerns warranted in acoustic sur- from backscatter models. Foote and Traynor veys (Olsen et al. 1983). Our results do not let us (1988) indicated that TS of walleye pollock Ther- conclude whether behavior is a signi®cant issue in agra chalcogramma measured in situ at 38 kHz paddle®sh surveys conducted in larger habitats compared well to modeled results, but modeled where transects are not replicated during a short averages were consistently lower than measured time frame. averages. Clay and Horne (1994) found a good Differences in acoustic counts of paddle®sh be- match between their models of Atlantic cod and tween side-looking and down-looking surveys measured results from dead, tethered cod reported conducted during the day and at night indicated at 38 kHz. Jech et al. (1995) reported that TS of that diurnal movement should be considered in live, tethered thread®n shad compared well with acoustic survey design. During the Horseshoe one model based on ®sh morphology (i.e., body Lake surveys, paddle®sh were more likely to be and swim bladder) and two additional models near the surface at night than during the day. Shal- based upon cylinder representations of ®sh mor- low night movement and seasonal differences in phology at 120, 200, and 420 kHz. Differences we depth selection of paddle®sh were observed by observed between measured and modeled paddle- Zigler et al. (1999) in an Upper Mississippi River ®sh TS may have resulted from tethering, anes- telemetry study. Complete acoustic coverage with thesia, or ®sh handling (MacLennan and Sim- both methods appears important. We would have monds 1992) or from errors in models. missed most of the ®sh in the top2mofthewater Pond experiments demonstrated the potential in- column without side-looking acoustics; these ®sh ¯uence of ®sh behavior on acoustic surveys. Our represented nearly one-third of large targets count- ®st survey of Hebron Pond allowed us to identify ed during the day and nearly three-fourths of large 24 large targets when 21 large targets were present, targets counted at night. Light intensity may also resulting from sampling some ®sh more than once, be a signi®cant issue during night surveys. Luecke whereas subsequent pond surveys failed to locate and Wurtsbaugh (1993) found that Bonneville cis- even half as many. Some high-frequency sounds coes Prosopium gemmifer dispersed closer to the of 110±140 kHz emitted from stationary trans- bottom during the full moon compared to the new ducers create avoidance responses from alewives moon in Bear Lake, Utah, creating dif®culty in (Dunning et al. 1992) and blueback herring Alosa detecting ®sh during the full moon and reducing aestivalis (Nestler et al. 1992). However, we do acoustic density and biomass estimates by 50%. PADDLEFISH ACOUSTICS 755

FIGURE 5.Ð(A) Length frequency distribution of paddle®sh sampled with gill nets in Horseshoe Lake, adult paddle®sh tethered to estimate target strength, and adult paddle®sh measured for anatomically based models and (B) length frequency distribution of adult paddle®sh converted back to eye±fork length for comparison with total length of bycatch.

Abundance estimates from acoustic counts of seine hauls in Wyoming lakes and reservoirs, adult paddle®sh in Horseshoe Lake were greater where purse seine bycatch was generally less than or more variable than population estimates derived 10%. Cryer (1996) reported a good match of 10 from depletion gillnetting. These differences could acoustic surveys of rainbow trout Oncorhynchus have been related to bycatch, sampling effort, sam- mykiss from acoustic counts in Lake Taupo, New pling methods, or approaches to analysis. Species Zealand, with expectations from a range of pub- composition is an important consideration in lished empirical models and a recent creel survey. acoustic surveys. Yule (2000) found similar den- Rainbow trout were the only large pelagic species sities of salmonids in acoustic surveys and purse- and could easily be separated by TS from the other 756 HALE ET AL. pelagic species, the New Zealand common smelt od that incorporates the spatial distribution of ®sh Retropinna retropinna, in steep-sided Lake Taupo. in the analysis and that is not dependant on the Horseshoe Lake gillnetting conservatively indi- presence of large numbers of ®sh or on their ho- cated that at least one-®fth of acoustic abundance mogenous distribution (Thompson 1991; Thomp- estimates could have been comprised of common son and Seber 1996). Adaptive cluster sampling carp, bigmouth buffalo, smallmouth buffalo, or involves oversampling an area when target organ- black buffalo. Although body sizes of adult pad- isms are identi®ed in random transects, and then dle®sh and bycatch species did not overlap con- estimating abundance via an analysis that incor- siderably, we do not have information about TS porates cluster size and geographic distributions of of these nontarget species, and vulnerability of by- organism clusters. This method is particularly suit- catch to our gill nets may have differed from that ed for estimating abundance of organisms found in of paddle®sh. If our acoustic abundance results low abundance but in clustered distributions, and were corrected for bycatch, side-looking surveys may be useful for surveys of paddle®sh or other would have provided a range of 150±250 adult large ®shes. We could not apply this approach post paddle®sh and down-looking surveys would have hoc because we did not sample adaptively and did provided a range of 2,771±10,791 adult paddle®sh. not have enough information to make reasonable The large discrepancy between down-looking assumptions about the shape of clusters. abundance estimates and those from side-looking Acoustic surveys generally require less ®eld ef- acoustic surveys and gill-net surveys also sug- fort and little or no ®sh handling compared with gested that sampling effort was important. Differ- alternative methods, the exact amounts depending ences in sampling effort were considerable be- upon the level of supplemental sampling required. tween down-looking and side-looking surveys. Al- In Horseshoe Lake, four acoustic surveys required though duration of the surveys was similar, the approximately 6 man-hours (calibration, setup, down-looking acoustic beam was considerably re- and surveys) compared to 180 man-hours for one duced relative to the side-looking beam due to depletion gillnetting survey. Within the Mississip- shallow lake depth. The low percentage of area pi River basin, paddle®sh catch rates have been sampled in the down-looking survey would be ex- reported at 0.5 ®sh/h for gillnetting, 6.5 ®sh/h for pected to increase variance (Thorne 1983), but it electro®shing, and from 0.2 to 1.3 ®sh/h for snag- may also have in¯ated abundance estimates or the ging with rod and reel (Grady and Conover 1998). in¯uence of bycatch. None of these methods provided estimates of abun- The smaller number of paddle®sh counted and dance, but rather general indicators or indexes. In their distribution may also have in¯ated ratio and contrast, we estimated that our mobile acoustic cluster estimates. Conners and Schwager (2002) surveys conducted at 2.5 m/s counted paddle®sh, have shown that a simple sample mean may have or large ®sh targets, at a rate of 16.8 ®sh/h and a strong nonnormal distribution when transects are did not require handling of ®sh. Paddle®sh injury sampled from a spatially clustered population. If has been associated with electro®shing (Scarnec- the distribution of paddle®sh or other large ®shes chia et al. 1999), and gillnetting and snagging pose is clustered, then results such as these may violate other potential sources of injury or mortality, but the standard design assumption that target organ- such concerns do not exist during acoustic surveys. isms are evenly distributed in space such that each The short time required to conduct acoustic sur- transect may be considered a random draw from veys in Horseshoe Lake demonstrated the potential an identical underlying distribution. Although it is cost savings and greater spatial coverage for sam- possible to deal with clustering through model- pling large ®sh compared to effort expended in based estimation such as kriging (MacLennan and methods such as gillnetting or electro®shing. Simmonds 1992; Petitgas 1993), the very small Acoustic surveys also provide an alternative to numbers of ®sh detected in our surveys con- methods of estimating paddle®sh abundance, strained the estimation of parameters for these which historically relied on reports of sport or models. Given this problem, down-looking results, commercial ®shery catches (Gengerke 1986). in particular, may have been improved by increas- Our results are the ®rst to indicate the potential ing both sampling area and number of transects, of ®sheries acoustics to estimate paddle®sh abun- and by stratifying sampling design by habitat dance, and we recommend use of acoustics for (Cochran 1977; Jolly and Hampton 1989). sampling paddle®sh and other large freshwater One alternative approach to ratio or cluster sam- ®shes. These ®ndings are signi®cant given the pau- pling estimates is adaptive cluster sampling, a meth- city of noninvasive techniques available to sample PADDLEFISH ACOUSTICS 757 paddle®sh and other large ®shes and the increasing L. Wells, and P. A. Unger. 1991. Acoustic measures need to quantify their abundance. Speci®c rec- of the abundance and size of pelagic planktivores ommendations include: (1) use of sonar frequen- in Lake Michigan. Canadian Journal of Fisheries and Aquatic Sciences 48:894±908. cies of 70 kHz or lower, with the largest acoustic Buerkle, U. 1987. Estimation of ®sh length from acous- beam angle possible given near-®eld consider- tic target strength. Canadian Journal of Fisheries ations for habitat sampled; (2) concurrent sam- and Aquatic Sciences 44:1782±1785. pling of ®sh populations with an alternative meth- Burczynski, J. J., P. H. Michaletz, and G. M. Marrone. od to identify bycatch; and (3) development of 1987. Hydroacoustic assessment of the abundance survey designs that rigorously consider sampling and distribution of rainbow smelt in Lake Oahe. North American Journal of Fisheries Management effort, strati®cation of habitat, and geographic dis- 7:106±116. tribution of target species. Clay, C. S., and J. K. Horne. 1994. Acoustic models of ®sh, the Atlantic cod (Gadus morhua). Journal of Acknowledgments the Acoustical Society of America 96:1661±1668. We thank J. Stafford and staff at the Hebron Cochran, W. G. 1977. Sampling techniques. Wiley, New York. State Fish Hatchery, Ohio Department of Natural Conners, M. E., and S. J. Schwager. 2002. Use of adap- Resources (ODNR) for assistance with care, hold- tive cluster sampling for hydroacoustic surveys. ing, and transportation of paddle®sh; S. Mimms, ICES Journal of Marine Science 59:1314±1325. Kentucky State University, for providing paddle- Cryer, M. 1996. A hydroacoustic assessment of a rain- ®sh; and D. Savage for project support through bow trout (Oncorhynchus mykiss) population in a permitting use of his property to conduct the deep oligotrophic lake. Pages 196±205 in I. G. Horseshoe Lake sampling. G. Disterdick and E. Cowx, editor. Stock assessment in inland ®sheries. Blackwell Scienti®c Publications, Cambridge, Mas- Wisner, The Ohio State University Equine and Bo- sachusetts. vine Veterinary Clinic, provided facilities, advice, Degan, D. J., and W. Wilson. 1995. Comparison of four and materials for radiographs. A. Mueller, Aqua- hydroacoustic frequencies for sampling pelagic ®sh coustics, Inc., provided assistance with analysis populations in Lake Texoma. North American Jour- and manuscript review. T. Stefanavage and L. Leh- nal of Fisheries Management 15:924±932. man of the Indiana Department of Natural Re- Dunning, D. J., Q. E. Ross, P. Geoghegan, J. J. Reichle, J. K. Menezes, and J. K. Watson. 1992. Alewives sources and C. Goings, R. Hoover, J. Mion, J. Na- avoid high-frequency sound. North American Jour- varro, and S. Xenakis of the ODNR provided as- nal of Fisheries Management 12:407±416. sistance with ®eld sampling. R. Hale and D. Degan Fleischer, G. W., R. L. Argyle, and G. L. Curtis. 1997. acknowledge funding provided by the Ohio Wild- In situ relations of target strength to ®sh size for life Diversity Fund and the Ohio River Trust Fund, Great Lakes pelagic planktivores. Transactions of and J. Horne acknowledges funding provided by the American Fisheries Society 126:786±794. the U.S. Of®ce of Naval Research. Foote, K. G. 1980. Importance of the swim bladder in acoustic scattering by ®sh: a comparison of gadoid and mackerel target strengths. Journal of the Acous- References tical Society of America 67:2084±2089. Anderson, W. G., R. S. McKinley, and M. Colavecchia. Foote, K. G. 1987. Fish target strength for use in echo 1997. The use of clove oil as an anesthetic for rain- integrator surveys. Journal of the Acoustical Soci- bow trout and its effects on swimming performance. ety of America 82:981±987. North American Journal of Fisheries Management Foote, K. G., and D. N. MacLennan. 1984. Comparison 17:301±307. of copper and tungsten carbide spheres. Journal of Beauchamp, D. A., C. Luecke, W. A. Wurtsbaugh, H. the Acoustical Society of America 75:612±616. G. Gross, P. E. Budy, S. Spaulding, R. Dillenger, Foote, K. G., and J. J. Traynor. 1988. Comparison of and C. P. Gubala. 1997. Hydroacoustic assessment walleye pollock target strength estimates deter- of abundance and diel distribution of sockeye salm- mined from in situ measurements and calculations on and kokanee in the Sawtooth Valley lakes, Idaho. based on swim bladder form. Journal of the Acous- North American Journal of Fisheries Management tical Society of America 83:9±17. 17:253±267. Gengerke, T. W. 1986. Distribution and abundance of BioSonics. 1999. Guide to using DT series hardware and paddle®sh in the United States. Pages 22±35 in J. software, manual version 4.0. BioSonics, Seattle. G. Dillard, L. K. Graham, and T. R. Russell editors. Brandt, S. B. 1996. Acoustic assessment of ®sh abun- The paddle®sh: status, management and propaga- dance and distribution. Page 385±432 in B. R. Mur- tion. American Fisheries Society, North Central Di- phy and D. W. Willis, editors. Fisheries techniques, vision, Special Publication 7, Bethesda, Maryland. 2nd edition. American Fisheries Society, Bethesda, Grady, J. M., and G. A. Conover. 1998. Mississippi Maryland. River basin paddle®sh research, coded-wire tagging Brandt, S. B., D. M. Mason, E. V. Patrick, R. L. Argyle, project. Mississippi Interstate Cooperative Re- 758 HALE ET AL.

source Association, Annual Report 1997, Columbia, in Lofoten obtained from underwater photographs Missouri. and their relation to target strength. ICES CM/B: Henley, D., L. Franklin, S. Hale, C. O'Bara, and T. Ste- 25. fanavage. 2001. Paddle®sh assessment in the Ohio Olsen, K., J. Angell, F. Pettersen, and A. Lovik. 1983. River sub-basin: current status and strategic plan Observed ®sh reactions to a surveying vessel with for management. Ohio River Fisheries Manage- special references to herring, cod, capelin and polar ment, Team Report ORFMT0101, Columbus. cod. FAO (Food and Agriculture Organization of Horne, J. K., and C. S. Clay. 1998. Sonar systems and the United Nations) Fisheries Reports 300:131±138. aquatic organisms: matching equipment and model Parkinson, E. A., B. E. Rieman, and L. G. Rudstam. parameters. Canadian Journal of Fisheries and 1994. Comparison of acoustic and trawl methods Aquatic Sciences 55:1296±1306. for estimating density and age composition of ko- Horne, J. K., and J. M. Jech. 1999. Multi-frequency kanee. Transactions of the American Fisheries So- estimates of ®sh abundance: constraints of rather ciety 123:841±854. high frequencies. ICES Journal of Marine Science Petitgas, P. 1993. Geostatistics for ®sh stock assess- 56:184±199. ments: a review and an acoustic application. ICES Jech, J. M., D. M. Schael, D. M. Schael, and C. S. Clay, Journal of Marine Science 50:285±298. 1995. Application of three sound scattering models Ricker, W. E. 1975. Computation and interpretation of to thread®n shad (Dorosoma petenense). Journal of biological statistics of ®sh populations. Bulletin of the Acoustical Society of America 98:2262±2269. the Fisheries Research Board Canada 191. Jennings, C. A., and S. J. Zigler. 2000. Ecology and Runstrom, A. L., B. Vondracek, and C. A. Jennings. biology of paddle®sh in North America: historical 2001. Population statistics for paddle®sh in the perspectives, management approaches, and research Wisconsin River. Transactions of the American priorities. Reviews in Fish Biology and Fisheries Fisheries Society 130:546±556. 10:167±181. Scarnecchia, D. L., P. A. Stewart, L. F. Tyckman, G. Van Jolly, G. M., and I. Hampton. 1989. A strati®ed random Eeckhout, B. C. Reed, and L. K. Graham. 1999. transect design for acoustic surveys of ®sh stocks. Caution on the use of electro®shing for paddle®sh. Canadian Journal of Fisheries and Aquatic Sciences Intermountain Journal of Science 5:35±38. 47:1282±1291. Schael, D. M., J. A. Rice, and D. J. Degan. 1995. Spatial Love, R. H. 1969. Maximum side aspect target strength and temporal distribution of thread®n shad in a of an individual ®sh. Journal of the Acoustical So- southeastern reservoir. Transactions of the Ameri- ciety of America 46:746±752. can Fisheries Society 124:804±812. Love, R. H. 1971. Dorsal-aspect target strength of an Snedecor, G. W., and W. G. Cochran. 1980. Statistical individual ®sh. Journal of the Acoustical Society of methods. Iowa State University Press, Ames. America 49:816±823. Thompson, S. K. 1991. Adaptive cluster sampling: de- Love, R. H. 1977. Target strength of an individual ®sh signs with primary and secondary unit. Biometrics at any aspect. Journal of the Acoustical Society of 47:1103±1115. America 62:1397±1403. Thompson, S. K., and G. A. F. Seber. 1996. Adaptive Luecke, C., and W. A. Wurtsbaugh. 1993. Effects of sampling. Wiley, New York. moonlight and daylight on hydroacoustic estimates Thorne, R. E. 1983. Application of hydroacoustic as- of pelagic ®sh abundance. Transactions of the sessment techniques to three lakes with contrasting American Fisheries Society 122:112±120. ®sh distributions. FAO (Food and Agriculture Or- MacLennan, D. N., and E. J. Simmonds. 1992. Fisheries ganization of the United Nations) Fisheries Reports acoustics. Chapman and Hall, New York. 300:269±277. Medwin, H., and S. S. Clay. 1997. Fundamentals of Warner, D. M., L. G. Rudstam, and R. A. Klumb. 2002. acoustical oceanography. Academic Press, New In situ target strength of alewives in freshwater. York. 2002. Transactions of the American Fisheries So- Mulligan, T. J., and R. Kieser. 1986. Comparison of ciety 131:212±223. acoustic population estimates of salmon in a lake Yule, D. L. 2000. Comparison of horizontal acoustic with a weir count. Canadian Journal of Fisheries and purse-seine estimates of salmonid densities and and Aquatic Sciences 43:1373±1385. sizes in eleven Wyoming waters. North American Nestler, J. M., G. R. Ploskey, and J. Pickens. 1992. Re- Journal of Fisheries Management 20:759±775. sponse of blueback herring to high-frequency sound Zigler, S. J., M. R. Dewey, and B. C. Knights. 1999. and implications for reducing entrainment at hy- Diel movement and habitat use by paddle®sh in dropower dams. North American Journal of Fish- Navigational Pool 8 of the Upper Mississippi River. eries Management 12:667±683. North American Journal of Fisheries Management Olsen, K. 1977. Orientation measurements of the cod 19:180±187.