In situ acoustic estimates of the swimbladder volume of Atlantic herring ( Clupea harengus )

Abstract

Acoustic measurements at 1.5–5 kHz on fish in the Gulf of Maine showed a swimbladder-resonance peak near 2.5 kHz at 160–190-m depth. Midwater trawls confirmed that the fish were likely to be Atlantic herring ( Clupea harengus ) of 19–29 cm length. Calculation using a model of swimbladder resonance gives swimbladder volumes of 1.2% of fish weight at 160–190 m. Extrapolation of this volume of gas using Boyle's Law suggests that at the sea surface, these herring would need to inflate their swimbladders by up to five to six times the volume required for neutral buoyancy. If these fish were to maintain this volume of gas with surface “gulping”, they would need to submerge from the sea surface with a 30% excess buoyancy. In general, swimbladders of the Clupeidae may have greater volumes of gas than if the fish were neutrally buoyant at the sea surface and the interpretation of HF-echosounder surveys may be additionally complex when the volume of gas and swimbladder volume are difficult to predict. Mechanisms of how herring obtain additional swimbladder gas are discussed.

1 Introduction

Marine fish with swimbladders generally maintain swimbladder-gas volumes of 5% of their body weight in order to maintain neutral buoyancy ( Jones and Marshall, 1953 ). In the herring family, Clupeidae, the swimbladder is hypothesized to provide a dual function, acting to provide buoyancy and as a reservoir of gas for the acoustico-lateralis system ( Blaxter et al ., 1979 ). Herring are physostomes, and as such lack an internal organ dedicated to maintain gas in their swimbladder. Instead, they possess two ducts that allow its inflation and deflation. The anterior duct connects the middle of the swimbladder to the pyloric caecae and is the primary duct for its inflation. Gas in a herring's swimbladder is hypothesized to have three possible origins. The first is that atmospheric air is obtained by “gulping” air at the sea surface ( Brawn, 1962 ; Blaxter et al ., 1979 ; Blaxter and Hunter, 1982 ; Blaxter and Batty, 1984 ). The second is that the gas is a product of bacterial action in the digestive tract ( Brawn, 1962 ) while the third is that the gas is excreted from the walls of the swimbladder or ducts ( Sundnes et al ., 1958 ; Fahlen, 1967 ).

Regardless of the mechanism, knowledge of the volume of gas in the swimbladder is important to fisheries acoustics because of the large proportion (90–95%) that the swimbladder contributes to acoustic backscatter at high frequencies ( McCartney and Stubbs, 1971 ; Foote, 1980 ). Since backscattering at high frequencies is sensitive to swimbladder cross-sectional area and shape ( Blaxter and Batty, 1990 ), and backscatter at low frequencies is sensitive to swimbladder volume ( Love, 1978 ), knowledge of swimbladder volume is important in understanding both low-frequency (LF) and high-frequency (HF) backscatter. A discrepancy of just a few decibels between predicted and observed target strengths can have a major impact on the conversion of acoustic energy to fish biomass in fisheries-acoustic surveys ( MacLennan and Simmonds, 1992 ).

A mid-frequency (MF), 1.5–5-kHz, broadband, towed sonar has been developed by the Naval Research Laboratory (NRL) for investigating the acoustic backscatter from fish in a continuous-survey mode ( Nero et al ., 2001 ). Acoustic data in this frequency range provide measurements of the swimbladder-resonance frequency, from which swimbladder volumes can be deduced. In September 2001, pre-spawning Atlantic herring ( Clupea harengus ) were surveyed in the Gulf of Maine and on the northern flank of Georges Bank with the NRL mid-frequency system and the Northeast Fisheries Science Center's (NEFSC) scientific echosounders (12, 38, and 120 kHz). The MF sonar was operated simultaneously with the NEFSC high-frequency, fisheries echosounders and sonar runs were followed closely with pelagic trawls for species identification and biological measurements. This study documents the swimbladder resonance of herring during several measurements with both systems.

2 Sampling methods

The experiment was conducted in two general areas, one near Cashes Ledge in the Gulf of Maine, 42°50′N 68°50′W, and the other on the north side of Georges Bank, 42°N 68°W ( Figure 1 ). Measurements were obtained from the NOAA Ship FRV “Delaware II” during the first two weeks of September 2001 in conjunction with the annual NEFSC acoustic survey of Atlantic herring. On several deployments, problems with electrical cables in the MF sonar resulted in bad transmitted or received pings. In spite of the cable problems, three sets of high-quality MF, HF, and biological measurements on Atlantic herring were obtained ( Table 1 ).

Figure 1

A map of the sampling sites.

Table 1

Deployment chronology giving sample dates and times with their associated MF–HF runs. Times are US Eastern daylight savings time. Run 1 was conducted on Cashes Ledge and Runs 2 and 3 were conducted on Georges Bank.

Deployment	Run	Type	Begin date	End date	Latitude	Longitude
48	1	HSMRT	09/10-09:10	09/10-09:52	42 50.49N	68 49.45W
49: Run 1	1	MF–HF	09/10-11:35	09/10-12:19	42 49.53N	48 49.09W
50	1	CTD	09/10-14:35	09/10-14:49	42 47.00N	68 46.98W
59	2	HSMRT	09/11-12:23	09/11-13:09	42 00.90N	67 44.80W
61	2	HSMRT	09/11-17:17	09/11-17:42	42 05.24N	67 53 16W
62	2	HSMRT	09/11-18:59	09/11-19:41	42 04.48N	67 52 19W
64: Run 2	2	MF–HF	09/11-20:33	09/12-00:07	42 03.37N	67 52.20W
66	3	HSMRT	09/12-08:56	09/12-09:28	41 52.42N	68 11.62W
67	3	CTD	09/12-12:57	09/12-13:08	42 09.84N	68 05.96W
69	3	HSMRT	09/12-13:43	09/12-14:37	42 08.09N	68 06.44W
70: Run 3	3	MF–HF	09/12-15:15	09/12-17:19	42 02.92N	68 07.65W

2.1 Trawling operations

A High-Speed, Midwater, Rope Trawl (HSMRT, modified from Dotson and Griffith, 1996 ), with operational dimensions of the mouth opening when towed of 10-m vertical by 30-m horizontal (∼300 m ² ), was used to sample fish from locations and depths of high acoustic backscattering. The net was towed at approximately four knots. Although trawl locations were chosen to correspond to acoustic data-collection sites ( Table 1 ), not all trawls were an ideal match; weather, boat traffic, and gear failures resulted in several missed opportunities. At the Cashes Ledge site, trawl 48 was about 1 nmi north of the MF–HF deployment (Run 1) and occurred about 2 h prior to it. At the Georges Bank site, five trawls (59, 61, 62, 66, and 69) accompanied the two acoustic measurements within a period of a half to eight hours and occurred at the site of the measurements or no further than 15 nmi away. CTD casts were made at the Cashes Ledge and Georges Bank sites.

Trawl performance was monitored with acoustic (wing and door spread, and mouth opening) and temperature-depth sensors. Trawl duration was adjusted depending on the catch. Catch was monitored with a SIMRAD SF900 sonar mounted on the headrope. Trawl catches were separated by species, and the weights of each species of catch were recorded. Individual lengths (fork length) of species other than Atlantic herring were measured to the nearest centimeter. Atlantic herring were measured for individual lengths (fork length to the nearest millimeter), individual weights recorded to the nearest gram, sex, maturity, and stomach contents noted, and otoliths were extracted for age estimation.

2.2 Mid-frequency (MF) scattering measurements

The MF acoustic source and receiver transducers (Naval Undersea Warfare Center, Underwater Sound Reference Division (USRD)) were mounted on a 3-m long pontoon raft that was towed 100 m behind the research vessel. The source transducer was mounted on an aluminum pole that extended 1 m below the sea surface. The receiver was a flexible towed array, tethered 3 m behind the source. The source transducer was an USRD Model G81, cylindrical ring transducer. The G81 beam pattern is radially symmetric with a cross-sectional beam width at 3 kHz of about 90° between 3-dB down points. The G81 was driven by an Instruments Inc. Model L-6 amplifier, with input signals generated by a computer with a QuaTech Inc., WSB-10, 12-bit digital to analog converter. An eight-ping frequency hop (FH) was used with each ping a continuous wave (CW) sine pulse (10-ms pulse duration) generated for a sequence of 1.5–5 kHz at 0.5 kHz increments. Each repetition of the FH lasted 16 s. Source levels were 188.9 dB at 1.5 kHz rising to 195.8 dB at 3.5 kHz and falling to 187.2 dB at 5 kHz.

The receiver was an USRD Model F81 line hydrophone, which was constructed of 40 omnidirectional elements spaced at 15.2-cm intervals and set in a castor oil-filled vinyl hose. Beam widths of the F81 were progressively narrower from 8.4° at 1.5 kHz to 2.6° at 5 kHz. An Ithaco Model 568 pre-amplifier and a Wavetek Model 716 filter conditioned the analog signal before digital sampling with a National Instruments Inc., PCI MIO 16XE, 16-bit analog to digital converter (20-kHz digital sampling rate). Sampling of the received signal began at the start of each outgoing pulse and continued well beyond the echo returned by the sea floor. To achieve a high signal-to-noise (SNR) ratio for fish in either surface, midwater, or bottom schools, the G81 was mounted with its axis of symmetry in the fore-aft direction, thus providing high ensonification at 1.5–5 kHz of targets occurring deep in the water column. The digital data were bandpass-filtered at 500 Hz around the pulse-center frequency. The Hilbert transform of the signal was then used to calculate the amplitude envelope of the time-series. Amplitude envelopes were smoothed using a rectangular window equal in width to the 10-ms pulse. The echo strength (ES) was calculated as  

(1)

where V is the amplitude envelope of the recorded time-series, gain the amplification in dB of the recording system, FFVS the free field voltage sensitivity of the F81, SL the source level of the G81, TL the transmission loss from source to target assuming spherical spreading, c the sound-speed, τ the pulse length and β is the beam width of the F81. The last term in this equation represents the total volume sampled by the system at range r.

Echo strengths were displayed in echograms as a function of range and FH number for each run and examined to determine the presence of individual fish targets, layers, or aggregations. Echo-strength regions were selected corresponding to scattering from the horizontal layers of scatterers. For each region, volume-scattering strength (S _V ) was calculated using  

(2)

where z is the depth at the top of the scattering layer. In one instance (FH 36–38 of Run 1), for what appeared to be a discrete echo from a fish school, the sampled volume term was added back to the echo strength to obtain the Target Strength (TS) of the school.

The MF system transducers were calibrated at a USRD calibration facility. Source levels were based on these calibrations and the measured current and voltage output of the amplifier. Previous measurements on known targets suggest that the system produces accuracy within ±1 dB ( Dubberley et al ., 2001 ; Nero et al ., 2001 ).

2.3 High-frequency (HF) scattering measurements

High-frequency scattering measurements were obtained with a SIMRAD EK500 scientific echosounder transmitting on hull-mounted transducers at 12 (single-beam, 16° beam width), 38 (split-beam, 12° beam width), and 120 (split-beam, 7° beam width) kHz. Pings for each frequency were transmitted simultaneously every 2 s. Pulse durations were 3 ms for the 12-kHz and 1 ms for the 38- and 120-kHz systems. Acoustic signals were digitized and processed by the EK500 (20 log R TVG and calibration gains applied) and echo-integration data (1-m vertical) were logged on a computer for post-processing. Post-processing was done using SonarData Echoview software (SonarData, Pty Ltd. GPO Box 1387 Hobart, Tasmania, Australia), and consisted of removing extraneous bottom echoes and surface noise (the upper 10 m were eliminated for the 38- and 120-kHz data, and the top 32 m for the 12-kHz data) and identifying herring aggregations.

2.4 Spectral images

Spectral images of the mid-frequency data were created by combining the eight data values within one FH and range bin (eight frequencies from 1.5 to 5 kHz and 187 range bins from 0 to 224 m). The frequency with the highest volume-scattering level was used to assign a color of red through blue (red, orange, yellow, light green, green, light blue, cyan, and blue). Color intensity was set using volume-strength intervals. In this case, three color intensities corresponded to the volume-strength intervals: <−50 dB, −50 to −45 dB, and >−45 dB.

The high-frequency (12, 38, and 120 kHz), volume-scattering data were combined into a synthetic color image where the intensity level of each of the frequencies determined the brightness level of the colors magenta (12 kHz), yellow (38 kHz), and cyan (120 kHz). Color intensity was indexed to eight levels of volume scatter. The lowest intensity was set for volume-scatter values <−95 dB, the next six color intensities were at 8.33 dB increments from −95 dB to −45 dB, and the highest color intensity was used for >−45 dB. Five hundred and twelve colors were possible (8 ³ ), with low levels of all three frequencies corresponding to white and high levels of all three frequencies corresponding to black.

2.5 LF swimbladder modelling

The acoustic spectra of fish measured with the MF sonar were compared to scattering spectra generated from a model of the swimbladder resonance of fish ( Love, 1978 ). Several model parameters were based on measurements made on fish obtained from the trawl hauls. The swimbladder model gave the backscattering cross-section of an individual fish, σ _bs assuming the swimbladder is an air-filled, viscous, and spherical shell ( Love, 1978 ). The acoustic cross-section in the backscattered direction, σ _bs , for a single swimbladder of radius r is given by  

(3)

where σ _bs is in m ² , r in m, f the ensonifying frequency in Hz, f ₀ the swimbladder's monopole resonance frequency in Hz, and H is a damping factor. The resonance frequency is  

(4)

where γ _α is the ratio of specific heats of air (γ _α =1.4), P the ambient pressure in Pa (Pascal), and ρ is the density of fish flesh in kg m ⁻³ . The term ζ is a swimbladder-resonance correction attributed to Weston (1967 , described below).

The damping factor is  

(5)

where c is the speed of sound in water in m s ⁻¹ and ξ is the viscosity of fish flesh in Pa s (Pascal second). Love's model includes a term in Equation (4) that accounts for the effects of swimbladder-wall tension on f ₀ and a thermal-damping term in Equation (5) . However, Love (1978) shows that these terms are negligible, so they have been omitted here for clarity. The physical properties of fish were ρ = 1071 kgm ⁻³ based on measurements on herring ( Brawn, 1969 ), and χ = 50Pa s, a value empirically determined as suitable for several families of physoclists: Scorpaenidae, Gadidae, and Macrouridae ( Love, 1993 ; Nero et al ., 1997 , 1998 ), and a physostome: Salmonidae ( Nero and Huster, 1996 ).

The swimbladder-resonance correction, ζ, is justified because most swimbladders are not spherical but instead resemble prolate spheroids. Such elongated spheroidal-shaped swimbladders have been shown to have a significantly higher resonance frequency than a spherical swimbladder ( Weston, 1967 ; Feuillade and Werby, 1994 ; Ye, 1997 ). Weston (1967) provides a correction factor  

(6)

which is a function of ɛ, the ratio of the minor to major axes of a spheroid. To model Atlantic herring for this study, we assumed the swimbladder is a prolate spheroid and calculated the value of ɛ, based on the assumption that below the sea surface, swimbladders of herring compress according to Boyle's Law, viz.  

(7)

The herring swimbladder is also assumed to have “fixed-end positions and a pressure-sensitive diameter” ( Ona, 2003 ), represented as a spheroid collapsing about its minor axis, a _z , given at depth by  

(8)

The length of the major axis, b, is fixed at 33% of total fish length based on measurements from dissection. In the present modelling exercise, the correction ζ increases f ₀ given in Equation (4) by a factor ranging from 1.03 near the sea surface to near 1.28 at 200-m depth. Love's model with Weston's correction to the resonance frequency has recently been used successfully to examine scattering from fish in several ocean regions ( Love, 1993 ; Thompson and Love, 1996 ; Nero and Huster, 1996 ; Nero et al ., 1997 , 1998 ). Feuillade and Werby (1994) have shown that the broadside target strength of spheroids is about 0.5 dB less than that of a sphere. Because this difference is well within the expected error in our measurements and modelling, it was not incorporated here.

Equations (3)–(8) were applied to the complete size distribution of herring, and calculated for the same depth range as the acoustic measurements were integrated. The conversion of herring length to weight is given by: W=0.0033L ^3.35 (W in g and L in cm), based on the length and weight of 427 herring caught in trawls from near the acoustic sites ( Figure 1 , Table 1 ).

2.6 HF modelling

Measurements of scattering with the HF echosounder (12, 38, and 120 kHz) were compared to a simple empirical HF model of a fish's dorsal-aspect backscatter ( Love, 1971 ). Love (his Figure 6 ), fits a simple two-part regression to TS measurements on several species of freshwater and marine fish. In his model, backscatter decreases as L/λ (length/wavelength) increases to approximately 14 (TS = 15.8 log L − 4.2 log λ − 22.9), and then increases as L/λ increases further (TS = 27.5 log L − 7.5 log λ − 36.2). More complex HF scattering models require detailed information on swimbladder size, shape, and aspect ( Clay and Horne, 1994 ), which are presently incomplete for Atlantic herring. Love's simple two-part regression model was preferred as a practical alternative.

3 Results

Spectral images of the mid- and high-frequency data reveal backscattering patterns in the volume-scattering data ( Figures 2–4 ), with the data presented on a common horizontal scale. At a vessel speed of approximately four knots, distances along the ship's track are approximately 6, 10, and 6 km for Runs 1, 2, and 3, respectively. Because of the wide beam of the MF sonar, these data are indexed to range while the HF data are indexed to depth. However, the range and depth scales in the two panels are comparable for target features occurring directly below the transducers. The rise in bathymetry in the middle of Run 2 occurred when the ship steamed south onto Georges Bank, turned 180°, and then steamed north off the bank.

Figure 2

Echograms at MF (upper panel) and HF (lower panel) for Run 1. The white boxes delineate areas for which data were integrated to obtain mean scattering levels. In the upper panel, eight frequencies are represented by colors red through blue (1.5–5 kHz) and in the lower panel, three frequencies are represented by magenta–yellow–cyan (12, 38, and 120 kHz). See text for dB levels.

Figure 3

Echograms for Run 2, details as in the caption for Figure 2 .

Figure 4

Echograms for Run 3, details as in the caption for Figure 2 .

All three runs detected a major scattering layer that occurred in the lower 20 m of the water column over bottom depths of 170–220 m. Trawl catches confirmed that Atlantic herring were the predominant fish species in these areas. Spectral images of the MF acoustic data display this layer as an olive color in Run 1 and a brighter yellow in Runs 2 and 3, indicating a peak in intensity near 2.5 kHz. In the HF spectral images, the herring layer appears as a gray to black layer, indicating approximately equal intensity on all three high frequencies. Runs 1 and 3 occurred during the day ( Table 1 ) with the herring layer distributed from about 20 to 50 m off the bottom. In Run 2, which took place during the night, the herring were more tightly concentrated within 10–15 m of the bottom, but with an indication of some fish higher in the water column. In the second half of Run 3, the highest concentration of herring occurred in a layer at 160–200 m with volume-scatter measurements giving densities of up to one fish m ⁻³ .

To investigate in more detail the absolute level of the MF and HF measurements, sections of the data were selected to calculate average volume-scattering strengths. These sections are outlined in Figures 2–4 . Average volume strengths from within these selections are given in Figures 5–7 . In addition, five equally spaced sub-samples from each selection were obtained for an indication of volume-backscattering variability. In all three runs, backscatter between 160 and 190 m had a distinct resonance peak at 2.5 kHz. The strongest resonance was collected on the large concentration of herring observed towards the end of Run 3 ( Figure 4 ). Levels from this intense layer were as high as −29 dB at a resonance of 2.5 kHz ( Figure 7 ).

Figure 5

Run 1 scattering levels compared to models. The circles are the integrated levels with thick lines depicting mean levels. Models: LF swimbladder model, thin line; HF model, thin, dashed line.

Figure 6

Run 2 scattering levels compared to models. The circles are the integrated levels for the first half of Run 2 and the crosses are the integrated levels for the second half of Run 2. Thick lines depict the mean levels. Models: LF swimbladder model, thin line; HF model, thin, dashed line.

Figure 7

Run 3 scattering levels compared to models. The circles are the integrated levels with the solid lines depicting mean levels. Models: LF swimbladder model, thin line; HF models, thin, dashed line.

The NRL swimbladder-scattering model ( Equation (3) ) was fitted, by eye, to the MF and HF volume-scattering data for each run. Matching the NRL swimbladder-scattering model to the resonance peak at 2.5 kHz required two steps. The first was to model the swimbladder scattering so that a peak occurred at 2.5 kHz. The second step was to adjust the backscattering level so that the overall amplitude matched the MF data. Matching the 2.5 kHz peak required several assumptions of swimbladder size and behavior ( Equations (6)–(8) ). The simplest model was to estimate the volume of gas a herring would require at the sea surface to resonate at depth. In this case, the swimbladder compresses in accordance with Boyle's Law with no loss or gain of gas. Fish-flesh density was set at 1.071 g cm ⁻³ ( Brawn, 1969 ) and seawater density was estimated to be 1.026 g cm ⁻³ (8–10 °C and a salinity of 34 from CTD data at the depth of the herring). Using these parameters, swimbladder volumes of herring between 160- and 190-m depth were estimated to be between 1.3 and 1.6 ml. Such volumes would require an uncompressed bladder volume of 22–32 ml at the sea surface, or about four to six times the swimbladder volume of a neutrally buoyant herring (5 ml).

The lengths of Atlantic herring were based on trawl-haul catches (HSMRT #s 48–69). Length frequency distributions of herring from all trawl hauls were similar, with a length range of 19–29 cm and a pooled mean length of 23.4 cm ( Figure 8 ).

Figure 8

The lengths of Atlantic herring (total lengths) caught in trawls near the MF acoustic-measurement sites.

After the 2.5-kHz resonance peak had been fitted, the amplitude of the backscattering curve was modified by adjusting numeric density, via the relationship: Sv = 10 log ₁₀ (N〈σ _bs 〉), where N is numeric density (fish m ⁻³ ) and 〈σ _bs 〉 is the mean backscattering cross-section. This assumes linear addition of individual backscatter ( Foote, 1983 ) and that all fish in the selection were the same species. Numeric densities of 0.025, 0.50, and 1.3 fish m ⁻³ were chosen to fit the scattering model to the data to give resonance peaks occurring at levels of −47, −33, and −30 dB ( Figures 5–7 , respectively). The HF modelling was also adjusted to the numeric densities used in fitting the LF swimbladder model. The fits ranged from good in Run 3, to poor in Run 1. The high-frequency scattering measurements consistently show a minimum Sv at 38 kHz, which the HF empirical model also approximates, but with the dip at about 90 kHz. The overall best fit occurred for Run 3 where both models gave a reasonable fit to all the data and the high-frequency tail of the swimbladder model overlays the HF model reasonably well. For Run 1, the HF data are well below the level of the MF data. One explanation for this poor fit in Run 1 is that Sv values in the MF data are artificially high due to a low SNR (i.e., low overall scattering levels were contaminated by noise).

In addition to the backscattering by Atlantic herring, several other scattering features were observed. A shallow backscattering layer at 40–70 m was detected in the second half of Run 1 and at three places in Run 3 ( Figures 2 and 4 ). Trawl 69 was towed at the depth of this shallow layer for a short time during its retrieval and in addition to the catch of herring, it also contained 16 juvenile silver hake ( Merluccius bilinearis ) of 2–8-cm length (mean 5.3 cm). The NRL swimbladder model was fitted to these silver hakes, but in order to fit the high resonance frequency and the narrow resonance peak reasonably ( Figure 9 ), the model required several unusual adjustments. First, swimbladder volumes were assumed to be 3% of fish-equivalent weight, a reasonable assumption for gadoids and hake ( Nero et al ., 1998 ). Second, to generate a high peak in the resonance, the fish-flesh viscosity term had to be reduced from 20 Pa s, the usual value for small fish, to a value of 10 Pa s. Third, the swimbladder-resonance frequency predicted by the model was too low. To increase the resonance frequency, the juvenile hake swimbladders were assumed to be only 1–2.4% of equivalent weight.

Figure 9

Integrated levels for pelagic layers. (A) Run 1 and (B) Run 3. The circles are the integrated levels with the thick lines depicting mean levels. The LF swimbladder model is shown as the thin line.

4 Discussion

Atlantic herring are the likely source of the main scattering layer observed during this experiment dominating trawl catches targeted at 160–200-m depth. Although silver hake were present in some of the trawls, they were either caught in lesser numbers, or the sampled lengths were too short to contribute to the strong resonance observed at 2.5 kHz. A small percentage of midsize silver hake (L = 22–29cm) were caught in trawls on Georges Bank. However, based on the low catch rates and modelling of these fish using parameters for Pacific hake ( Nero et al ., 1998 ), these silver hake would be resonant at frequencies of 1.6–1.8 kHz and Sv levels would be 6 dB below that of the herring. In addition, Atlantic herring consistently dominate pelagic trawls conducted as part of the annual Atlantic herring acoustic survey in the Georges Bank region (Northeast Fisheries Science Center, unpublished data). In the Cashes Ledge site, a large number of Acadian redfish, Sebastes spp., were caught in trawl #49. While it is believed that these fish aggregate on or near the bottom and were caught during a brief period when the trawl made contact with the sea floor, it is possible that the redfish were in the water column and could have contributed to the poor fit between the swimbladder-scattering model and the data in Run 1.

The shallow backscattering layer detected in Runs 1 and 3 were modelled as juvenile silver hake with unusual adjustments to achieve a very high peak in the resonance curve, suggesting a somewhat “unconstrained” bladder resonance. This may indicate a more bubble-like resonance than has normally been encountered ( Love, 1993 ; Nero et al ., 1998 ).

In addition to the scattering layers attributed to herring and hake, several pelagic aggregations were detected by the mid-frequency system during Run 1. An example of this was from a single “school-like” object detected at a range of about 120–140 m, and at FH numbers 36–38 ( Figure 2 , upper panel). The depth of the aggregation is not greater than 140 m, and likely to be less as it did not appear on the HF echograms. This aggregation appears to coincide with an internal wave feature on the HF echograms, at about ping number 280 ( Figure 2 , lower panel). The best-fit of the swimbladder model to this aggregation, shown in Figure 10 , was obtained using a small (5–7 cm) swimbladder. One hypothesis, based on the depth of the bottom of the internal wave feature on the HF echogram, is that the backscatter is due to a physoclist fish with an uncompressed swimbladder at a depth of 50–80 m. The TS of the aggregation at resonance (+5 dB) suggests it contained approximately 25 000 fish (44 dB above a single fish TS). Several other possible pelagic aggregations were evident in the first half of Run 1 but no attempt was made to model these schools because of the poor quality of the MF signatures. The good fit of the swimbladder model to the first school, suggests that the resonance was from a fish with a swimbladder radius of about 0.2–0.4 cm, possibly juvenile silver hake or other small fish.

Figure 10

Integrated levels for a school observed in Run 1 with the LF swimbladder model (thin line).

Based on the length frequency distribution of Atlantic herring and the assumption of a neutrally buoyant swimbladder at the sea surface, the swimbladders should be resonant at about 5 kHz at 160–190-m depth. This frequency is well above the measured resonance of 2.5 kHz. The resonance at 2.5 kHz at these depths in Runs 1, 2, and 3 indicate that the herring swimbladders would need to have had volumes about five to six times greater than those of fish which were initially neutrally buoyant at the sea surface. The effects of Boyle's law on swimbladder compression for various depths of initial neutral buoyancy are demonstrated in Figure 11 . This figure shows that herring, resonant at 2.5 kHz at 160–190-m depth, would be neutrally buoyant at 40- and 50-m depth and would possess an extra buoyancy of 20–30% of their body mass at the surface.

Figure 11

The effect of Boyle's Law on swimbladder volume for neutral buoyancy models of 0–60 m. The models are shown as seven curves running from the top to bottom of the figure. Line segments crossing these curves indicate the depths of identical resonance frequency. The models were calculated for a mean herring of 23.4-cm total length which when neutrally buoyant would have a swimbladder volume of approximately 5 ml.

To test whether herring were positively or negatively buoyant at the surface, an ad hoc experiment was conducted during the survey. On retrieval of trawl #66, six live herring were selected to estimate their buoyancy by measuring the approximate weight (7 g increments) required to sink them in seawater. Of the six, one herring was negatively buoyant and the remaining five fish gave buoyancies of 5.6–9.8% of fish weight. This suggested that herring are positively buoyant at the sea surface, and that these herring contained up to at least a three times greater volume of gas than a neutrally buoyant fish at the sea surface.

How much gas is present in herring? Gas release is common in fisherman's lore ( Brawn, 1962 ; Blaxter and Hunter, 1982 ) and has been observed from schools ascending towards the sea surface and maintaining position in a current ( Thorne and Thomas, 1990 ), as well as from schools under attack from predators ( Nottestad, 1998 ). Gas release has also been associated with sound production ( Wilson et al ., 2003 ). One may argue that the amount of excess gas may never exceed one full swimbladder volume at the sea surface ( Blaxter et al ., 1979 ; Blaxter and Hunter, 1982 ; Blaxter and Batty, 1984 ). Conversely a large quantity of excess gas must be present in order for gas release to occur at depth. Thorne and Thomas (1990) found that Pacific herring in Puget Sound released excess gas at depths of 40 m and they argue that if Brawn's (1962) observation of a 32% reduction in pressure is required to induce herring to release gas, then the herring in Puget Sound may have been neutrally buoyant at depths as great as 60 m. This is in agreement with the present study.

Our results suggest, in fact, that the swimbladder volume of Atlantic herring can be greater than a “full” swimbladder. The volume of gas is likely to vary daily and seasonally, being dependent on fish migratory behavior, feeding, and reproductive condition ( Ona, 1990 ). The ability of herring to maintain a volume of gas that is greater than one full swimbladder volume is problematic to acoustic surveys of fish because the swimbladder generally contributes between 90 and 95% of the backscattered acoustic energy at high frequencies ( McCartney and Stubbs, 1971 ; Foote, 1980 ). Changes in swimbladder size and form can strongly impact the observed backscatter at high frequency and can introduce bias in the acoustic estimates of fish size and biomass ( Blaxter and Batty, 1990 ; Ona, 1990 , 2003 ). Knowledge of the origin of the gas and the rate at which it can be added to the swimbladder might help in achieving a better understanding of how to predict the gas content and how to interpret the results of acoustic surveys.

4.1 Origin of excess gas

Three mechanisms have been postulated as a source for excess gas in herring: first, the physiological secretion of gas into the swimbladder from the swimbladder epithelium; second, the production of gas from biochemical action of Bacillus bacteria in the gut and transferal to the swimbladder via the pneumatic duct; and thirdly, air gulped at the sea surface with transferal to the swimbladder via the pneumatic duct.

The physiological mechanism for the secretion of swimbladder gas in herring remains unresolved. Fahlen (1967) demonstrated that in the family Clupeidae, particularly the Atlantic herring, there is no development of blood-supplied, swimbladder epithelia equivalent to the rete-mirabili of physoclists. However, some gland cells of the pneumatic-duct epithelia were identified, as secretory cells and there is a good vascular contact with the innermost layers of the wall of the pneumatic duct, thus not completely ruling out the possibility that some gas secretion might occur along it. However, experiments aimed at testing and looking for a gas-secretion mechanism in herring have not demonstrated strong evidence of such a process ( Brawn, 1962 ; Fahlen, 1967 ).

Several studies have examined the potential for fermentation within the digestive tract as a source of swimbladder gas ( Brawn, 1962 ; Fahlen, 1967 ; Blaxter et al ., 1979 ). Obst (1919) cited in Brawn (1962) indicated that several species of Bacillus bacteria were present in the herring digestive system. Most of Brawn's experiments failed to demonstrate an increase in swimbladder gas after herring fed heavily on plankton, although one of six experiments did suggest gas production ( Brawn, 1962 ). Measurements of herring-swimbladder gas show that it contains ratios of O ₂ , CO ₂ , N ₂ , and Argon in the same proportion as atmospheric air ( Fahlen, 1967 ; Blaxter et al ., 1979 ). Tests for increased CO ₂ or methane gas, likely products of gut fermentation, have been negative ( Fahlen, 1967 ; Blaxter et al ., 1979 ). However, in the two studies where gas analysis occurred, the herring were not fed plankton as heavily as they were in Brawn's experiment. Thus, gas production through fermentation remains a possibility.

The swallowing of gas at the sea surface has been noted in several studies ( Brawn, 1962 ; Blaxter and Hunter, 1982 ; Blaxter and Batty, 1984 ). However, a big-gulp of a volume that is five to six times the volume of a non-distended swimbladder would suggest a substantial effort at the sea surface. Could herring gulp and hold such a large volume of gas? Obtaining a volume of gas which is about five to six times the normal “full” swimbladder volume could be possible if the buccal cavity, esophagus, and caecum could hold such a volume, and if the fish could swim downwards with sufficient thrust to overcome an excess buoyancy equivalent to 20–30% of body weight. Examination of herring anatomy and approximations of body-cavity dimensions suggests that five to six times the “normal” swimbladder volume held in the buccal cavity, esophagus, caecum, and bladder would certainly be at the extreme physical limit of these structures. For the 23 cm length herring in this study, the excess buoyancy would be about 25 g or about 25 mN of buoyant force. Studies of sustained swimming in 21 cm length bluegill sunfish demonstrate that they can generate about 75 mN of thrust during sustained swimming at a speed of 1 body length per second ( Drucker and Lauder, 2001 ). Presumably a herring of similar size, capable of much higher swimming speeds ( Blaxter and Hunter, 1982 ), could easily generate enough thrust in a short burst to descend from the surface to depths of 30–40 m. Jumping at the surface would help initiate this descent ( Thorne and Thomas, 1990 ).

Interestingly, both this study and the study of Thorne and Thomas (1990) observed the lower limit of herring neutral buoyancy to be near 40–50 m. This depth limit may be coincidental, or may support the hypothesis of an atmospheric origin of their swimbladder gas. If the gas was from a physiological or digestive mechanism, one might expect neutrally buoyant herring and observations of gas release from fish at depths greater than 50 m. Swimbladder-gas production from fermentation or secretion may occur but these mechanisms probably occur at a slower rate in comparison to surface “gulping”. This study lends support to a primary atmospheric origin of swimbladder gas. Acoustic surveys and estimates of herring TS should consider that they can have a full complement of swimbladder gas to near 40–50-m depth.

We greatly appreciate the assistance of the Captain and crew of NOAA Ship FRV “Delaware II” and the Northeast Fisheries Science Center, Woods Hole, MA. Bjørn E. Axelsen is thanked for suggestions and comments that greatly improved the manuscript. This work was funded by the Office of Naval Research and Naval Research Laboratory through program element 6274N/UW-747-014.