Large zooplankton has an important role for the functioning of the ecosystem in many lakes. Most are predatory, which makes them both competitors as well as food for many planktivorous fish species. In general, it has been elusive to measure the abundance of this group of zooplankton with traditional sampling methods, particularly in large volumes of water. In this study we tested the potential and adequacy of multi-frequency hydroacoustics as a method to map the spatial patterns in abundance of the large zooplankton in Lake Vänern, Lake Vättern and Lake Mälaren. In addition, we used a plankton net to groundtruth biovolume estimates vertically, and a Tucker trawl for the horizontal distribution. Two frequencies were used to separate the acoustic backscattering from fish, mysids and plankton. The main target species were Leptodora kindtii, Bythotrephes longimanus and Limnocalanus macrurus.
The plankton communities were very heterogeneous both within and among the lakes. Lake Vänern and Lake Mälaren were dominated by Leptodora, whereas Lake Vättern was dominated by Limnocalanus. Bottom depth at the sample site was the most important community structuring factor.
Our results indicate that the biomass of large predatory zooplankton is comparably high and in most surveyed areas even higher than the biomass of planktivorous fish. Consequently, they are more important for the dynamics of lake food webs than previously assumed. Using multi-frequency hydroacoustics show promising results and with some alterations to the sampling design this would be a valuable addition to the traditional zooplankton monitoring in large lakes.
There has been a surge of interest in large zooplankton in the last decade, mainly due to the large effects of the invasion of Bythotrephes in North America (Yan et al., 2011). Several of the large bodied species are predatory, and thus direct competitors to planktivorous fish as well as important prey. The largest species that are found in large lakes in Northern Europe are the predatory cladocerans Bythotrephes longimanus (Figure 1a) and Leptodora kindti (Figure 1b), and the copepod Limnocalanus macrurus (Figure 1c), an omnivore (Warren, 1985). There are also carnivorous/omnivorous copepods of the genus Heterocope, but these are present at substantially lower abundances than Limnocalanus in the Swedish large lakes. These species are important prey organisms for the dominant planktivorous fish species, Smelt (Osmerus eperlanus) and Vendace (Coregonus albula) (Vallin, 1969; Northcote and Rundberg, 1970; Nilsson, 1979; Northcote and Hammar, 2006). Previous studies have indicated that large zooplankton occur more frequently in the diets of fish than biomass estimates from regular sampling would suggest (Nilsson, 1979). They have also been shown to be important for early life stages of the commercially important Pike-Perch (Ginter et al., 2011). These species have been particularly well studied in waters where they recently have been introduced (Yan et al., 2011). Once they successfully colonize, they have been found to profoundly influence the trophic structure of the pelagic ecosystem (Rennie et al., 2011). The most striking example is the effects of the invasion of Bythotrephes longimanus in North American lakes where it has given rise to species shifts and reduced abundance and diversity of native zooplankton communities (Barbiero and Tuchman, 2004). In contrast, in Europe they have been found to occur in many lakes with well-functioning ecosystems (Hessen et al., 2011).
There are no standard methods for sampling of zooplankton in European lakes but sampling is often carried out using various plankton nets. Large zooplankton are hard to quantify in traditional zooplankton monitoring programmes, possibly as a result of sampling small volumes of water which is less optimal for these animals that occur in lower numbers than smaller zooplankton and are known to aggregate in patches (Boudreau and Yan, 2004; Weisz and Yan, 2010). One alternative method to plankton nets that could be more suitable to handle the patchy occurrence of large zooplankton is hydroacoustics that in recent years has been successfully used to quantify other organisms than fish, e.g. Krill, Mysids (Rudstam et al., 2008b; Axenrot et al., 2009) and insect larvae (Knudsen and Larsson, 2009). The evolution of technology and methods for analysis have opened new opportunities to quantify the abundance of invertebrates and thus also zooplankton. To do this, the hydroacoustic reflection from zooplankton must be detected and separated from the reflection of fish (Jurvelius et al., 2008). Today, this is possible by covering the same volume of water using multiple frequencies and by masking the relatively stronger reflections of fish.
In this study we wanted to answer three questions: (1) Can large zooplankton be monitored using a combination of Tucker trawl hauls and multi-frequency hydroacoustics? (2) How do these estimates of large zooplankton densities compare to regular zooplankton monitoring programmes using other methods? (3) How abundant are large zooplankton and how are they distributed in the three largest lakes of Sweden?
Data were collected in the three largest lakes of Sweden, Lake Vänern, the largest lake in the European Union, Lake Vättern and Lake Mälaren. The lakes have very different characters, and span from deep to shallow and hyperoligotrophic to mesotrophic/eutrophic (Table 1). They are all surveyed annually in a national environmental monitoring programme. Zooplankton is monitored by collecting water at various depths all through the water column with a Ruttner collector and a Clark-Bumpus net with a 120 μm mesh in Lake Vänern and Lake Vättern and with a Rodhe collector in L. Mälaren (5 litre volume). Fish is monitored in hydroacoustic surveys in the night-time along parallel transects. Fish species composition and size distributions are verified using mid-water trawling. Both surveys are performed in August and September.
Zooplankton samples in our study were collected and compared with concurrent hydroacoustical data in a depth gradient at six deep sites in Lake Vänern, four in Lake Vättern and two in Lake Mälaren. Two of the sites in each lake are also zooplankton sample stations in the regular monitoring programmes, thus enabling a direct comparison with the data obtained in our study. Plankton data were collected using a large plankton net and a Tucker trawl to cover both the vertical and horizontal distribution of large zooplankton. In Lake Vänern, sampling was conducted both during night-time and day-time to test the importance of diel variation on zooplankton abundance and noise levels from other animals such as insect larvae and mysids.
The plankton net (Aquatic Research Instruments, simple plankton net) was deployed from the surface to the vicinity of the bottom and then back to the surface. It had a diameter of 0.6 m, length 1.8 m and a mesh size of 500 μm. The sampled volume was measured with a flowmeter that was placed in the opening of the net (OceanTest Equipment Flowmeter MF315).
The Tucker trawl (opening = 0.25 m2) had three separate nets and cod-ends (mesh size = 500 μm). The nets could be opened separately during trawling in order to sample different depths. Trawling speed was approximately 2 knots, equaling to circa 1 m s−1. Trawling depth was adjusted by varying the length of wire and measuring the angle of the wire according to the formula: trawled depth = sin(wire-angle × π / 180) × wire length. However, the exact trawled depths were recorded using a probe (CTM208, software SDA v. 1.83, SST GmbH). The same probe was also used to measure the water temperature (surface to bottom) at each sampling site. Samples were taken from three depths at each site, close to the surface and just above and below the thermocline. The sites in Lake Mälaren were considerably shallower, thus only the two shallowest depth zones were sampled.
The zooplankton samples were preserved in ethanol and later analyzed in the laboratory. The samples were counted in a stereomicroscope after subsampling in a Wiborg whirling vessel. Where possible, at least 200 individuals of Bythotrephes, Leptodora, Limnocalanus and Heterocope were counted in each sample (median 416 individuals). All treatment of samples and calculations of biomasses and biovolume zooplankton mm−3 of water were made according to the Swedish national standard for zooplankton monitoring (SS-EN 15510:2006).
Due to technical problems with the flow-meter, the sampled volume with the plankton net was adjusted for the samples where it malfunctioned. The sampled volume of these hauls were adjusted using an empirical model (r2 = 0.97, p < 0.001, N = 20) based on the data with flowmeter data: adjusted volume (m3) = 2.95 + 0.672 × 2 × depth (m) × plankton net opening area (m2).
The hydroacoustic surveys were performed directly adjacent to zooplankton sampling sites in daytime and at night, in darkness, on Lake Vänern, and exclusively at night in Lakes Vättern and Mälaren, in August and September. The hydroacoustic data were collected using two echo sounders, 38 and 120 kHz, with hull mounted transducers at 1.5 m depth on R/V Asterix (Simrad EK60 with ES 38B and ES 120 7C). Both transducers have a theoretical beam width of 7 degrees and were mounted to cover the same sample volume (Korneliussen et al., 2008). The mounting depth and frequencies renders an efficient near-field limit at four meters depth. The echo sounders concurrently emitted sound pulses (pings) with a ping rate set to 3 s−1 and pulse length to 512 μs. The power was set to 1000 and 100 W for 38 and 120 kHz, respectively. Vessel speed when trawling and collecting hydroacoustic data was 2 knots, equaling 1 m s−1. Data were collected over 1842 m in Lake Vänern, 4300 m in Lake Vättern and 1477 m in Lake Mälaren. The echo sounders were calibrated according to recommendations (Foote et al., 1987).
Hydroacoustic data were analyzed using Echoview 5.1 (Sonar Data). Initially, noise was removed taking into account that mysids and zooplankton are weak scatterers. Fish echoes in 38 and 120 kHz hydroacoustic data were identified and used to construct a mask to filter out all backscattering (Sv, dB; Simmonds and MacLennan, 2007) generated by fish. Due to the strong difference in backscattering properties of fish and mysids/zooplankton, small areas around the fish echoes were also filtered out.
Based on the results of previous studies (Axenrot et al., 2009), the remaining filtered hydroacoustic backscattering at 120 kHz was then compared to the groundtruth from zooplankton sampling to evaluate if the density estimates were accurately representing the abundance of macrozooplankton. Filtered hydroacoustic data were matched to the exact columns sampled in the Tucker trawl hauls and vertical tow hauls. Thereafter we conducted stepwise multiple regression procedures to analyze the correlations between back-scattered energy and the density of various animals caught in the hauls. All data on densities of animals were log-transformed prior to analyses. Pearson product-moment correlation coefficients were computed to evaluate the relationships between trawled depth and zooplankton densities (numbers m−3). Differences between day and night were conducted with Mann-Whitney U-test. Three samples were excluded from the analysis due to problems with the masking of fish. These three samples also had very low densities of zooplankton (<50 mm3 m−3).
Large zooplankton and other invertebrates, mainly mysids and Chaoborus, were caught in the Tucker trawl hauls (Table 2). The densities of these organism groups were related to the hydroacoustic signal strength. The backscattered energy (Sv) was affected by the amount of zooplankton and the amount of Chaoborus. Sv = −97.74 + 18.8 × log (Chaoborus [numbers m−3] + 1) + 4.5 × log (large zooplankton [mm3 m−3]), r2 = 0.935, r2adj = 0. 924, p = 0.025, N = 16. The low densities of mysids that were observed had no effect on Sv (T = −0.62, p = 0.548).
Limnocalanus was more abundant at larger depths (r = 0.448, n = 20, p = 0.048). In contrast, Leptodora was less abundant at larger depths (r = −0.446, n = 20, p = 0.049). Bythotrephes abundances were highly variable and consequently their distribution did not partition significantly with depth (r = −0.320, n = 20, p = 0.168). Bythotrephes are known to aggregate in dense patches (Weisz and Yan, 2010) leading to very high variability among samples. This was evident also in our study systems where the coefficient of variation varied from 93–110% in the three different lakes.
The amount of zooplankton was not affected by the light level in Lake Vänern, where both day and night sampling with the Tucker trawl was tested (Mann–Whitney W = 27, ndark = 6, nlight = 3, P = 0.51). The median abundance at night was 184 mm3 m−3 and in daytime 277 mm3 m−3. Neither mysids nor Chaoborus were caught in daylight. The results from the vertical plankton net samples revealed a large variation in species composition and the density of large zooplankton both within and between lakes (Figure 2). The glacial-relict Limnocalanus was more abundant in the deeper parts of the lakes (r = 0.556, n = 38, p < 0.0001). Chaoborus was only found in the more eutrophic Lake Mälaren. Some species were only found in very low numbers, such as Caligus lacustris (two individuals in Lake Vänern). Heterocope also occurred in relatively low numbers, particularly in Lake Vänern where it was almost absent and in Lake Vättern were it constituted less than 2% of the total biomass of large zooplankton.
The densities of large zooplankton were considerably higher in this study than in the regular monitoring program during August and September (Table 3). The median density was 158 ind m−3 in the study sampling and 36 in the regular monitoring (Mann.Whitney W = 1592, nstudy = 38, nregular = 30, P < 0.0003). The total biomass of large zooplankton species was high in all three lakes. Based on the Tucker trawl samples, mean biomass in Lake Vänern was 110 kg ha−1 (±21.2 S.E.), in Lake Vättern 39 kg ha−1 (±12.2 S.E.) and in the two basins of L. Mälaren 78 kg ha−1 (±8.0 S.E.). These figures can be compared with the biomass of pelagic fish calculated from hydroacoustic (120 kHz) and mid-water trawling data in August and September 2011. Pelagic fish biomass adjacent to the zooplankton stations in Lake Vänern was 35.5 kg ha−1 (±9.42 C.I.), in Lake Vättern 13.7 kg ha−1 (±4.69 C.I.) and in the two basins of Lake Mälaren, Granfjärden 127 kg ha−1 (±48.0 C.I.) and Prästfjärden 78.2 kg ha−1 (±29.1 C.I.).
Our study indicated that the density of large zooplankton in all three studied species was higher than what has previously been known for these systems implying that large zooplankton have a more important role in pelagic foodwebs than previously thought. Planktivorous fish, here Smelt and Vendace, are believed to be the main plankton-feeding organisms in the majority of large lakes in the Northern hemisphere, but see Gal et al. (2006). These planktivorous fish have been shown to feed selectively on large zooplankton, probably since they offer more energy per food item and thus, even though they are less numerous than other zooplankton, may constitute more preferable food items. The estimated biomasses of large zooplankton exceeded the biomasses of pelagic fish in Lake Vänern and Lake Vättern while they were at a similar or slightly lower level in the two basins of Lake Mälaren. One possible explanation to the higher estimate of fish biomass in Lake Mälaren could be backscattering from high densities of Chaoborus, erroneously interpreted as fish. The biomass estimates for large zooplankton are also higher than the biomass of mysids in this study and in an earlier study in Lake Vättern (Axenrot et al., 2009).
Given that these large zooplankton also are competitors to planktivorous fish their occurrence may influence the trophic dynamics of plankton food webs. There are many well established examples of top-down control by predatory fish on lake production, with phytoplankton or filamentous algae at the bottom of the foodweb (Persson et al., 1992). Plankton foodwebs in small lakes have often been used as classic examples of reticulate interactions between species and a higher tendency for trophic cascades to occur. The presence of a large biomass of large predatory zooplankton would give rise to an additional, often neglected, trophic level in plankton food webs (Rennie et al., 2011). This could alter the interpretation of lake foodweb dynamics, e.g. how predation effects from top consumers cascade down to lower trophic levels (Carpenter et al., 1985; Pace et al., 1999). In addition, the production of fish in the lakes with predatory zooplankton is likely to be lower due to energy loss due to the added trophic level.
The spatially heterogeneous distribution of zooplankton induces problems when sampling with traditional methods that only cover small sample volumes. This has been particularly problematic when sampling Bythotrephes that has a tendency to aggregate due to currents (Weisz and Yan, 2010). We suggest that large zooplankton instead can be monitored using hydroacoustics with two frequencies that allow coverage of substantially larger water volumes. However, it was impossible to discern between different species of zooplankton on the one hand, and between zooplankton and other invertebrates on the other hand. As no mysids were caught in the hauls in the daytime this problem might be less pronounced. This is in line with previous studies of mysids (Levy, 1991; Axenrot et al., 2009). Of the tested methods, the Tucker trawl was the most suitable sampling device for collecting groundtruthing data.
We show that hydroacoustics in combination with Tucker trawling for groundtruthing can be used to estimate the abundance of large zooplankton. The characteristic backscattering properties of mysids (Rudstam et al., 2008a) and the strong reflection from Chaoborus might shadow the weaker acoustic reflection of zooplankton. Consequently, we suggest that future hydroacoustic monitoring of zooplankton should be conducted in daylight when both the mysids and Chaoborus are close to the bottom and can be easily separated from the zooplankton. Developing combined, integrated monitoring of fish, mysids and zooplankton would substantially improve the potential to accurately assess the interactions among the main organism groups in pelagic ecosystems of large temperate lakes.
We acknowledge the excellent working conditions on R/V Asterix and the help with counting mysids and additional sampling from Björn Kinsten. We are also grateful for valuable assistance with the hydroacoustics analysis from Professor Lars Rudstam.