Size-selective predation by three estuarine zooplanktivorous fish species

Keywords: Atypichthys strigatus, diet, estuarine ecosystem, gut contents, laser optical plankton counter, optimal foraging theory, planktivory, prey choice, Scorpis lineolata, Trachurus novaezelandiae.

Optimal foraging theory is a broad framework that seeks to understand foraging behaviour, including predation that transfers biomass from lower to higher trophic levels. Underlying this theory is the assumption that an animal will make foraging decisions that maximise the amount of energy ingested, while minimising the energy used during feeding (Pyke et al. 1977). Although sometimes criticised for being too simple and not representative of the natural environment (Pierce and Ollason 1987), this theory continues to provide a useful framework for exploring foraging behaviour and has been used recently to make and test predictions about foraging for a range of taxa, including marine mammals (Foo et al. 2016; Tyson et al. 2016), birds (Hernández-Pliego et al. 2017), lions (Barnardo et al. 2020) and fish (Thygesen et al. 2016).

Marine ecosystems are strongly structured by size (Sheldon et al. 1972; Andersen et al. 2016; Hatton et al. 2021), resulting in small prey being highly abundant compared with larger prey. Yet, small prey contain less biomass per individual and may be harder to detect than are larger prey (Hansen et al. 2013). This presents marine predators with a choice of whether to feed on the abundant easy to catch prey or target larger prey?

Zooplanktivorous fish are a key trophic link between abundant zooplankton and larger predators. On temperate rocky reefs, zooplankton can support over 50% of the fish community biomass, with much of this flowing through small zooplanktivorous fish (Truong et al. 2017; Goddard et al. 2022). The foraging strategy of zooplanktivorous fish presents an interesting test of optimal foraging theory. Zooplankton are highly abundant (often >1000 individuals m⁻³) and small zooplankton are often an order of magnitude more abundant than large zooplankton (Sheldon et al. 1972; Heneghan et al. 2016), meaning that the chance of a zooplanktivorous fish encountering small prey is much greater than is their chance of encountering large prey. However, a large copepod can contain up to 15 times more biomass than a small copepod (Chisholm and Roff 1990), which may make large zooplankton a preferential prey source. This creates a situation where zooplanktivorous fish could feed randomly on the basis of encounter rates of zooplankton (commonly known as filter feeding), essentially eating mostly smaller zooplankton or they could target larger zooplankton (particulate feeding) if the trade-off in terms of biomass return for energy used in searching for and capturing the larger prey is favourable. If consuming the greatest amount of biomass for the least effort is the goal of zooplanktivorous fish, then perhaps their feeding strategy is mediated not by the abundance of different prey but by the biomass of each prey type in the environment. It is plausible that they would target the prey type with the highest biomass in the environment, rather than abundance. Some species have been observed to switch between filter-feeding and particulate-feeding behaviours in response to changes in prey density (Jansen et al. 2019). A similar but alternative theory may be the targeting of calories rather as biomass as the calories are a direct measure of energy (Cumminns and Wuycheck 1971; Balogh et al. 2022). This is most likely if the different prey types have large differences in calorific content but not biomass.

Prey selectivity of zooplankton by zooplanktivorous fish has previously been explored, with a focus on mouth size and gill raker dimensions proving mixed results. Although mouth size is important for small fish (<30 mm), fish are quickly capable of feeding on almost all zooplankton and, at larger fish sizes, the limiting factor becomes handling time and capturing prey (Wanzenbock 1995). Similarly, gill raker size has been shown to not be a limiting factor on the capture of small particles, although it may reduce the efficiency of capture, leading to increased relative capture of larger particles (Langeland and Nøst 1995; Budy et al. 2005). Overall, prey selectivity in zooplanktivorous fish is not simply driven by morphology, but there is also active selection of different size prey, particularly for larger zooplanktivorous fish, which have the ability to capture all sizes of prey and can significantly alter their own behaviour while foraging (Tanaka et al. 2006).

The goal of our study was to explore the foraging strategy of three common zooplanktivorous fish in Sydney Harbour. To investigate selective feeding, we measured the size-structured zooplankton abundance and biomass in the environment over the summer period, and compared this with the size-structured diet of planktivorous fish collected over the same period. The specific aims were to (1) examine the average size-structured zooplankton abundance and biomass in Sydney Harbour over a 3-month period, (2) determine the prey size and diet composition of three planktivorous estuarine fish, and (3) compare the prey size of the gut contents with the size structure of zooplankton available in the water to quantify size-selective predation.

Zooplankton were sampled in the lower reach of Sydney Harbour between November 2013 and February 2014 (summer) at three sites (Site 1: −33.834°, 151.278°, Site 2: −33.839°, 151.277°, Site 3: −33.849°, 151.266°), along the southern shore close to the mouth of the estuary and the dominant tidal flow. The total distance between sites was 2.1 km. All sites were sampled in the morning on 10 days, during five ebb and five flood tides (sampled 2–3 h after the predicted high or low tide). At each site, three horizontal replicate plankton tows were made with a 40-cm diameter, 100-µm mesh net at 1-m depth. A mechanical flowmeter (Model 2030R, General Oceanics Inc., Miami, FL, USA) was attached to the net to calculate the sampling volume. A detailed description of the Sydney Harbour ecosystem is available in Johnston et al. (2015).

The zooplankton size distribution and biomass from all towed samples was determined using a laboratory-based laser optical plankton counter (LOPC; Herman et al. 2004), coupled to a pump system (Moore and Suthers 2006). When a particle passed through the beam of the LOPC, the attenuance of light was detected and recorded as the corresponding equivalent spherical diameter (ESD) of the particle. The zooplankton size was classified into a size-frequency distribution with 30-µm bins. Only particles between 300 and 3000 µm ESD were included, because particles outside this size range were unlikely to be sampled accurately by the plankton net (Moore and Suthers 2006). Zooplankton biomass was calculated from the volume of a prolate spheroid (ratio of 1:3, width:length) and the specific gravity of water (Suthers et al. 2006; Garcia et al. 2022).

Because the goal of this study was to investigate foraging behaviour of zooplanktivores rather than variation in the zooplankton community, all zooplankton samples were averaged together to provide a representative zooplankton snapshot over the whole sampling period. This aligns with the fish collection discussed below.

Over the same time period as the zooplankton sampling, Atypichthys strigatus (Günther, n = 17), Trachurus novaezelandiae (Richardson, n = 24) and Scorpis lineolata (Kner, n = 22) were collected either by using unbaited hook-and-line or spearfishing from the study area. These methods have been used successfully in other studies to sample these species while avoiding contamination of gut contents with bait (Gaston and Suthers 2004; Champion et al. 2015). On the basis of previous research, these species are suspected zooplanktivores and were observed to be the most abundant around our sampling sites. It was confirmed that these species are some of the most abundant small fish in Sydney Harbour by using data from fish surveys undertaken by the Reef Life Survey (Edgar and Stuart-Smith 2014). Individual fish were immediately placed on ice and later frozen, until dietary analysis took place. Fish were collected throughout the sampling period, irrespective of tide, and although fish were not collected evenly in space, all were collected within 500 m of the zooplankton sampling sites. In all, 58 of the 63 fish (92%) were collected on the same day as the zooplankton samples, with five individuals of A. strigatus being collected opportunistically on different days.

The gut contents of each fish were weighed and prey items were identified to a coarse taxonomic resolution. The fullness and percentage (by volume) of plant matter, zooplankton and unidentifiable material were recorded for each gut. Because it was not feasible to use the LOPC for partially digested gut contents, the size distribution (ESD) of zooplankton in the gut contents was manually determined from the length and width of zooplankton, to compare with the size distribution of zooplankton in the water column (the LOPC data). From each gut, a random sample of each zooplankton taxonomic group was photographed using a Leica M80 Microscope with Leica Application Suite (ver. 4.4, Leica Microsystems GmbH, Wetzlar, Germany). ImageJ (ver. 1.48, see https://imagej.nih.gov/ij/; Schneider et al. 2012) was used to measure the length (l, mm) and width (w, mm) of each prey item in the photographs. Length and width measurements were converted into an ESD (µm) by assuming the shape of an ellipsoid and using the following equation:

To determine the size range of zooplankton having the highest incidence of consumption by estuarine planktivorous fish, the ESD measurements of identified prey within the fish guts were compiled into frequency histograms for each fish (30-µm bins; corresponding to those of the LOPC). The prey-size distributions of all individuals from each species were then averaged to obtain an average prey-size distribution for each species.

All analysis was conducted using R (ver. 4.0.2, R Foundation for Statistical Computing, Vienna, Austria). A split-plot ANOVA was used to test for differences in total zooplankton biomass between tides and sites, with tide and site as fixed factors and day as a random factor. It was a split-plot design because only one level of tide (ebb or flood) was sampled per day (the ‘plot’). The zooplankton biomass was log₁₀-transformed to meet the assumptions of normality and homogeneity of variance.

To assess selective predation in our zooplanktivorous fish, we used a simulation approach similar approach to Chappell and Smith (2016). We generated three null models of expected prey distributions under the assumption that non-selective feeding would result in prey-size distributions reflecting the zooplankton distribution in the environment. There was one model for each species. Using the abundance (%) of each 30 µm ESD zooplankton size class in the water, we created a simulated population of known size distribution. From this population, we then drew 2000 random samples of x prey items, where x represents the average number of prey items found in each species gut. From these 2000 samples, we generated a mean size-frequency distribution with a 95% confidence interval. We then assessed size-selective predation by comparing the observed predation rates for each size class with the null predictions. If a size class occurred in the fish guts more frequently than predicted by the null simulation, this suggested that this size class was disproportionately preyed on. To estimate the magnitude of the selectivity, we again followed Chappell and Smith (2016) and present the proportional effect (PE) ratio, which is the ratio of observed predation to the expected predation. A PE < 1 signifies underselection, whereas a PE > 1 signifies overselection. Compared with traditional selectivity indices such as those of Chesson (1978), the simulation approach allows for robust analyses of low sample sizes and provides a benchmark of null selection to compare against (Chappell and Smith 2016).

In total, 86 plankton tows were analysed using the LOPC. Four samples, each from a different site or day, contained too much gelatinous material for accurate analysis; so, they were excluded. The zooplankton biomass varied over an order of magnitude both among days and within days among sites. The smallest zooplankton biomass was recorded at Site 1 with 106.0 mg m⁻³ (25 November 2013), and the largest was 1722.2 mg m⁻³, at Site 3 (27 February 2014). The largest range in zooplankton biomass across the three sites on a single day was 1120.0 mg m⁻³ (27 February 2014). No significant difference in zooplankton biomass was found between ebb and flood tides (ANOVA: F_1,8 = 1.17, P = 0.31), but Site 3 contained significantly more zooplankton biomass than did Site 1 across all tides (ANOVA: F_2,72 = 5.48, P < 0.01, Supplementary Fig. S1). No significant interaction was found between tide and site (ANOVA: F_2,72 = 0.83, P = 0.44).

Despite this variation in total biomass among samples, when the percentage biomass and abundances in each size class were investigated, consistent patterns were observed and we present an overall average zooplankton distribution, which shows consistent declines in abundance with size, and a peak in biomass between 495 and 705 µm ESD (Fig. 1). Among sites, there were only minor differences in the biomass percentage size distributions, with Site 3 having a lower percentage biomass than the other sites in the small bins (<375 µm ESD) and Site 1 having a slightly more even distribution of biomass (a lower peak between 495 and 705 µm ESD; Supplementary Fig. S1).

Fig. 1.  Mean zooplankton size structure in the lower Sydney Harbour during our study. Error bars show 1 s.e.

In total, 4140 prey items were identified in the guts of 17 Atypichthys strigatus, 22 Scorpis lineolata and 24 Trachurus novazelandiae individuals. No guts were empty, although seven contained fewer than ten identifiable prey items. A. strigatus, T. novaezelandiae and S. lineolata all consumed zooplankton. In A. strigatus and T. novaezelandiae, zooplankton comprised 100% of the identifiable gut contents. In S. lineolata, zooplankton comprised 40%, with plant matter making up the other 60%. Copepods were the most abundant prey items for all species (Fig. 2a). They were found in 97% of all guts (Fig. 2b) and represented 64, 66 and 41% of all prey items by count in A. strigatus, S. lineolata and T. novaezelandiae respectively. A. strigatus had the greatest average number of prey items in their guts (145.1 ± 21.4 s.e.), followed by S. lineolata (46.9 ± 6.8 s.e.), and T. novaezelandiae had the fewest (26.8 ± 4.4 s.e.; Table 1). S. lineolata was the only species to consume plant material, sand grains or barnacle cirri. T. novaezelandiae had the smallest species richness in its diet, containing eight prey types, compared with 13 and 14 prey types for A. strigatus and S. lineolata respectively.

Fig. 2.  (a) Average number of identifiable individuals for each prey type per gut (+s.e.). Note the log₁₀ scale on the y-axis. (b) Frequency occurrence (%) of each prey type, which shows the percentage of fish guts that contained at least one identifiable individual of the prey type.

Table 1.  Summary of gut contents and fish total lengths.

In total, 1997 prey items from gut contents were measured and converted to an equivalent spherical diameter to be comparable with the zooplankton measurements from the LOPC. The three species of fish showed a similar pattern of prey sizes in their guts, although T. novaezelandiae contained smaller prey (Figs 3, 4). The smallest size class of prey was 240–270 µm ESD, with only 0.06% (±0.02 s.d.) of prey in A. strigatus and this size class was not observed in the other two species. This smallest size class was smaller than that resolved by the LOPC, so was not included in comparisons with zooplankton in the environment. The largest size class containing prey was 2940–2970 µm ESD, containing only a single prey item, observed in the gut of a T. novaezelandiae. The median prey size classes of each species were as follows: A. strigatus, 540–570 µm ESD; S. lineolata, 540–570 µm ESD; and T. novaezelandiae, 480–510 µm ESD.

Fig. 3.  Comparison of observed zooplanktivorous fish prey-size (solid line) and the expected prey-size distribution (dashed line) on the basis of abundance of different zooplankton size classes in the environment for (a) Atypichthys strigatus, (b) Scorpis lineolata, and (c) Trachurus novaezelandiae. Error bars show 95% confidence intervals around the mean. Size classes greater than 1200 µm ESD are not displayed because they contained few prey items (less than 0.02% total).

Fig. 4.  Comparison of observed zooplanktivorous fish prey-size (solid line) and the expected prey-size distribution (dashed line) on the basis of total biomass of different zooplankton size classes in the environment for (a) Atypichthys strigatus, (b) Scorpis lineolata, and (c) Trachurus novaezelandiae. Error bars show 95% confidence intervals around the mean. Size classes greater than 1200 µm ESD are not displayed because they contained few prey items (less than 0.02% total).

Strong evidence of size-selective predation was found for A. strigatus and S. lineolata. When compared with the null model of expected diet proportions based on the assumption of random feeding and the observed size structure in the environment, all species showed an underselection of prey in both small (<480 µm ESD for A. strigatus, <450 µm ESD for S. lineolata and <420 µm ESD for T. novaezelandiae) and large (>870 µm ESD for A. strigatus, >690 µm ESD for S. lineolata and >630 µm ESD for T. novaezelandiae) size classes (Fig. 3, Supplementary Table S1). This corresponded to strong evidence of an overselection of medium size particles in A. strigatus (480–840 µm ESD) and S. lineolata (480–690 µm ESD). By contrast, only weak evidence of selectivity was evident in T. novaezelandiae, where the small and large prey underselected, whereas the moderate-size prey showed only weak evidence of selectivity, although this could be due to the low average number of prey items within their guts creating large 95% confidence intervals (Fig. 3, Supplementary Table S1).

When compared with the null model of expected diet proportions calculated using the biomass of each size class in the environment rather than the abundance, the evidence for size-selective feeding was reduced and our observed prey sizes aligned more closely with expectation for A. strigatus and S. lineolata (Fig. 4). Whereas there continued to be some evidence of underselection at small and large prey sizes, there was vastly reduced evidence of positive selectivity occurring in any size class (Supplementary Table S1). The peak in observed prey size corresponded to the peak in size classes that contained the most biomass in the environment, suggesting that A. strigatus and S. lineolata were likely to be targeting available biomass. By contrast, T. novaezelandiae showed slightly more evidence of selectivity in the biomass model than did the abundance model, with the peak in size classes consumed being smaller than would be expected on the basis of available biomass, suggesting that T. novaezelandiae was likely to be feeding on the basis of abundance, not biomass.

This study showed evidence of size-selective predation occurring in three estuarine zooplanktivorous fish, with two species (A. strigatus and S. lineolata) showing strong evidence, whereas T. novaezelandiae showed only weak evidence of size-selective predation. By comparing the observed prey-size compositions with expected prey-size compositions on the basis of feeding relative to environmental abundance and biomass in each size class, we propose that size-selective predation is driven by the increased amount of biomass available in the environment for preferred size classes relative to the other size classes. Our finding supports the theory of optimal foraging theory. Because the goal of foraging is to consume the greatest biomass for the least effort, the size classes with the greatest biomass in them represent the most ‘profitable’ food sources. Understanding the foraging decisions made by planktivorous fish is vital because they often link zooplankton as an abundant resource with the fisheries typically comprising higher trophic levels (Pikitch et al. 2014).

Previous studies have defined T. novaezelandiae as a planktivore (Kingsford 1989; Bulman et al. 2001; Dawson et al. 2020), whereas A. strigatus and S. lineolata have previously been defined as piscivores (Bulman et al. 2001) or planktivores (Kingsford 1989; Glasby and Kingsford 1994; Champion et al. 2015). However, this study found only evidence of planktivory with some benthic foraging only for A. strigatus and omnivorory for S. lineolata. It is likely that A. strigatus and S. lineolata may have a flexible diet that can vary with ontogeny and in both time and space, particularly in offshore locations such as those in Bulman et al. (2001). The classification of T. novaezelandiae is consistent with other members of the Trachurus genus (Tanaka et al. 2006).

There is strong evidence for prey-size selection occurring in all three species. Prey smaller than 465 µm ESD and larger than 900 µm ESD were found in the gut contents significantly less often than would be expected on the basis of their environmental abundance. This was matched with significantly more prey of a moderate size (480–~780 µm ESD) being observed in the gut contents of both A. strigatus and S. lineolata than expected on the basis of abundance. Owing to low numbers of prey in T. novaezelandiae gut contents, there was insufficient power to detect any positive prey selection and, instead, the contents matched the expected consumption of moderate-sized prey. There was variation among species, with A. strigatus showing the strongest evidence of prey selectivity while also consuming a higher proportion of larger prey (>705 µm ESD) than do the other species. Whereas there are no comparable studies of prey-size selection for T. novaezelandiae and S. lineolata, it has previously been observed that, in coastal environments, A. strigatus had a preference for slightly larger zooplankton (Champion et al. 2015). This may be because in the offshore location, the available zooplankton were also larger than they were in the estuarine samples in the current study (Champion et al. 2015).

When compared with biomass available in each size class of zooplankton prey (rather than raw abundance), the observed prey-size distributions were a much closer match to the expected distributions. There continued to be an underselection of small and large particles, but the peak in prey size consumed closely matched the size classes of zooplankton in the environment containing the most zooplankton. Avoidance of small prey as an active choice was demonstrated previously in a mesocosm experiment, which demonstrated that turbidity was not an influence on selection by planktivorous bluegill sunfish (Gardner 1981). The present study suggests that the reason for this avoidance may be that the energetic costs of capturing the small prey are not as efficient as when targeting the size classes with the most biomass. In the future, it would be useful to investigate whether prey selectivity changes with the ontogeny of the predator because this has been observed in larval and juvenile yellow perch that switch from a high capture-efficiency technique to a lower capture but higher biomass-return strategy as they mature (Graeb et al. 2004).

As our prey-size measurements for the environmental zooplankton and gut contents were obtained using different methodologies (LOPC vs manual sizing), it is possible that there could be a methodological bias between the measurements. The LOPC has been rigorously validated and used in many studies and has been shown to provide accurate measurements across the size range observed in our study (Herman et al. 2004; Herman and Harvey 2006). By contrast, manual measurements are variable in method and may be subject to bias, particularly because zooplankton in gut contents are not in pristine condition. Previous manual measurements of copepods have shown to be highly accurate (within 1% accuracy), giving us confidence in our method (Araoz 1991). Our measurement method followed that of Skjoldal et al. (2013) and because the majority of zooplankton prey observed in this study were typical copepods with an elliptical shape (Araoz 1991), our method of calculating their volume is valid for most prey items, although it does ignore appendages, which had often fallen off. This may result in a small underestimation of size; however, because the appendages contain very small amounts of total volume and, therefore, biomass (<5%), it is reasonable to assume that our measurements and prey size estimate would be within 5% accuracy and any variation around this would not change the interpretations of the patterns observed in this study.

Copepods were the most abundant prey item in the gut contents of all three species, followed by cladocerans for A. strigatus and T. novaezelandiae and large dinoflagellates for S. lineolata (along with plant material), with other prey types being almost an order of magnitude less abundant. Copepods and cladocerans have very similar calorific contents (Cumminns and Wuycheck 1971), suggesting that the selection in this case was most likely driven by the available biomass and search time trade-off (which will correlate with available calories).

The overlapping prey-size range also provides insight into the niche partitioning occurring in Sydney Harbour; although all three of our studied species consume zooplankton of the same size, they occur in different habitats and are consuming the zooplankton in different parts of the estuary. A. strigatus is an extremely reef-associated zooplanktivore, S. lineolata is also reef associated but also eats plants, and T. novaezelandiae is a zooplanktivore but is less associated with reefs, particularly within estuarine environments.

Although not consistent, there was, on average, higher zooplankton biomass at Site 3 (inner site) than at Site 1 (outer site). This suggests that zooplankton may accumulate inside the estuary and not all is discharged on the ebb tide. This accumulation may be due to either estuarine production or retention within the estuary (Avila et al. 2012) or active use of tidal currents by zooplankton (Simons et al. 2006), and further research looking at estuarine gradients in zooplankton biomass and productivity would provide valuable insight in how estuarine and coastal zooplankton support higher trophic levels. The idea of retention is supported by previous research showing that 50% of the water in this lower-estuary region of Sydney Harbour is retained and not exchanged with the ocean for up to 80 days, increasing up to 90% in the inner estuary (Das et al. 2000).

Regardless of the large variation observed in zooplankton biomass and abundance in the environment, once this was standardised to percentage composition of size classes, there was a consistent trend in the proportions of each size class present. This showed that although there are fluctuations in the overall zooplankton abundance, the size structure of the zooplankton community is stable. This stable size structure of the zooplankton community potentially enables the zooplanktivorous fish to match their prey-size preference to the greatest available biomass that occurs in specific size classes.

Increased understanding of lower trophic level predation dynamics will enable ecosystem modellers to better capture predator–prey dynamics within their models. It is now well recognised that modelling zooplankton specifically in ecosystem models is important (Heneghan et al. 2016), and that size-based modelling approaches may offer significant advantages over traditional food-web models (Blanchard et al. 2017), particularly when predators have diverse prey. By empirically matching predator–prey dynamics to specific size classes of zooplankton with a mechanism (biomass availability), our findings should enable more confidence in the modelling of zooplanktivorous fish predation.

This study has both demonstrated size-selective zooplankton predation by estuarine fish and shown that the targeted size range is likely to be preferred because it contains the largest amount of total biomass, although the extent of the selectivity varies by species. This results in the targeted size classes being the most efficient prey source in terms of return and effort during foraging. This has important considerations for zooplanktivorous fish trophic ecology because it highlights how these species do not feed randomly on zooplankton in the water. The information highlighted in this paper will enable the creation of more accurate lower trophic level and size-based ecosystem models.

All code and data are available at https://github.com/HaydenSchilling/Estuarine-Size-Selective-Predation.

The authors declare that they have no conflicts of interest.

This study was conducted as part of the Sydney Institute of Marine Science Sydney Harbour Research Project, which provided financial assistance through their Seed Funding Program. J. A. Smith and J. D. Everett were funded by the Australian Research Council (LP120100592 and DPI120100728 respectively). D. P. Harrison was funded by an Australian Postgraduate Award.

Supplementary material is available online.