Introduction

Feeding is known as a key factor affecting the larval mortality of pelagic fishes (Cushing 1990; Dickmann et al. 2007; Houde 2008). Since feeding in early life stages seems to have a major impact on fluctuations in the recruitment of these fishes, numerous studies have been conducted on the feeding ecology of larval and juvenile pelagic fishes. Partly due to the difficulty of rearing pelagic fishes in the laboratory, most of the studies in the last four decades were based on field-collected samples. Examples of such field studies include Young and Davis (1992), Uotani (1985), Cass-Calay (2003), Reiss et al. (2005), Dickmann et al. (2007), and Intxausti et al. (2017), which reported similar trends for various pelagic fish species: larger larvae and/or juveniles ingested more and/or larger prey; smaller individuals tended to have empty guts; and the main prey items were crustaceans consisting mainly of copepodites.

However, possible biases related to net sampling make it uncertain whether the results obtained from field studies truly represent the situation of live fish in the sea (reviewed elsewhere, e.g. Young et al. 2009; Intxausti et al. 2017). A first example is related to the calculation of selectivity index for certain prey organisms. This calculation from field samples is based on a precondition that potential prey organisms and fish were sympatrically distributed in the source sea area. However, this premise is questionable considering that zooplankton in the sea are patchily distributed at the scale of meters or even smaller (Young et al. 2009), and that plankton nets are usually towed for a much longer distance. As a second example, the main prey of larval and juvenile pelagic fishes found in field studies, i.e. crustaceans mainly copepodites (Ciechomski 1967; Tudela et al. 2002; Sampey et al. 2007; Morote et al. 2008, 2010; Catalan et al. 2010; Yasue et al. 2011), may have appeared so merely because their residues were easy to identify by regular microscopic analysis, and prey items which were decomposed without leaving any digestive residue might have been overlooked (Pepin and Dower 2007). Third, for net-collected samples, low numbers of larvae containing prey do not necessarily mean that they had low feeding incidence. This is because of the possibility that the weakest or malnourished larvae might have been selectively captured and/or the prey in their guts might have been regurgitated or excreted due to the stress of capture and handling (Arthur 1976; Conway et al. 1998). Given the fact that smaller larvae of pelagic fish are more likely to expel their gut contents as a result of sampling trauma than larger larvae (Yamashita 1990), the reason why gut content tends to increase with body size (Conway et al. 1998; Islam and Tanaka 2009; Catalan et al. 2010) may be that larger larvae with a more developed digestive tract can have a greater ability to retain the ingested prey during net sampling. These possibilities, which may affect the reliability of the net sampling methodology itself to an extent not yet known, have prevented field researchers from definitively determining feeding characteristics of larval and juvenile pelagic fishes in previous field studies (e.g. Islam and Tanaka 2009; Intxausti et al. 2017). Therefore, evaluating the extent of the influence of possible biases related to net sampling would help us understand the precise feeding ecology of pelagic fish larvae and juveniles.

To clarify these points, empirical studies that meet the following conditions are needed: (i) experimental fish should be fed with a sympatrically distributed zooplankton assemblage; (ii) the zooplankton assemblage should include not only crustaceans but also other taxa without shells, and the experimental period should be carefully controlled so as to prevent flimsy prey being digested; and (iii) experimental fish should be fixed in a way that minimizes regurgitation or excretion of ingested prey. This study targeted on juvenile chub mackerel (Scomber japonicus) and larval/juvenile Japanese anchovy (Engraulis japonicus) since comparable field studies were abundant (e.g. Ozawa et al. 1991; Sassa et al. 2008; Shoji et al. 2009; Islam and Tanaka 2009; Okazaki et al. 2019a; Taga et al. 2019). Both species are widely distributed in the Northwest Pacific including the whole costal area of Japan (Suhara et al. 2013; Nakamura et al. 2020). Not only are these commercially important as adults, but larvae of Japanese anchovy are also fished and traded in Japan under the commercial name of shirasu. For both species, field studies concerning their feeding ecology in early life stages have reported that larvae and juveniles mainly feed on copepods and appendicularians, while utilizing various preys in other taxa supplementarily (Sassa et al. 2008; Islam and Tanaka 2009; Okazaki et al. 2019a; Taga et al. 2019). In this study, fishes were hatchery-reared for use in an experiment in which they were fed with a wild-caught zooplankton assemblage for 15–30 min and then fixed under anesthesia so as to prevent the ingested prey items getting regurgitated or excreted. Using these samples of two species, we examined the relationship between fish size and prey number/size in the gut and the selectivity on each prey item.

Methods

Egg hatching and fish rearing

Eggs of chub mackerel and Japanese anchovy were obtained from induced spawning of captive broodstock maintained at the Hakatajima Station, Japan Fisheries Research and Education Agency (133.10°E, 34.20°N, Imabari, Japan). For chub mackerel, spawning was induced by hormonal injection following the procedure of Nyuji et al. (2012). In order to obtain a sufficient number of eggs, we used two parent groups of different ages: 1 year old and 3 years old. Twenty individuals of 1-year-old fish and thirteen individuals of 3-year-old fish of each sex were injected intramuscularly with 400 µg kg–1 of body weight gonadotropin-releasing hormone analogue (GnRHa) on 6 and 20-May and 1-Jun-2021, and maintained in 50,000-L (3-year-old individuals) and 20,000-L (1-year-old individuals) square tanks with circulating seawater. Approximately 10,000 eggs were collected evenly from the two parental age groups three times on 19, 24, and 31-May. For Japanese anchovy, approximately 15,000 naturally spawned (without hormonal injection) eggs were collected three times on 26-October and 3 and 15-November-2021 from c. 50 parental fish, which had been held in two 2,000-L plastic tanks. The eggs were placed in three 1,000-L plastic tanks, each for a different spawning date, and incubated under a photoperiod cycle of 13 h light and 11 h dark. The tanks were maintained at an average water temperature of 22 °C. The larvae were fed with a mixture of rotifers (Brachionus) and planktonic algae (Isochrysis galbana or Nannochloropsis oculata) once per day until 8 days post hatch (dph) for chub mackerel and 29 (first and second groups) or 24 (third group) days post hatch for Japanese anchovy, after which they were also fed with newly hatched brine shrimp (Artemia salina) once per day and reared until 27 to 30 dph (chub mackerel) or 35 to 40 dph (Japanese anchovy). Ration level was kept high (rotifers: > 20 individuals/mL; brine shrimp: > 0.5 individuals/mL) throughout the experimental period. All chub mackerel used in the experiment were juveniles, whereas Japanese anchovy included both larvae and juveniles. Following Islam and Tanaka (2009), we regarded individuals larger than 18 mm SL as juveniles.

Zooplankton collection and feeding experiment

The feeding experiment was conducted three times (hereinafter each referred to as a trial) on 9 and 21-Jun and 2-July for chub mackerel and on 3, 12, and 24-December for Japanese anchovy. Wild zooplankton were collected mainly using a light trap (illustrated in Fig. 1a; see also Fig. S1 for a real picture) on the night before each trial.

Fig. 1
figure 1

Outline of the experimental procedure. Zooplankton were collected mainly by the light trap (a). Zooplankton aggregated under the light were drawn into the plastic pipe and collected in the plankton net. Water flow in the pipe was generated by the air bubbles moving upward through the pipe. The whole apparatus was surrounded by a fish net of 10 mm mesh so as to prevent the aggregated zooplankton being eaten by fish. Collected zooplankton were once introduced in the plankton net (0.1 mm mesh) and kept for 4 to 5 h (b). Deposits were vacuumed out by a siphonic system to remove dead or weakened zooplankton. All of the left plankton were concentrated to a mass of 2 L, a proportion of which was introduced into the experimental beakers (c)

Full size image

Zooplankton assembled under the light were introduced into the plastic pipe by the water flow generated by ascending air bubbles, then led to the plankton net (0.33 mm mesh, φ 45 cm), and finally collected in the cod end jar attached to the end of the plankton net. The whole apparatus was surrounded by a fish net of 10 mm mesh so as to prevent the aggregated plankton being eaten by large predators. Zooplankton were collected throughout the night until the cod end jar was retrieved the next morning. For all trials, zooplankton were subsidiarily collected using a drifting plankton net in the tide during the daytime of the day before each trial so as to prepare zooplankton species that are not photopositive. Three same-size nets (0.33 mm mesh, φ 45 cm) were used for this zooplankton collection, in which the nets were drifted for a total of 1 h. For the first trial on chub mackerel, zooplankton collected from an aquaculture pond (near the Hakatajima Station) on the day of the trial were also used as an additional source. One of the three nets was used for the collection from the aquaculture pond, and a total of 1,000-L bucketed water was filtered through this net. All of the collected zooplankton were once introduced in the plankton net (0.1 mm mesh) placed in a 500-L tank, and kept under aeration at 22 °C for 4 to 5 h (Fig. 1b). Deposits were vacuumed out to remove dead or weakened zooplankton. All of the left zooplankton were concentrated to a mass of 2 L in a beaker (hereinafter referred to as the source), from which zooplankton were introduced in each trial into two 2-L experimental beakers containing filtered seawater and fish. Since the concentration of source zooplankton varied by date, we counted the number of zooplankton included in 1 ml of source water and calculated the number of zooplankton contained in the source water prior to every trial. Based on this calculation result, we determined the ratio of source seawater containing zooplankton to seawater containing fish that would make the experimental beakers have a zooplankton concentration of 1 individual/ml (Fig. 1c). Accordingly, seawater was introduced into each experimental beaker together with 5 individuals of juvenile chub mackerel or 10 to 11 individuals of larval/juvenile Japanese anchovy, followed by zooplankton. In order to confirm that the experimental fish would eat familiar prey under these experimental conditions, one additional 2-L experimental beaker was prepared for all trials, containing brine shrimp of 1 individual/ml and the same number of experimental fish (not illustrated in Fig. 1). The experimental beakers were lit from the four sides and the top and bottom by LED illumination to prevent zooplankton converging on any particular part thereof. The illumination intensity was 160 lx and the water temperature was kept at 22 °C through room temperature control. The trials were terminated after 15 min (chub mackerel) or 30 min (Japanese anchovy) by humane killing with an overdose of 0.1% 2-phenoxyethanol. We carefully monitored the experimental beaker after this anesthetizing procedure and did not observe any apparent regurgitation/excretion. After standard length (SL) was measured, the fish were kept in 10% buffered formaldehyde solution. Unused source zooplankton were fixed and preserved in seawater with 5% buffered formaldehyde solution for a later examination of the taxonomic composition of the source zooplankton. Gut content analysis was carried out on 30 chub mackerel and 59 Japanese anchovy. The gut contents of each fish and the source zooplankton were examined under a microscope, and all the zooplankton were identified to the lowest taxonomic level possible. For each taxonomic group of the source zooplankton, the length of up to 20 individuals was measured. For the gut contents, both length and width were measured unless they were digested to an extent that would not allow for accurate measurement. Since ingested brine shrimp was apparent from the fish surface due to its orange color for most individuals, whether fish ate brine shrimp or not was primarily judged by external observation. For large chub mackerel juveniles, whose gut contents were unclear from the surface, gut contents were dissected out.

Data analysis

All the statistical procedures were conducted using the statistical software R version 4.1.2 (R Core Team 2021). To clarify whether SL had a significant effect on prey number in the gut, generalized linear models (GLMs) based on the Poisson distribution were applied. The response variable was prey number in the gut, and the explanatory variable was SL. In addition, generalized linear mixed models (GLMMs) based on the Gamma distribution were applied to find any significant relationship between prey size (length and width) in the gut and SL. The response variable was prey length or prey width, the explanatory variable was SL, and the random effect was individual ID. Ingested preys whose size could not be measured accurately were excluded from this analysis. The data from all three trials were pooled in these GLM and GLMM analyses for each species. The selectivity of fishes for prey i was quantified using Chesson’s (Chesson 1978) α-selectivity index:

$${a}_{i}=\left({{r}_{i}{p}_{i}}^{-1}\right){\left(\sum _{j=1}^{m}{{r}_{j}{p}_{j}}^{-1}\right)}^{-1}$$

where ri and pi are the proportions of prey i in the diet and in the water in the experimental beakers, respectively. The value of α ranges between 0 and 1 and deviates asymmetrically from the reciprocal of the considered m prey items. We adopted this index because it provides a quantitative description of the relative preference for a prey type in relation to the other available prey types regardless of prey density (Pearre 1982). The unidentified organisms (unidentified crustacean fragments and unidentified remains) among the gut contents were excluded from the calculation as well as a prey item that was found in the gut but not in the source water (namely Harpacticoida, for the second trial on chub mackerel and for the whole calculation on Japanese anchovy). In this study, 25 types of prey for chub mackerel juveniles and 18 types of prey for larval/juvenile Japanese anchovy were considered. Therefore α’s critical value was set at 1/25 and 1/18, respectively. Values lower than this threshold indicate rejection and values higher than this indicate selection to each of the prey items. Due to the asymmetry of the index, we avoided discussing differences in negative selectivity. The index was computed for each of the three trials on chub mackerel but only for the second trial on Japanese anchovy. This is because most of the Japanese anchovy individuals had empty guts in the first and third trials. The size distribution (average and standard error) of each zooplankton taxon in the source was calculated to examine the relationship between α and prey size. For this calculation, the source zooplankton for all three trials were pooled. For most of the taxa, the number of individuals detected per trial was less than 20, and the length data of all individuals were used. However, more than 20 individuals were detected in some of the taxa (e.g. Protozoa, Chaetognatha, and Brachyura zoea), in which case the size data of randomly selected 20 individuals were used for calculation. The “glmmTMB version‎ ‎1.1.2.3” package (Brooks et al. 2017) was used for GLMM analysis. The significance level of all the GLM and GLMM analyses was set at α = 0.05. Sample sizes were 30 (chub mackerel) and 59 (Japanese anchovy) for the GLM analyses of ingested prey number, and 11 (chub mackerel) and 6 (Japanese anchovy) for the GLMM analyses of the relationship between prey size (length and width) and SL. The GLM analyses were based on the Poisson distribution because the response variable was discrete, whereas the GLMM analyses were based on the Gamma distribution because the response variable was continuous and contained only positive values.

Results

Taxonomic composition of source plankton assemblages

All of the source plankton assemblages used in the experiment met the criterion that they should include not only crustaceans but also other taxa without shells. The most abundant species in the field-caught plankton assemblages was Noctiluca scintillans (shown as Protozoa in Figs. 4 and 5, all of which was N. scintillans) throughout all the trials, occupying between 35 and 78% of the source zooplankton. These assemblages also contained various crustaceans such as copepodites and adults of copepods (hereafter referred to as copepod(s)), larval decapods, Cladocerans, Gammaridea, Mysidacea, and Stomatopoda as well as organisms that do not have external skeleton such as fish larvae, fish eggs, Chaetognatha, Mollusca, and Protozoa, (see Figs. 4 and 5 for details). The number of individuals of Calanoida species identified in the unused source zooplankton is shown in Tables S1 (for juvenile chub mackerel) and S2 (for larval/juvenile Japanese anchovy). Note that none of the copepod nauplii was included in any of the plankton sources.

Fig. 2
figure 2

Relationships between prey number in the gut and standard length of juvenile chub mackerel and larval/juvenile Japanese anchovy. Solid lines indicate the generalized linear models (GLMs) fitted to the data. Gray shadings on either side of the lines indicate the 95% confidence interval. Both models were based on the Poisson distribution

Full size image
Fig. 3
figure 3

Relationships between prey size (length and width) in the gut and standard length of juvenile chub mackerel and juvenile Japanese anchovy. Solid and dotted lines indicate the generalized linear mixed models (GLMMs) fitted to the data, representing significant and non-significant relationships, respectively. Gray shadings on either side of the lines indicate the 95% confidence interval. Vertical lines on the plots indicate standard error. All models were based on the Gamma distribution

Full size image

Prey number and size

For chub mackerel juveniles, 18 individuals ranging from 16.25 to 35.93 mm SL contained prey whereas 12 ranging from 15.83 to 36.61 mm had empty guts. For Japanese anchovy larvae, 29 individuals ranging from 11.85 to 32.21 mm SL contained prey whereas 30 ranging from 15.04 to 31.16 mm had empty guts. The number of prey in the gut increased with body size in both species (P < 0.01; Table 1; Fig. 2).

Table 1 Summary of modeling results of generalized linear models (GLM), generalized linear mixed models (GLMM), and the Wald test to examine effects of fish size (standard length: SL) on the number (tested by GLMs based on the Poisson distribution) and size (tested by GLMMs based on the Gamma distribution; length or width) of prey items in the gut
Full size table

In the case of chub mackerel juveniles, prey length increased with body size (P = 0.03) whereas the relationship between prey width and body size was not significant (Table 1; Fig. 3, details of the random effect are shown in Table S3). For anchovy larvae, both characteristics of prey increased with body size. Prey width, not prey length, showed stronger correlation with body size of anchovy larvae (P < 0.01 and P = 0.04, respectively; Table 1; Fig. 3, details of the random effect are shown in Table S3), showing a contrasting trend with chub mackerel juveniles. All the individuals in the beaker containing brine shrimp ingested this prey, suggesting that the poor ingestion rate of small individuals of either species is not attributed to the experimental conditions.

Diet composition and selectivity

Juvenile chub mackerel

Copepodites and decapods were the main prey items ingested by chub mackerel juveniles 15.83 to 36.61 mm SL in all trials, contributing between 54 and 80% to the diet by number (Table 2).

Table 2 Diet composition of hatchery-reared chub mackerel juveniles in three trials. The number of ingested individuals and its ratio to the total number of ingested prey in each of the three trials are listed for each prey organism. Size ranges of the individuals tested were 15.83–24.57 (first trial), 16.25–25.07 (second trial), and 26.09–36.61 mm SL (third trial)
Full size table

Copepodites consisted of calanoid and poecilostomatoid species, both of which were ingested with positive selectivity except in the third trial (Fig. 4). Poecilostomatoids ingested in the first and second trials were both Corycaeus affinis, which were rarely distributed in the source water. This led to the very high selectivity for this prey item. Four of the ingested Calanoida species were identified: Paracalanus parvus s.l., Pseudodiaptomus marinus, Centropages abdominalis, and Labidocera rotunda. All of them were rare with a count of up to two individuals per trial, and none of these species dominated in the gut of chub mackerel juveniles. Decapods were ingested in the second and third trials, most of which were zoeal Brachyura and others were megalopa Brachyura and zoeal Anomura, both with a count of only one individual per trial. Even though many Brachyura zoea were found as gut contents especially in the third trial, the selectivity for them was moderate because of their abundance in the source water. Contrastingly, megalopa Brachyura and zoeal Anomura were both ingested with high selectivity due to their scarcity in the source water (circular plots in Fig. 4c). Despite the short experimental period of 15 min, some of the ingested prey were digested to such an extent that accurate identification was impossible (Table 2). Other prey items observed include Cirripedia nauplius and cypris, fish egg, Onychopoda (all of which was Evadne tergestina), and Gammaridea, all with positive selectivity.

Fig. 4
figure 4

Length (a), frequency in the source assemblages (b), and Chesson’s α index (c) of each prey item used in the feeding experiment on the juveniles of chub mackerel. Prey length data were pooled from the three trials. Vertical lines on the plots indicate standard error. Gray square, triangle, and circle plots indicate the first, second, and third trials, respectively. Horizontal dashed lines indicate (a) the largest and smallest lengths of the ingested prey and (c) the critical value for selectivity. Brachyura Zoea and Protozoa are shown in the separate panel for better visibility of the frequency (panel b) of the other prey items. Zero values are not plotted

Full size image

Juvenile Japanese anchovy

As with the case of juvenile chub mackerel, copepodites and decapods were the main prey items ingested by juveniles of Japanese anchovy 22.9 to 32.2 mm SL in the second trial, contributing 68% to the diet by number (Table 3). Copepodites consisted of calanoid and poecilostomatoid species, both ingested with positive selectivity (Fig. 5). Poecilostomatoida ingested was Corycaeus affinis as with the case of chub mackerel. Two of the ingested Calanoida species were identified: Temora turbinata and Labidocera rotunda accounting for 15% and 8% of the diet by number, respectively (Table 3). Most of the ingested decapods were zoeal Anomura and others were zoeal Brachyura and zoeal Caridea, both with a count of only one individual per trial. Zoea of Anomura and Caridea were positively selected whereas zoeal Brachyura were negatively selected (Fig. 5). As with the case of juvenile chub mackerel, unidentified crustacean and other remains accounted for a substantial fraction of the ingested prey (26% in total; Table 3). Other prey items observed include three individuals of Hyperiidea (Table 3), the selectivity for which was negative owing to its very high abundance in the source water (Fig. 5).

Table 3 Diet composition of hatchery-reared Japanese anchovy juveniles in the second trial. The number of ingested individuals and its ratio to the total number of ingested prey are listed for each prey organism. Size range of the individuals tested was 22.9–32.2 mm SL. Data of the first and third trials were excluded because most of the individuals had empty guts
Full size table
Fig. 5
figure 5

Length (a), frequency in the source assemblages (b), and Chesson’s α index (c) of each prey item used in the feeding experiment on the juvenile Japanese anchovy. Only data from the second trial were used. Horizontal dashed lines indicate (a) the largest and smallest lengths of the ingested prey and (c) the critical value for selectivity. Hyperiidea and Protozoa are shown in the separate panel for better visibility of the frequency (panel b) of the other prey items. Zero values are not plotted

Full size image

Discussion

The main value of this study lies in examining the feeding ecology of larval and juvenile pelagic fishes, which are quite difficult to artificially hatch and rear, under experimental conditions that would minimize the unavoidable uncertainties associated with field net sampling. The obtained results, indicating that prey number and size increased with fish size and the fish demonstrated strong selectivity for crustaceans including copepodites, complement the findings of previous field studies.

Despite careful monitoring, neither apparent regurgitation nor excretion of ingested prey was observed in our samples fixed under anesthesia. Nevertheless, prey number and size (length and/or width) increased with fish size in both juvenile chub mackerel and larval/juvenile Japanese anchovy as reported in previous field studies of pelagic fish larvae and juveniles (Cass-Calay 2003; Morote et al. 2010; Intxausti et al. 2017). Empty guts were frequently seen in small individuals, which is in accord with previous field studies of larval/juvenile chub mackerel (Taga et al. 2019) and Japanese anchovy (Islam and Tanaka 2009). The increasing trend in the ratio of individuals with food in their guts was clear in larval/juvenile Japanese anchovy and moderate in chub mackerel, which coincides with previous field studies of each species (Uotani 1985; Islam and Tanaka 2009; Taga et al. 2019). Empty guts in small individuals are unlikely to be attributed to the experimental conditions, because all fish ate brine shrimp under the same experimental setup. The fact that trends similar to those seen in field-collected juvenile or larval pelagic fishes were observed in this laboratory study suggests that the qualitative trends observed in previous field studies represent the true situation of live fish in the sea at least within the size range tested in this study.

The measure of prey size (length or width) that showed a stronger correlation with fish size differed between two species: prey length for juvenile chub mackerel and prey width for larval/juvenile Japanese anchovy. Not surprisingly, however, prey width for juvenile chub mackerel, a prey-size criterion that did not show a significant effect on fish size, still showed a tendency of positive correlation. Therefore, there remains a possibility that larger sample size may lead to the detection of a significant correlation in prey width for chub mackerel. In fact, Hunter and Kimbrell (1980), who investigated a larger number of wild chub mackerel larvae (up to c. 17 mm; N = 86), reported a significant positive relationship between fish size and prey width, which was the non-significant criterion in our study. On the other hand, both measures of prey size showed a significant correlation to the fish size of Japanese anchovy larvae, with prey width showing a stronger correlation than prey length. It is unclear whether the contrariety of strongly associated factors between juvenile chub mackerel and larval/juvenile anchovy (prey length vs. prey width) is due to the species or life-stage difference. Nevertheless, these results shed light on the importance of measuring multiple characteristics of prey when examining the relationship between prey size and predator size, as suggested by previous studies (Fahy 1980; Dodrill et al. 2021).

Juvenile chub mackerel exhibited strong selectivity for Corycaeus affinis (shown as Poecilostomatoida in Fig. 4), Gammaridea, and Cirripedia cypris (α > 0.34). They also positively selected zoea of Anomura and Brachyura, Megalopa of Brachyura, Calanoida species, fish egg, and Onychopoda. In contrast, Noctiluca scintillans (shown as Protozoa in Fig. 4), the most abundant species in the source water, were not ingested at all even though their size range was equivalent to that of other preferred taxa, suggesting that chub mackerel juveniles had strong negative selectivity for this species. Noctiluca scintillans and fish eggs have subglobular forms of similar size, and do not actively move. The clearly contrasting selectivities on these two prey items suggest that chub mackerel juveniles do not select prey based merely on its size and mobility. Annelida, Chaetognatha, Cumacea, fish larva, and Mysidacea were not ingested either, presumably because they were too big to eat. In addition, juvenile chub mackerel did not eat adult and zoeal Caridea, Cnidaria, Harpacticoida, Isopoda, Mollusca, and Stomatopoda even though they were within the ingestible size range. Juvenile Japanese anchovy showed positive selectivity for Calanoida species, Corycaeus affinis (shown as Poecilostomatoida in Fig. 5), and zoea of Caridea and Anomura. Corycaeus affinis has been reported as an important prey for adults of Japanese anchovy (Yoneda et al. 2022), showing the particular importance of this prey species for both larvae and adults of Japanese anchovy. Interestingly, zoea of Brachyura were not selected even though they were the most abundant zoea species in the source water, showing a clear difference from juvenile chub mackerel. As in the case of chub mackerel juveniles, Noctiluca scintillans (shown as Protozoa in Fig. 5), the most abundant species, were not selected at all. Chaetognatha, Cnidaria, Cumacea, Gammaridea, Hyperiidea, Isopoda, Luciferidae, Mysidacea, Ostracoda, Polychaeta, and Hyperiidea were not selected quite likely due to their large size. Adults of Caridea were not selected even though they were within the ingestible size range. Our data clearly show that juveniles of Japanese anchovy and chub mackerel selectively eat crustaceans, as suggested from the previous field studies. The crustaceans ingested by the two species consisted of not only copepodites but also larvae of other crustaceans such as zoeal decapods, indicating that zoea of Brachyura or Anomura can be preferred prey if dominantly distributed in environmental water, as suggested by Yamamoto and Katayama (2012). Some field studies mentioned appendicularians, in addition to copepodites, as important prey for larvae and juveniles of chub mackerel (Sassa et al. 2008; Kume et al. 2021) and Japanese anchovy (Uotani 1985). However, since appendicularians were not collected, selectivity for this group could not be evaluated in this study.

The fact that laboratory-reared fishes that were fed only with rotifers and brine shrimps showed clear selectivity on wild zooplankton indicates that intrinsic factors may play an important role in the feeding selectivity of larval and juvenile fishes as suggested by Peck et al. (2012). A novel finding provided by our data is that both species can selectively prey on preferred foods that are rare while avoiding non-preferred foods that are abundant. This implies that juveniles of Japanese anchovy and chub mackerel are capable of distinguishing between apparently similar types of prey to find the target prey. This kind of selective ability in larval and juvenile fishes is thought to be supported by their sense of perceiving the morphology and behavior of prey in addition to the size (Nunn et al. 2012; Okazaki et al. 2019b). Our study has opened new possibilities for further laboratory studies on detailed mechanisms of the feeding selectivity of larval and juvenile pelagic fishes. For example, recording the feeding behavior of these fishes and analyzing their responses to the prey and the behavior of the prey will help to understand how and why certain types of prey are selected or not. Given the difficulty of predicting preferred prey for fishes in early life stages based only on theory, such as optimal foraging theory (Gerking 1994), this sort of approach will be beneficial for understanding the feeding ecology of larval and juvenile pelagic fishes.

There is a concern that global climate change will induce a geographic or seasonal match/mismatch between prey and predators (Cooley et al. 2022). The same type of empirical study as shown in the present paper will help the development of models that can project the growth and survival of commercially important fish species under such future climate conditions. Inadequate development of zooplankton modeling has long been one of the bottlenecks for future projections of fish production (Ito et al. 2010). However, its accuracy has recently improved owing to the development of plankton community models that allows various ecological and physiological traits of prey plankton, such as size and biogeochemical functions, to be introduced into global models (Ward et al. 2014; Negrete-García 2022). In contrast to the improved performance in the simulation of plankton community fluctuations in a dynamic environment, performance in the simulation of prey selectivity for organisms at higher trophic levels, such as fish, is on a plateau due to the lack of knowledge about the real prey selectivity of fish and the mechanisms which determine the prey selectivity. A series of experiments proposed by this study would be also valuable in the field of global modeling, because empirical approaches can help to reflect the realistic prey selectivity of the target fish species in the above-mentioned multi-trait prey plankton models and promote the development of models projecting ecosystem responses to global climate change.

The preferred size of prey under experimental conditions is known to change over several tens of minutes (Gerking 1994). Confer and O’Bryan (1989) demonstrated a phenomenon called “first feeding burst” through their feeding experiment on larval perch and trout: they preferred large prey species shortly after the start of the experiment, but this preference transitioned within about 10 min to smaller species. The results of our trials, which were conducted for 15–30 min, might have been significantly affected by the first feeding burst. Further experiments over an extended period of time will help us understand the characteristics of feeding tactics over a longer term. The fact that even crustaceans were digested to an unidentifiable condition within short periods of time under 22 °C would be informative for field scientists from the viewpoint of providing solid bases for making a convincing interpretation of the situation of gut contents of their samples. It is also important to note that copepod nauplius, an organism often reported as the main prey of small larvae of pelagic fish (e.g. Dickmann et al. 2007; Intxausti et al. 2017; Taga et al. 2019), were not collected with our method. The lack of nauplius does not seem to have a substantial effect on the result of chub mackerel because larvae of this species are known to change their main prey from nauplius to copepodites when they reach 9–11 mm SL (Taga et al. 2019), and all the juveniles used in our study were bigger (< c. 15 mm SL). However, in the case of Japanese anchovy, this may be one factor that led to the high ratio of empty guts in small individuals, given a field study that reported that small larvae of this species feed on copepod nauplius (Yasue et al. 2011). A weak point of this study is the small number of replicates and specimens included in the analyses, which might have affected the results. For example, as mentioned above, a measure of prey size that was not significant might turn out to be significant if validated with a larger sample size. However, the results of this study were considered reasonable in view of both theoretical and field studies. Moreover, the differences in results among the trials of chub mackerel were very small (Table S3). For these reasons, we assume that increasing the number of samples would not lead to a critically different conclusion.

Conclusions

This study demonstrated what type and size of zooplankton would or would not be selected as prey organisms by larval/juvenile Japanese anchovy and juvenile chub mackerel under experimental conditions minimizing the uncertainties inseparable from field net sampling. The obtained results—prey number and size increased with body size and the fish showed strong selectivity for crustaceans including copepodites and decapods—substantially support the conclusions of previous field studies. Nevertheless, results may vary depending on the developmental stage and/or species of target fish. In addition, different species compositions of source plankton assemblages might lead to significantly different results. Further experimental studies of this sort, which would complement the results obtained from prevailing field studies, are essential for a precise understanding of the feeding ecology of small pelagic fishes.