Mysis relicta can be observed on echograms as a sound scattering layer when they migrate into the water column at night to feed on zooplankton. However, quantitative measures of mysid abundance with hydroacoustics requires knowledge of mysid target strength (TS), a method of removing fish echoes and contribution from noise, and an understanding of the effect of range on the ability of hydroacoustics to detect mysids (the detection limit). Comparisons of paired net data and acoustics data from July 7, 2005 yielded a mysid TS of −86.3 dB (9 mm animal) and a biomass TS of −58.4 dB (g dry wt)−1. With ambient noise levels (Sv of −125 dB at 1 m depth) and this TS, we can detect a mysid density of 1 m−3 at 60 m depth with a signal to noise ratio of 3 dB. We present a method to remove backscattering from both noise and fish and apply this method and the new TS data to whole lake acoustic data from Lake Ontario collected in July 25–31, 2005 with a 120 kHz echosounder as part of the annual standard fish survey in that lake. Mysis abundance was strongly depth dependent, with highest densities in areas with bottom depth > 100 m, and few mysids in areas with bottom depth < 50 m. With the data stratified in five bottom depth strata (> 100 m, 100-75 m, 75–50 m, 50–30 m, < 30 m), the whole-lake average mysid density was 118 m−2 (CV 21%) and the whole-lake average mysid biomass was 0.19 g dry wt m−2 (CV 22%) in July 2005. The CVs of these densities also account for uncertainty in the TS estimates. This is comparable to whole-lake density estimates using vertical net tows in November, 2005 (93 m−2, CV 16%).
Mysid shrimp (Mysis relicta) are common in deep northern lakes and can be a major component of the food web (Lasenby et al., 1986). This species is 3–22 mm long, migratory and omnivorous, feeding in the upper hypolimnion and lower metalimnion at night and on or near the bottom substrate during the day. In Lake Ontario, the biomass and rates of zooplanktivory of Mysis relicta have been estimated to be similar to that of alewife (Alosa pseudoharengus), the most abundant planktivorous fish in the lake (Johannsson et al., 2003; Gal et al., 2006). Mysids are also important prey for a variety of fish, including alewife, rainbow smelt (Osmerus mordax), and slimy sculpin (Cottus cognatus) (Mills et al., 2003). Therefore, reliable measures of the abundance of mysids are important for understanding food web dynamics in Lake Ontario.
Available estimates of mysid abundance are based on vertical net tows (Johannsson et al., 2003). But net tows are time consuming to both collect and process, and therefore expensive. Furthermore, it is difficult to use net tows to obtain information on fine-scale vertical and horizontal distributions. Hydroacoustics can provide such information, and is used routinely to measure the distribution and abundance of marine euphausiids and other invertebrates (e.g. Simard and Lavoie, 1999; Foote and Stanton, 2000; Hewitt et al., 2003). Acoustics has also been used to measure distribution and abundance of invertebrates in lakes, including copepods and cladocerans (Megard et al., 1997; Hembre and Megard, 2003; Holbrook et al., 2006), pelagic amphipods (Rudstam et al., 1992; Melnik et al., 1993; Trevorrow and Tanaka, 1997), and insect larvae (Eckmann, 1998; Kubecka et al., 2000; Knudsen et al., 2006). Although a mysid layer is obvious in acoustic echograms, and the behavior of this layer has been used to study vertical migration for decades (Teraguchi et al., 1975; Rudstam et al., 1989; Levy, 1991; Johannsson et al., 2003), the use of echosounders for quantitative estimates of freshwater mysids has been limited (Gal et al., 1999a, 1999b, 2004).
In Lake Ontario and in several of the other Great Lakes, acoustic surveys for forage fish (primarily alewife, rainbow smelt, and various coregonids) using 120-kHz echosounders are conducted each year (Schaner and Lantry, 1999; Mason et al., 2001; Warner et al., 2006). These acoustic surveys can also be used for quantitative estimates of mysid abundance if the backscattering cross section σbs (or its logarithmic transformation target strength, TS) of mysids at 120 kHz is known. Available information on M. relicta TS is limited to higher frequencies (420 kHz, Gal et al., 1999b). We also need measures of the detection range of mysids at 120 kHz and a method to remove the contribution from noise and fish from the acoustic backscattering originating from mysids.
In this paper, we present a method to measure the abundance of mysids in Lake Ontario using acoustic data from the 120 kHz annual fish surveys. We will: 1) describe how to remove the contribution from noise and fish to acoustics backscattering; 2) present data on mysid TS at 120 kHz (from July 7, 2005); 3) calculate the detection range for mysids given the ambient noise level during the survey; 4) present data on the distribution and abundance of mysids in Lake Ontario using the Ontario Ministry of Natural Resources (OMNR) and New York State Department of Environmental Conservation's (NYSDEC) acoustic forage fish survey (from July 25-31, 2005); and 5) compare these results with estimates from a whole lake net survey conducted by the Department of Fisheries and Oceans, Canada, in November, 2005.
Mysis target strength (TS) study
We collected acoustic data and vertical net tow samples at 21 locations with different bottom depths during the night of July 7, 2005, off Oswego, NY (location in Fig. 7), with the U.S. Geological Survey research vessel Kaho. Net samples were obtained with a 1-m diameter conical opening and closing net (mesh size 1 mm) at a number of locations with bottom depth from 170 to 50 m. In all cases, a restricted section of the water column was sampled based on the echograms. The net was retrieved at 0.3 m s−1 to maximize the efficiency of capture of mysids (Nero and Davies, 1982). All sampling was done under minimal red light. We assumed that the net was 100% efficient and calculated volume sampled from the opening of the net and tow length. Boat drift will not affect sample volume as long as the net opening remains perpendicular to the surface of the water. This should have been the case during sampling because July 7 was a calm night and the wire angles were not noticeably different from vertical. Mysids were preserved in alcohol in the field, and then enumerated and measured (up to 100 per sample) in the laboratory. Length measurements (tip of rostrum to the cleft of the telson) were converted to biomass using a length to dry weight regression (Ln (W, dry wt, g) = −12.55 + 2.72 Ln (L, mm), derived from Johannsson (1995) after correcting an error in the original equation). Temperature profiles were collected at 30, 75, 130, and 170 m water depth with a SeaBird depth and temperature profiler.
Acoustic data were collected with a Biosonics Dt-X 120-kHz split beam unit (7.8° beam width, pulse duration 0.6 ms, 1 ping s−1). The transducer was mounted on a tow body positioned 5 m away from the starboard side of the boat at a depth of 1.5 m. The tow body was balanced as to remain horizontal both when the ship was moving and when stationary. Absorption coefficient and sound speed were calculated for 10°C, the average measured temperature between the surface and 40 m depth on July 7. The unit was calibrated by the manufacturer in May 2005, and confirmed to be within 0.1 dB of manufacturer's specification in October 2005 using a −40.4 dB standard copper sphere. Data were collected with a lower threshold of −130 dB.
Whole lake survey
The whole lake acoustic data were collected as part of the annual OMNR-NYSDEC survey for forage fish (alewife and rainbow smelt) in Lake Ontario July 25–31, 2005. Five cross-lake transects in the main lake and three shorter transects in the Kingston Basin were surveyed at night. Data were collected with a similar echosounder as used for TS measurements described above (Biosonics Dt-X, 120-kHz split beam, 7.2° beam width, 0.4-ms pulse length, 1 pings s−1) deployed on a tow body with the transducer at 1.8 m depth. Survey speed varied from 5.5 to7 knots, depending on conditions. This unit was calibrated with a standard −40.7 dB tungsten steel sphere, and no correction was necessary from factory calibrations. All data were collected with a lower threshold of −100 dB.
The two acoustics units used in this study should give comparable data. The difference in pulse length and beam width between the two instruments is accounted for by the calibration and the difference in threshold is not important as data below −100 dB were not useable in the mysid layer due to ambient noise levels.
Auxiliary data were collected with midwater trawls to identify fish targets, and with a Brancker Research temperature profiler at several times during each transect. In addition, a HOBO light meter was deployed to measure shipboard light levels (in lux) every 10 s. In 2005, Tucker trawls were used to investigate smaller targets in the epilimnion. This net caught young-of-year alewife, young-of-year rainbow smelt, and 3-spine sticklebacks (Gasterosteus aculeatus). These are also the main adult fish species caught in offshore Lake Ontario in years when using the standard larger midwater trawl (Schaner and Lantry, 1999; Gal et al., 2006).
Analysis of acoustic data
Analysis of acoustic data from both the Mysis TS study and the whole lake survey was done with EchoView 3.4 (SonarData, 2004, Fig. 1). First, it was necessary to remove scattering due to ambient noise including noise from engines, electrical generators, and the electronics in the echosounder. Volume backscattering strength (Sv) originating from ambient noise was measured to be lower than −125 dB during both the Mysis TS study and the whole lake survey. This noise level increases with depth due to the application of a time varied gain (TVG) function. We removed noise by subtraction of the expected noise level at each depth calculated from a noise level of −125 dB at a range of 1 m and the applied TVG function (Korneliussen, 2000). This noise level was lower than the factory specification (Sv of −123.8 dB at 1-m range).
Second, we removed the contribution from fish to Sv. Fish are large targets relative to mysids and could seriously bias quantitative acoustic estimates of mysids if present. To increase the probability of removing all contributions from fish, we increased the number of data points removed around each target as follows. The data were resampled as 10 data pixels per meter (pixel size of 10 cm) and dilated by replacing each pixel with the maximum found in the surrounding 7 × 7 pixels. Pixels with values above a certain threshold in the dilated data matrix (the fish exclusion threshold, FET, in most cases chosen to be −60 dB—see results) were replaced with “no data” in the original data set (Mask in Fig. 1). We applied the FET in the TS domain where the recorded value (“uncompensated TS”) is proportional to target size at all depths. Re-sampling and dilation are needed to remove weaker echoes from fish associated with the tails of the pulse and when located further from the center of the beam. All areas with excessive noise were also removed from the analysis. Thus, the procedure we developed for extracting mysid volume backscattering coefficients from the 120-kHz data requires four parameters to be specified: a fish exclusion threshold, a noise level at 1 m, a vertical pixel size, and the size of the dilation filter. The first part of the results and discussion section elaborates on the choice of these parameters.
Volume backscattering coefficients (sv, in m2 m−3, Sv = 10 Log10 (sv)) for the TS study was calculated for the depth layers and time periods corresponding to the mysid samples after removing the contributions from noise and fish. We also excluded epilimnetic waters that mysids do not occupy (Gal et al., 2004). Net catches (both density and biomass) and sv values were used to calculate acoustics backscattering cross section (σbs) per mysid and per unit dry weight at 120 kHz.
For the whole lake survey, the mysid layer was identified from the appearance on the echograms, from the location of the layer relative to the temperature profiles, and from the behavior of the scattering layer in response to shipboard light. The acoustic return from mysids is known to decrease dramatically in response to light because mysids will avoid high light by rapidly sinking; a behavior that leads to changes in orientation and therefore to reduced mysid TS. We have observed this behavior in both the field and in the laboratory (Gal et al., 1999b). Consistent with these earlier observations, we noted an immediate decline in the deeper acoustic backscattering layer when the ship's deck lights were turned on, as registered by the logging light sensor (Fig. 2). This identifies this layer as predominantly M. relicta. This behavior was not present in the epilimnetic scattering layer. Mysids are the only abundant larger invertebrate in the hypolimnion of Lake Ontario (Kuns and Sprules, 2000). The largest hypolimnetic copepods (Limnocalanus and Senecella) are less than 3 mm and are expected to have much lower TS with less directional response compared to Mysis (Megard et al., 1997; Foote and Stanton, 2000).
Acoustic data from the mysid layer was exported in 1-min intervals after removing noise and contributions from fish. We used time intervals instead of distance to allow direct comparison with the light sensor data and removed any section with registered light levels. One minute corresponds to a 160-220-m section of a transect at standard survey speed of 5.5-7 knots. Acoustic data from above the mysid layer were excluded, as that depth layer includes backscattering from other zooplankton and from larval fish that will not be removed by the FET. Acoustic data within 2 m from the bottom were also excluded. Thus, all data presented here are from the top of the mysid layer to 2 m above the bottom or the detection limit (60 m, see results), whichever was shallower (Fig. 2).
Mysid density and biomass were calculated for each 1-min interval from the measured sv and the average σbs determined on July 7, 2005. We also exported latitude, longitude, and bottom depth for each interval. Overall density in the lake was calculated by stratifying the data by bottom depth (BD) into four regions 1) BD > 100 m. 2) BD between 100 and 75 m, 3) BD between 75 and 50 m and 4) BD between 50 and 30 m. Bathymetric data from Lake Ontario were from Virden et al. (2000). No data were collected in BD < 30 m and mysid density was assumed to be 0 in those areas. For calculations of variance, each transect was considered one (deep region of the lake) or two sampling units (north and south sections of each transect considered independent). Standard statistical analyses were done with JMP5.1 (SAS) and maps were produced with ArcGIS 9.2 (ESRI).
Results and Discussion
Development of analysis method and choice of fish thresholds
The target strength of fish is substantially greater than that of mysids and it is therefore important to remove fish echoes from volume backscattering attributed to mysids. Correct noise removal is also important because mysids are small targets often observed in deep water. We investigated the importance of the re-sampling and dilation parameters and the noise level on mysid volume backscattering (Table 1). Changing the re-sampling and dilation parameters did not have a large effect. Noise levels were more important. Noise levels higher than −120 dB (at 1 m) would mask mysid backscattering at low and moderate densities and cause an underestimation of lake wide abundance. If noise levels increased from −125 to −110 dB, the mysid scattering would decline by 40% in areas with moderate mysid densities (Table 1). Higher noise levels will also decrease the detection limit. Thus it is important to minimize noise when using acoustics to measure mysid abundance.
The choice of fish exclusion threshold (FET) is important (Fig. 3). The FET should maximize the inclusion of mysid backscattering and minimize the inclusion of fish backscattering in the estimate. The difference between mysid and fish TS is substantial at 120 kHz; the average TS of a 100-mm rainbow smelt is −45.3 dB (Rudstam et al., 2003) and of a 100-mm alewife is −43.7 dB (Warner et al., 2002), which is over four orders of magnitude higher than the TS of a 9 mm mysid (−86.3 dB, see below). Fish TS is variable, however, and depends strongly on tilt angle (Simmonds and MacLennan, 2005). This is true for both rainbow smelt and alewife (Rudstam et al., 2003; Warner et al., 2002). In addition, fish will have lower echo levels when located at the edge of the acoustic beam. A fish returning a −54 dB echo when located in the middle of the beam would return a −60 dB signal if located at the half power beam angle (2-way decline of 6 dB, see Simmonds and MacLennan, 2005). Remember that the FET is applied in the TS domain. But dense concentrations of mysids can result in echo levels over −60 dB in the TS domain (Fig. 2 and Fig. 3). It is therefore not possible to choose one threshold that removes all fish backscattering while retaining all mysid backscattering. The choice of FET is by necessity a compromise between the need to remove most fish echoes while retaining most mysid backscattering.
We used two sources of information for choosing the FET. First we used literature information on the expected TS distributions of the fish species that co-occur with mysids in the meta and hypolimnion of Lake Ontario (yearling and older rainbow smelt and adult alewife, Gal et al., 2006). Over 95% of the TS from yearling and older rainbow smelt are larger than −55 dB in Lake Champlain (Parker Stetter et al., 2006). For adult alewife, 95% of the TS measured on 130-mm alewife in a net cage were larger than −58 dB (Rudstam et al., unpublished data). These results suggest that a threshold in the TS domain needs to be smaller than −64 dB to remove the contribution from all fish even when the fish is located at the edge of the beam. Second, we plotted sv as a function of the FET (Fig. 3). In areas of moderate mysid and fish densities (“Mysis”), sv declines with FET but there is a saddle between −55 and −65 dB when most fish backscattering is removed while most mysid scattering is retained (Fig. 3). With FET smaller than −65 dB, mysid backscattering is also excluded. In areas with low mysid TS, because of their response to light (“Light”), sv increases slowly for thresholds larger than −60 dB indicating that some fish backscattering is still present at a FET of −60 dB. Together, this suggests that a threshold of −60 dB will remove most fish backscattering while retaining most mysid backscattering. Therefore we chose a FET of −60 dB for most areas. However, when mysids are dense (“Dense Mysis”), this threshold removed too much mysid backscattering (Fig. 3) and we therefore applied an FET of −55 dB in those regions. These regions were obvious on echograms as regions without obvious fish traces and where the whole mysid layer could be masked out with a −60 dB FET. As a result, it was possible to visually identify areas requiring the higher FET and manually adjust this value. This occurred at depths between 18 and 25 m in some sections of transects 2 and 3 (Fig. 7). Holbrook et al. (2006) used similar thresholds for removing fish (−61 to −63 dB) from zooplankton acoustic signals in Lake Superior, but it is unclear from their paper if this was a threshold applied in the Sv or TS domain.
It is better to replace the areas with fish echoes with “no-data” rather than a value of 0. Replacing these pixels with “no-data” is equivalent to assuming that mysid density in this volume of water is the same as in surrounding water. Replacing these pixels with a 0 value is equivalent to assuming mysids are not present within this volume of water. Although mysids can avoid fish in the laboratory (Boscarino et al., 2007), it is unlikely they can avoid fish in the whole pulse volume removed. We would not expect to find mysids in fish stomachs if they were not present with fish in the lake, and this is clearly not the case as mysids are important prey items for both alewife and rainbow smelt in Lake Ontario (Johansson et al., 2003; Gal et al., 2006). In addition, targets in the whole pulse volume contribute backscattering to each pixel, not just the volume of water immediately surrounding the fish (Simmonds and MacLennan, 2005). The pulse volume increases with depth, beam width, and pulse length. At 20-m depth, a beam width of 7o, and a pulse length of 0.4 ms, the pulse volume is larger than 1.5 m3. We believe this is too large a volume to be devoid of mysids even when a fish is present.
Target strength of mysids at 120 kHz
Mysids were present in the echogram from the Mysis TS study as a dense layer from just below the thermocline to 40 or 50 m depth (Fig. 2). This distribution is similar to earlier observations in Lake Ontario (Johannsson et al., 2003; Gal et al., 2004) and can be predicted from the response of mysids to light and temperature (Gal et al., 2004; Boscarino et al., 2007). Vertical net hauls confirmed that this layer consisted of M. relicta, and that few mysids were found above the thermocline.
We caught 3,845 mysids with lengths between 3 and 18 mm (average 9.0 mm) in 21 net hauls (Fig. 4). The mysid population showed two distinct length peaks representing animals born in 2004 and 2005. M. relicta has an 18 to 24 month generation time in Lake Ontario (Johannsson et al., 2003). In four samples, the catch was less than 20 animals and the average size less than 6 mm because few older animals were caught. These data points were excluded a priori for two reasons. First, small mysids likely have different TS than larger mysids and may therefore bias the relationship. Second, if mysids are both rare and small, they may no longer be the dominant back-scatterers as other zooplankton also contributes to backscattering. In the remaining 17 samples, the average length of the animals ranged from 7.7 to 10.3 mm and included animals in both size groups. The total number of paired net-acoustics comparisons used for the regressions was therefore 17. Including all 21 samples a posteriori increases the R2 of the relationship (from an R2 of 0.68 with 17 data points to an R2 of 0.72 with 21 data points).
There was a significant correlation between volume backscattering strength (Sv, dB) and both mysid density and mysid biomass obtained with vertical net hauls in July 2005 (log-transformed data, R2 = 0.68 and 0.67, respectively, N = 17, Fig. 5). The slope was close to 0.1 for both regressions. A slope of 0.1 in the dB scale is equivalent to a linear relationship between sv and density on a linear scale (any other slope in the dB scale represents a power function between sv and density, which indicate that the σ bsof an individual mysid is dependent on mysid density). Thus, the contribution of backscattering from individual mysids was additive, as expected from acoustic theory (Simmonds and MacLennan, 2005). The additive property of acoustic backscattering has been shown to hold for fish unless densities are extremely high (Foote, 1983). We calculated an average σbs by dividing sv by mysid density or biomass for each data pair, and calculating averages and confidence limits from these data points. The average σ bs per mysid was 2.35· 10−9 (1 SE 3.5· 10–10) or in logarithmic form a TS of −86.3 dB (95% CL −87.8 dB to −85.1 dB, N = 17). The average backscattering per g dry weight was 1.44· 10−6 m2 g−1 or in logarithmic form −58.4 dB (95% CL −60.1 to −57.2 dB, N = 17). These results are slightly different from values calculated from the regressions in Fig. 5 because of the logarithmic transformations used in that figure. Note that these measures assume that the catchability of the net is 100%, which is likely not the case. Therefore, our TS values may be biased high and the mysid density estimates therefore biased low (see also Gal et al., 1999b).
A TS of a 9-mm mysid of −86.3 dB is somewhat low compared to expectations from theory and other measurements in the literature. A theoretical model for backscattering from a 9-mm bent fluid-filled cylinder (Stanton and Chu, 1993) predicts a TS of −82.4 dB. The same model with parameters from Gal et al. (1999b) predicts a TS of −77 dB at 120 kHz for a 9-mm M. relicta. But theoretical models can give widely different results. David et al. (1999) used three scattering models to estimate TS of 14 to 16-mm mysid Siriella jaltensis and obtained values between −100 dB and −79 dB depending on the model used. Empirical measures of mysid TS are rare. Gal et al. (1999b) reported TS of M. relicta at 420 kHz to be −73.1 dB using comparison with net hauls and between −74.8 and −76 dB using in situ estimates and comparison with optical plankton counters. Wiebe et al. (1990) measured TS at 420 kHz of 14 to 31-mm Neomysis rayii to be between −80.9 and −72.9 dB with no clear length dependence. A pelagic amphipod (Macrohectopus branickii) with similar size, shape and behavior as M. relicta has a TS of −82 dB at 200 kHz for a 15-mm animal (Rudstam et al., 1992). Although these literature values are greater than our measurements, most of them are for larger animals and higher frequencies, for which we generally expect larger TS (but see Knudsen et al., 2006).
We are only aware of literature values for TS per unit dry weight of mysids at 420-430 kHz. The equation in Gal et al. (1999b) yields an Sv of −54.0 dB for a biomass of 1 g dwt m−3 of M. relicta at 420 kHz, which is higher than our estimate and may be due to the higher frequency used by Gal et al.,Holbrook et al. (2006) presented a relationship between zooplankton dry weight and backscattering from Lake Superior that includes mysids. Using their equation, the volume back scattering for 1 g dwt m−3 is −62.6 dB at 430 kHz. Although lower than our estimate, their data includes substantial contributions from copepods and cladocerans that are expected to yield lower acoustic backscatter per unit weight due to their smaller size relative to the acoustic wavelength (Foote and Stanton, 2000).
Given a TS of −86.3 dB and a noise level of −125 dB at 1 m, we can calculate the expected signal to noise ratio (SNR) for a mysid density of 1 animal m−3 at different depths. If we accept a SNR of 3 dB (a factor of 2), the detection limit is 60 m (a noise level of −125 dB at 1 m is amplified to −89.3 dB at a range of 60 m by the TVG function). This detection range depends on density; the detection range for a density of 5 mysids m−3 (Sv of −79.3 dB) is 120 m. A detection range of 60 m should have been sufficient to insonify most mysids during the whole lake survey. Peak mysid abundance was between 20 and 30 m depth during this survey (quarter moon on July 28, 2005), and we know from other studies during the summer that most mysids are above 50 m in low moon light (Gal et al., 1999a, 2004). However, during full moons and during the fall when the thermocline deepens, some mysids are found in deeper water than 60 m (Gal et al., 2004) and the detection limit can then be a concern for whole lake mysid estimates.
Distribution of mysids in Lake Ontario
Mysid areal abundance in the 1-min sections ranged between 0.4 and 667 mysids m−2(depths from the thermocline to the bottom or to 60 m, Fig. 6). Average abundance for the five longer transects (1–5, Fig. 7) was 161 m−2 (range for whole transect averages were 130-211 m−2). Few mysids were detected in the shallow Kingston Basin (transect 6, average 10 mysids m−2). We assumed that no mysids were present in epilimnetic waters. This is supported by observations of acoustics backscattering with and without shipboard light (Fig. 2), previous net surveys in various regions of the lake (Kuns and Sprules, 2000; Johannsson et al., 2003, Gal et al., 2006), and laboratory experiments (Boscarino et al., 2007). Mysids prefer temperatures between 6 and 8°C, and never entered 18°C and above in laboratory experiments (Boscarino et al., 2007). Fish densities were too high close to shore to effectively remove their contribution with our method. Therefore we also excluded areas with bottom depth less than 30 m. We do not expect mysids to be present in that region of the lake (Johannsson et al., 2003).
Abundance was correlated with bottom depth (R2 = 0.44, Fig. 6). Higher mysid abundances were found when bottom depth was > 100 m. Abundance declined as bottom depth decreased and was very low at bottom depth < 50 m (Fig. 6, Table 2). This pattern, also observed in past surveys and in the 2005 net surveys, has been attributed to higher fish predation in shallow areas (Johannsson et al., 2003). Mysids may also actively avoid these areas either because of a lack of suitable temperature and light regimes, or because of high concentrations of fish. In the laboratory, M. relicta avoid water from alewife holding tanks (Boscarino et al., 2007).
Because of the known strong depth dependence of mysid abundance, we estimated abundance by stratifying the data in sections with bottom depth > 100 m, 75–100 m, 50–75 m and 30–50 m (Table 2). Whole lake average abundance per unit area was obtained by weighting the abundances in each region by the proportion of the lake with that bottom depth. The resulting acoustic lake-wide average mysid abundance for July 2005 was 118 m−2 (Table 2). This corresponds to an average biomass of 0.22 g dry wt m−2, with the mean weight of the mysids caught July 7 (1.88 10−3 g dry wt mysid−1). Corresponding acoustic biomass derived from the measured area backscattering and the estimated TS per g dry wt was 0.19 g dry wt m−1. This discrepancy is likely due to the non-linear relationships between length, weight, and acoustic backscattering. To compare with fish biomass expressed in kg ha−1, we converted dry weight to wet weight assuming a dry/wet weight ratio of 20% (Morgan, 1976). Wet weight mysid biomass is then 11.1 kg ha−1 (from abundance and average weight) or 9.4 kg ha−1 (directly from acoustics), which is similar to the average biomass estimated for forage fish in Lakes Ontario and Michigan (Warner et al., 2006; Gal et al., 2006). Because mysids are predominantly zooplanktivores and feed at specific rates similar to fish, these results further confirm that mysids must be important predators on zooplankton in the Lake Ontario food web (Johannsson et al., 1994, 2003; Gal et al., 2006).
Uncertainty in these acoustic estimates of whole lake mysid density includes uncertainty in average σbs and net catchability, spatial coverage, mysids present below the 60 m detection limit, application of the fish exclusion threshold, definition of the top of the mysid layer, calibration, noise, and sea state. Although the combined uncertainty is difficult to estimate, the main contributors are believed to be from uncertainty in σbs and sampling coverage (Simmonds et al., 1992; Rose et al., 2000; Demer, 2004). It is possible to combine the uncertainty in σbs and sampling coverage using formulas for the variance of a product of two independent variables (Mood et al., 1974). The combined CV accounting for these two processes was 21% for density and 22% for biomass. This does not include net efficiency during the TS study, which would bias σbshigh and therefore abundance estimates low. Any mysids below 60 m were not included and this would also bias the abundance estimate low. Calibration uncertainty is likely small compared to these other sources as both units were stable over time.
We compared these acoustically derived estimates with net hauls from a whole-lake survey conducted in November 2005. Although the whole lake acoustics survey and the whole lake net survey are separated in time by 3 months, we do expect similar whole lake densities in July and November because mysids live up to 2 years with low reproductive rates (10–50 young and usually one brood per life time). In Lake Ontario there are several overlapping generations and peak production of young is in April and August (Johannsson et al., 2003). In this survey, 79 vertical net hauls from close to the bottom to the surface were collected at 48 stations with bottom depths between 50 and 236 m covering the whole lake (Johannsson and Bowen, unpubl. data). Using the average from each station as an independent sample and the same stratification as for the acoustic estimate, we calculated the average lake-wide abundance from net hauls to be 93 m−2 with a CV of 16% (Table 2). This is similar to and not significantly different from the estimate based on acoustics (118 mysid m−2, CV 21%).
Acoustic data can be used to construct maps of mysid density for Lake Ontario (Fig. 7). In addition to confirming the decrease in mysid abundance towards the shore, we observed a decrease in abundance in the middle of the lake in transects 4 and 5 (Fig. 7). This is not directly related to water depth but could be related to an area of lower productivity in the center of the counter clockwise gyres that are often present in Lake Ontario (Beletsky et al., 1999).
We have developed a method for extracting useful information on mysid abundance and distribution from fisheries acoustics data that is routinely collected as part of a forage fish assessment in Lake Ontario. Although conceptually simple, our method would have been difficult to implement without recent advances in acoustic analysis software. For example, Holbrook et al. (2006) used manual removal of fish echoes to estimate zooplankton abundance using acoustics in Lake Superior, a method they considered too time consuming for use for a whole survey. Others have relied on finding sections of the echograms with no fish (e.g. Gal et al., 1999b, 2004). Our method is more efficient and allows for estimates of mysid abundance on a whole lake scale with reasonable effort. However, it may be more difficult to implement if hypolimnetic fish density is high.
Improved fish exclusion algorithms can probably be derived if data are available from more frequencies (Fernandes et al., 2006). A lower frequency would have a larger difference between fish and mysid TS, which would help identify and remove fish backscattering from the analysis, thus reducing the subjectivity in the choice of FET. The combination of 38 and 120 kHz has been used successfully in marine applications for separating regions with krill and fish (e.g. Cochrane et al., 1991; Swartzman et al., 1999; McKelvey and Wilson, 2006). These marine applications use the relative difference between frequencies to produce synthetic echograms of this difference in backscattering (e.g. Arnone et al., 1990; Korneliussen and Ona, 2003) that can be used to classify regions dominated by different organisms. Our approach has the advantage of allowing the removal of the contributions from single fish within mysid layers from data collected with one frequency, but it does require sufficient separation between the acoustic return of fish and the invertebrates of interest. Alternative approaches for when this is not the case, have been proposed to separate backscatter from the phantom midge larvae Chaoborus and fish (Eckmann, 1998; Malinen et al., 2005).
This paper is the result of research supported by New York Sea Grant project R/CE-23 with additional funding provided by the Ontario Ministry of Natural Resources, New York State Department of Environmental Conservation, Department of Fisheries and Oceans Canada, USGS–Biological Research Division, and the Environmental Protection Agency Lake Ontario lower trophic level assessment project. The views expressed are those of the authors and do not necessarily reflect the views of the granting institutions. The U. S. and Canadian Governments are authorized to produce and distribute reprints for governmental purposes notwithstanding any copyright notation that may appear herein. We are grateful to the crews of the Kaho and the Seth Green for help with data collection and to Matthew Wilson at SonarData for help with development of the analysis method. This article is contribution # 244 of the Cornell Biological Field Station and contribution # 1424 of the USGS, Great Lakes Science Center.