The Effects of Growth, Predation, and First-Winter Mortality on Recruitment of Bluegill Cohorts

Introduction

Much research has focused on understanding recruitment processes in freshwater and marine fishes, and mechanisms affecting survival of the early life history stages have been examined extensively (see Miller et al. 1988). The principal factors thought to regulate recruitment include such environmental conditions as temperature, physical habitat, and weather (Kramer and Smith 1962; Summerfelt 1975; Aggus 1979); food availability and competition for food (Prout et al. 1990; Welker et al. 1994; Ludsin and DeVries 1997); and predation (Forney 1971; Houde 1987; Rice et al. 1987b). In many cases, one or more critical periods of high mortality may act to substantially reduce recruitment in fishes (May 1974). The time from egg fertilization to the onset of first feeding has been suggested as critical in many species (Kramer and Smith 1962; Toetz 1966; Forney 1971; Rice et al. 1987b). Recruitment may be further influenced by high mortality during the first winter of life (Oliver et al. 1979; Adams et al. 1982; Post and Evans 1989; Miranda and Hubbard 1994).

First-winter mortality is often size dependent, whereby smaller fish generally experience higher mortality than larger individuals in a cohort (Oliver et al. 1979; Toneys and Coble 1979; Post and Evans 1989; Miranda and Hubbard 1994). Work with centrarchid basses suggests starvation due to depletion of energy reserves over the winter nonfeeding period is the physiological mechanism causing higher losses of smaller individuals (Oliver et al. 1979; Adams et al. 1982; Miranda and Hubbard 1994). Depletion of energy reserves is greater in small bass because they store proportionately fewer lipids and utilize this energy source more rapidly than larger individuals of the same cohort (Oliver et al. 1979; Adams et al. 1982). In addition to directly causing fish losses, starvation or poor nutrition can indirectly lead to reduced recruitment by influencing risk of predation (Miller et al. 1988; Jonas and Wahl 1998). Increased predation risk may result from either reduced swimming capabilities (Rice et al. 1987a) or reduced growth rates (Werner and Gilliam 1984; Post and Prankevicius 1987).

For populations experiencing size-selective winter mortality, the size distribution at the end of the first growing season can be an important determinant regulating survival to the following spring (Gutreuter and Anderson 1985). Length differences among age-0 largemouth bass Micropterus salmoides have been attributed to extended hatching periods, which can lead to differences in prey availability and growth within a cohort (Miller and Storck 1984; Adams and DeAngelis 1987; Ludsin and DeVries 1997). Bass hatched earlier in the spawning season are better able to utilize fish prey, grow faster, are larger in fall, and have higher survival overwinter than fish hatching later in the growing season. A similar pattern of higher survival among fish hatched early in the spawning season has been demonstrated for bluegills Lepomis macrochirus in piscivore-free waters (Beard 1982) and at northern latitudes where protracted spawning was precluded by a short growing season (Beard 1982; Cargnelli and Gross 1996). In contrast, Garvey et al. (2002) found higher survival of later hatches of sunfish (mixed bluegill and pumpkinseeds L. gibbosus) in a Canadian lake and suggested that predation may have influenced survival of earlier hatches. Given the potential importance of predation and time of spawning on fish recruitment, additional information is needed regarding the recruitment success of bluegills (1) in systems containing fish predators and (2) at lower latitudes where spawning is protracted.

We investigated the mechanisms influencing bluegill recruitment during multiple years in a lake having an established fish community and a protracted bluegill spawning period. First, we used otoliths to determine ages and spawning dates of bluegills and compared spawning distributions of recently hatched larvae with distributions of juveniles surviving to fall. This otolith-based analysis allowed us to identify differences in growth and survival that varied with spawning date. We then assessed the influence of predation on survival by combining extensive field sampling and a bioenergetics model to estimate the numbers of age-0 bluegills consumed by largemouth bass. Finally, we evaluated first-winter mortality by comparing fall and spring estimates of bluegill abundance and size structure in shoreline rotenone samples. Using this combination of approaches, we were able to empirically assess the relative importance of spawn date, growth, predation, and overwinter mortality on year-class strength and size structure of bluegills.

Methods

Larval bluegill demographics

Larval bluegills were sampled at weekly intervals at night from April through August 1987 through 1989. Two Miller high-speed samplers (1.6 m long × 14 cm in diameter, 0.5-mm mesh) were towed from the bow of a boat at a constant speed of about 2 m/s; a calibrated flowmeter was used to estimate the volume of water filtered. Oblique tows sampled from the surface to a depth of 4 m (the maximum depth that we could safely tow without hitting lake bottom) along the central axis of the lake. All larvae were preserved in 70% ethanol, identified to species, counted, and measured for total length (TL; mm). Total densities of bluegill larvae were expressed as the number of fish per cubic meter of water filtered and averaged across tows (N = 2 per sample date). Larval densities also were calculated for each 1-mm length interval.

We estimated the ages of bluegill larvae from daily increments on otoliths to determine larval spawn dates, growth, and mortality. Sagittal otoliths were removed from a subsample of larvae (N > 290/year; minimum of 3 per 1-mm length interval per sample date) and mounted proximal side down with thermoplastic cement on glass microscope slides. Larger otoliths were wet-ground on 600-grit emery cloth to expose daily rings (see Taubert and Coble 1977); smaller otoliths required no further preparation before aging. All otoliths were viewed by two readers with a microscope at a magnification of 400× or greater. Before aging wild fish, readers aged and re-aged a set of otoliths from known-age bluegill (2–71 d old) until there was no difference between estimated and known ages of fish. Daily rings on wild bluegill otoliths were counted three separate times, and those with counts varying by 10% or less were averaged to estimate age in days. Both readers reexamined otoliths with counts that did not agree within 10% until a consensus was reached; if consensus was not reached, the otolith was eliminated from the data set. Ages for all larvae in a sample were assigned based on the average age of larvae from the matching length interval in the aged subsample. Because estimated ages represent time from swim-up (Taubert and Coble 1977), we added 7 d to each otolith-derived age to determine the date that larvae were spawned (Morgan 1951; Childers 1967; Bain and Helfrich 1983). Spawning date distributions for newly hatched bluegills (5 and 6 mm TL) were calculated with larval density data summed over 7-d spawning intervals that corresponded to larval sampling dates.

Growth and mortality rates for larval bluegills were determined with otolith-based methods that modeled growth and mortality with exponential equations (Essig and Cole 1986; Zigler and Jennings 1993). We used 5–11-mm larvae for these estimates because catches of smaller and larger larvae were low in ichthyoplankton tows. Young bluegills in Ridge Lake began to transition from the limnetic to the littoral zone at about 12 mm TL, based on reduced catches in ichthyoplankton tows and the presence of 12-mm bluegills inshore during seining and rotenone sampling. Instantaneous daily growth was estimated monthly with the equation

where L is the total length (mm) of each larva, a is the length axis intercept, G is the coefficient of instantaneous growth, and t is the age of each larva in days. We chose to estimate growth with otoliths because this method has been shown to be more precise for larval fish than length-based methods (Zigler and Jennings 1993). Mortality was estimated for bluegills spawned early (May and June) and late (July and August) in the year because sample size was insufficient for monthly estimates. A single otolith-increment distribution was generated for each estimation period by summing larval densities by increment group from all samples within the period. Instantaneous daily mortality was estimated as the negative slope of the descending limb of the log_e transformed increment frequency curve with the equation

where N_t is the number of larvae at age t, N₀ is the number of larvae in the first fully recruited increment group (abundance axis intercept), Z is the coefficient of instantaneous mortality, and t is age in days. Confidence intervals of 95% for growth and mortality estimates were calculated by standard regression techniques (Zar 1974).

Juvenile bluegill demographics

We estimated the size structure and abundance of juvenile bluegills in September of each year with three shoreline rotenone samples spaced about equally around the lake. Bluegill yearlings were sampled similarly with rotenone during May 1988 through 1990 to assess overwinter mortality. A 4-mm-mesh block net (30.5 m long × 1.8 m deep) was deployed to enclose a rectangular area 53–96 m² and rotenone was applied to the enclosed area at a 2–3 mg/L concentration (Timmons et al. 1979). Dying fish were recovered immediately as they surfaced after which we removed macrophytes from the enclosure, checked them for dead fish, and pulled the block net to shore in seine-like fashion. All recovered bluegills were counted, and a subsample was measured for total length (mm; N = 200). Densities were calculated (number of juveniles/m²) and averaged across rotenone samples (N = 3/year). Densities also were calculated for each 5-mm length interval.

To examine the effects of spawn date on bluegill recruitment, we removed otoliths from 70 to 90 haphazardly selected (presumably random) bluegills from fall rotenone samples (minimum N = 10 fish/5-mm length interval). Age (d) and spawn date were determined for each fish following procedures described above for larval bluegills. Likewise, spawn dates were assigned to all bluegills in a sample based on the average age of fish from the matching length interval in the aged subsample. Spawn date distributions for juveniles were calculated with density data summed over 7-d spawning intervals that corresponded to larval sampling dates. Frequency distributions of juveniles and larvae were compared to assess their relative survival over the summer.

We evaluated first-winter mortality by comparing the densities of age-0 bluegills from fall samples to those of yearlings sampled in spring. The occurrence of size-dependent mortality was assessed by examining fall and spring length-frequency distributions and a quantitative length-increment technique that identifies the influence of size-selective mortality and growth on sequential size distributions (Post and Evans 1989). We first expressed length distributions as percentiles and then calculated the differences in length increments between percentiles for consecutive fall and spring samples. We next plotted length increment against total length of the fall sample and used the resulting linear relation to diagnose relative changes in growth and mortality between small and large individuals (Post and Evans 1989). We did not attempt to identify spawn dates for bluegills sampled during spring because daily rings on otoliths become indiscernible during the winter slow-growth period and the precise date of annual growth cessation was not known.

Bluegill predation losses

To determine the losses of each weekly bluegill cohort to predation, we used a bioenergetics model for largemouth bass (Rice et al. 1983) to calculate food consumption from observed growth and water temperatures. Monthly estimates of age-0 bluegill consumption from the model were used to adjust spawn date distributions of bluegills consumed by largemouth bass to assess the relative effect of bass predation on survival of bluegill cohorts. Water temperatures as well as largemouth bass growth and food habits data were collected as input parameters for the bioenergetic model. Water temperature was recorded daily with a maximum–minimum thermometer placed 1 m below the lake surface. Maximum and minimum values were averaged to provide the mean daily temperature used in the model. We assumed that largemouth bass did not thermoregulate because they were almost always captured in littoral areas less than 1 m deep.

From April to October of each year, age-1 and older largemouth bass were sampled monthly for diet and size structure via daytime angling and day and night shoreline electrofishing (3,000-W AC, 230 V, three-phase). We were able to use angler-supplied fish to supplement electrofishing samples because a complete creel census was conducted at the lake and estimates of size structure and diet composition are generally similar for largemouth bass sampled by electrofishing and angling (Santucci and Wahl 1991). Age-0 bass were sampled every other week by seining (July–August) and electrofishing and shoreline rotenone sampling (September–October). Stomach contents were removed with clear acrylic tubes, except for fish less than 75 mm TL, which were sacrificed for diet analysis. Standard and backbone lengths of bluegills in stomachs were converted to total lengths (as per Santucci and Wahl 1993) and age (d) was determined from length-age regressions (r > 0.97, P < 0.001) developed each year using combined data from larval and juvenile bluegill samples. Spawn dates for age-0 bluegills recovered from largemouth bass stomachs were determined, and each fish was assigned to a 7-d spawning interval corresponding to the larval bluegill sampling dates.

We estimated the population size of largemouth bass during September and October of each year. Age-1 and older fish were captured for the marking census by electrofishing and recaptured within 1 month by electrofishing and angling. All fish were marked with a fin clip (upper caudal) and measured for total length (mm) and weight (g). Population size was estimated with the Chapman modification of the Petersen formula (Ricker 1975). Abundance of age-0 largemouth bass was estimated from fall shoreline rotenone samples (per Timmons et al. 1979). To estimate largemouth bass ages and monthly growth, we removed scales from all individuals captured by fall electrofishing and scales and otoliths from a subsample of 10 fish/25-mm length interval taken from rotenone and electrofishing recapture samples. Largemouth bass captured during fall were aged with scales and otoliths; whereas those captured earlier in the year were assigned ages by following the temporal progression of modes of monthly length- and weight-frequency histograms.

Population size for largemouth bass estimated during fall, assumed to approximate the mean population size during June through September, was used to compute total population consumption from model-generated estimates of mean consumption per individual within each age-class. Largemouth bass older than age 3 did not consume age-0 bluegills, so we modeled food consumption for bass ages 0–3. The proportion of food consumption consisting of age-0 bluegills was determined from volumetric diet composition. Because the food consumption rate estimated as a proportion of maximum ration for an entire season is not appropriate (Rice and Cochran 1984), we estimated consumption for largemouth bass over 30-d intervals from June through September. Monthly estimates from the model of the total weight of age-0 bluegills consumed were used to adjust bluegill spawning date distributions developed from monthly largemouth bass diet samples. Adjusted distributions were then summed across months to estimate total largemouth bass consumption of each 7-d bluegill cohort.

Whole-lake extrapolations

To examine the relative importance of predation by largemouth bass on the mortality of various early life stages of bluegills, we estimated the whole-lake abundance of limnetic bluegill larvae and littoral-zone juveniles surviving to fall and spring. The limnetic region of the lake was defined as the water volume from lake surface to the thermocline for depths greater than 2 m. The littoral region was defined as the area of the lake between the 2-m depth contour and the lake shoreline. Larval estimates included 5- and 6-mm newly hatched larvae and 10- and 11-mm larvae (i.e., those not yet migrated to nearshore areas). Ichthyoplankton densities (number/m³) from weekly high-speed tows were multiplied by the volume of the limnetic region and summed across weeks to determine annual abundance. Juvenile densities (number/m²) from fall and spring rotenone samples were multiplied by the total area of the littoral region. These values were compared with bluegill losses to largemouth bass predation determined from bioenergetic model estimates.

Statistical analysis

Annual spawn-frequency distributions of juvenile bluegills surviving to fall were compared with distributions of larval bluegills via the Kolmogorov–Smirnoff test. Mortality of larval bluegills was compared between fish hatched early (May–June) and late (July–August) in the season via analysis of covariance (ANCOVA). Patterns in larval growth rates over time were assessed with linear regression procedures. Pearson correlation analysis was used to examine the relationship between larval bluegill growth, density, and water temperature. We compared densities of juvenile bluegills between fall and spring samples with one-way analysis of variance (ANOVA) for a randomized complete block design (blocked by year). Fall and spring length-frequency distributions were compared each year with the Kolmogorov-Smirnoff test. Statistical significance for all tests was set at α = 0.05.

Results

Effects of Spawn Date on Bluegill Growth and Survival

Bluegills had a protracted spawning period that lasted from 87 to 108 d and produced several peaks in larval abundance each year (Figure 1). Spawning was initiated during May, soon after rising water temperatures exceeded 20°C, and continued into August. In each year, peaks in abundance of newly hatched larvae occurred during May, June, and July, and there was a general decline in larval abundance throughout the summer. The initial spawning effort in May generated a mean across years of 146 larvae/m³, compared with 81 and 43 larvae/m³ for the two subsequent months. Peak larval bluegill densities across years ranged from 11 to 275 fish/m³ during May, 7–89 fish/m³ during June, 5–37 fish/m³ during July, and 1–10 fish/m³ during August.

For juvenile bluegills sampled in fall, the range of dates on which they were spawned was similar to that observed for 5- and 6-mm larvae (Figure 1). However, spawning-date distributions of juveniles differed from those of the newly hatched larvae in each of the 3 years examined (Kolmogorov–Smirnoff tests, P < 0.025). The period of greatest larval abundance occurred early in the spawning season in all years, whereas bluegills from later spawns showed the highest contribution to fall survivors. In 2 of 3 years, the earliest spawned bluegills contributed almost no fish to fall recruitment (Figure 1A, C). The lack of recruitment of early spawned bluegills was not explained by our estimates of mortality of limnetic larvae 5–11 mm TL. Instantaneous daily mortality rates ranged from 0.166 to 0.303 across years and were similar each year for fish spawned early and late in the season (ANCOVA, heterogeneity of slope, P > 0.13; Table 1).

Table 1. Estimates of instantaneous daily mortality for larval bluegills 5–11 mm total length in Ridge Lake, Illinois, 1987–1989. Sample groups are larval bluegills spawned in May and June (early) or July and August (late). Instantaneous daily mortality was estimated by the otolith-increment method (Essig and Cole 1986) and is the slope of the linear regression describing the relation of larval age and relative abundance

Average absolute growth (±95% confidence interval [CI]) of 5–11-mm bluegills across all months and years was 0.55 ± 0.14 mm/d. Estimates of instantaneous growth (G) were lowest for bluegills spawned in May and highest for fish spawned later in the year (Figure 2). A strong positive correlation existed between instantaneous growth rate and month spawned for all years combined (r = 0.79, P = 0.002). Larval bluegill growth also was positively correlated with temperature (Pearson correlation; r = 0.64, P = 0.03) but was not related significantly to larval fish density (r = −0.12, P = 0.72).

Effects of Predation on Recruitment to Fall

Primary food items in largemouth bass stomachs were age-0 and older bluegills, crayfish, and aquatic insects. Diet composition data were combined and averaged across years because diets of largemouth bass less than 20 cm did not differ among years (χ² test; P > 0.40) and differences in diets of larger bass were due to declines in crayfish volumes over time (30% to 5% from 1987 to 1989; P < 0.001). Age-0 bluegills were consumed only by age-3 and younger largemouth bass (<35 cm TL) and bluegills contributed less to the diets as predator size increased (Figure 3). Age-0 bluegills made up 89% of the diet for age-0 largemouth bass, 25% for age-1 largemouth bass, and less than 9% for larger fish. In contrast, the proportion of age-1 and older bluegills and crayfish increased with increasing largemouth bass size. Other invertebrate and vertebrate fauna (terrestrial worms and insects, other fish taxa, and herpetofauna) also were eaten but individually they made up only a small proportion (<5% by volume) of the diet.

The density of largemouth bass capable of consuming age-0 bluegills was high in Ridge Lake, but density varied among years and predator age-classes (Table 2). Densities for largemouth bass age-0 through age-3 were 327 fish/ha in 1987, 597 fish/ha in 1988, and 400 fish/ha in 1989. Abundance was always highest for age-0 fish, but density did not decline consistently with age for older bass because strong year-classes remained relatively abundant through time. Growth of largemouth bass also varied among years, and age classes and, at least for age-0 fish, appeared to be inversely related to abundance (Table 2).

Table 2. Largemouth bass population data used in bioenergetics modeling and estimates of total biomass of age-0 bluegills consumed by largemouth bass in Ridge Lake, Illinois, June–September 1987–1989. The abundance of largemouth bass was estimated in the fall from shoreline rotenone surveys (age-0 fish) or Petersen mark–recapture estimates (age-1–3 fish). The variable P is the proportion of maximum consumption (C_max) that was actually consumed, as determined by bioenergetic modeling. Age-0 bluegills were not consumed by age-3 largemouth bass in 1987

Age-0 largemouth bass consumed more age-0 bluegills than any other predator age-class (Table 2). The biomass of age-0 bluegills consumed declined dramatically with predator age in 1987 and 1988 (i.e., the oldest age-group consumed less than 1.5 kg/ha), whereas in 1989 age-0, age-2, and age-3 largemouth bass consumed nearly the same biomass of bluegills, possibly reflecting a decline in the availability of larger prey items in the lake. Despite differences in the demographics of age-0 bluegill consumed among years and largemouth bass age-classes, bluegills from each spawning bout were eaten. Predation losses obtained using the bioenergetics model were consistently higher for fish spawned during the early half of the spawning season than they were for those spawned later in the year (χ² test; P = 0.001; Figure 4). Predation losses were highest for fish spawned in late May and early June in 1987 and 1988 and during June in 1989.

Changes in Population Demographics over Winter

Juvenile bluegills suffered substantial mortality between fall and spring. Mean density of age-1 bluegills in early May (3.5 fish/m²) was far lower than the density of age-0 fish sampled the previous September (22.7 fish/m²; ANOVA: F = 11.66, df = 1, P = 0.005). Losses during the overwinter period ranged from 75% to 88% across years (Table 3).

Table 3. Mean density (±1 SE) and percent mortality of age-0 bluegills from fall (mid-September) and spring (early May) shoreline rotenone samples (three per season) in Ridge Lake, Illinois, 1987–1990

Whereas overwinter mortality was high for juvenile bluegills, a comparison of fall and spring length-frequency distributions showed little size-selective mortality (Figure 5). Bluegill size structure was remarkably similar between fall and spring samples in all years, except that spring distributions consistently were shifted to larger sizes by about 4 mm for all size-classes. This shift was probably due to either late fall or early spring growth. Initial unadjusted distributions differed between fall and spring each year (Figure 5; Kolmogorov–Smirnoff test, P < 0.001). However, after accounting for overwinter growth by adding 4 mm to all fall size-classes, we found fall and spring distributions were similar each year (Kolmogorov–Smirnoff test, P = 0.08, 0.79, and 0.22 for 1987, 1988, and 1989 cohorts, respectively). Slope of the length-increment plot for the 1988 cohort did not differ from zero, indicating that size-selective winter mortality did not occur that year. Positive increment slopes in the other years suggest that a higher proportion of small bluegills may have been lost during winter. However, length-frequency distributions indicated that these differential losses were relatively minor and fish as small as 20 mm survived to spring during all years.

Whole-Lake Estimates of Abundance and Mortality

Between 9 and 32 million bluegill larvae hatched in Ridge Lake during the 3 years examined. Our estimates of abundance indicated high bluegill mortality for all early life stages (Table 4). Few newly hatched larvae survived to the large larval stage (mean ± SE = 8.9 ± 3.0%), whereas more large larvae survived to become fall juveniles (35.6 ± 20.5%) and more fall juveniles survived over the winter (17.9 ± 4.0%). As a percentage of all larvae produced, total losses of young bluegills were high for limnetic larvae (87–97%) and low for fall and spring juveniles (<10%; Table 4). In contrast, estimated losses due to predation by largemouth bass were low for limnetic larvae (<13%) and high for juveniles in the littoral zone through early fall (58–93%).

Table 4. Whole-lake estimates of abundance, mortality, and losses to predation by largemouth bass for various bluegill life stages in Ridge Lake, Illinois, 1987–1989. Estimates are based on limnetic lake volume for larval bluegills and on littoral lake area for juveniles. Predation losses were determined from bioenergetic model estimates of the total numbers of age-0 bluegills consumed by largemouth bass from May 1 through September 15. Small larvae are 5–6 mm and large larvae 10–11 mm total length

Discussion

We gained insight into the survival and recruitment patterns in bluegills by examining otolith microstructure and following demographics during the first growing season and winter of life. We found that high mortality during early life stages indicated that no single period was critical to bluegill recruitment success. In contrast, Beard (1982) found that fall recruitment of bluegill juveniles in piscivore-free ponds was set during the first 4 d following dispersal of fry from the nest. In lakes or ponds with piscivore populations, abundance and survival of larval bluegills is not positively related to juvenile survival (Cargnelli and Gross 1996; Partridge and DeVries 1999), suggesting that year-class strength is probably determined by events occurring after the larval stage. We found that cumulative mortality throughout early life may be a more important determinant of bluegill year-class strength than any single, critical event. At latitudes supporting extended spawning, bluegills produce large numbers of young that face numerous sources of mortality during the summer growing season in both limnetic and littoral habitats. Winter provides another period of high mortality that can significantly reduce cohort strength without substantially altering size structure. Our study provides the first direct evidence that predation is a major factor regulating recruitment of age-0 bluegills and that predators affect bluegill survival at several early life stages. The high relative importance of age-1 and older bluegills in largemouth bass diets suggests that these piscivores may continue to influence bluegill recruitment for some time after the first year.

Larval fishes experience high mortality if prey densities are low when larvae shift from endogenous to exogenous food (May 1974). For bluegills, this early critical period occurs 4–9 d postfertilization (Toetz 1966), or about the time larvae disperse from the nest. This survival bottleneck may have gone undetected in our study because bluegill larvae were not fully recruited to our gear until they were 5–6 mm TL. If high mortality did occur during egg or swim-up stages, it did not appear to have a dramatic effect on recruitment. Densities of larval bluegills were high each year, and numerous young were recruited to fall (>39,000/ha) and spring (>10,000/ha) despite high cumulative mortality throughout early life. Bluegills also are known to consistently produce strong year-classes in other lakes (Coble 1988). As has been suggested for marine fishes (Norcross and Shaw 1984), high population fecundity and protracted spawning in bluegills may mitigate many of the abiotic and biotic mechanisms that influence recruitment of other freshwater fishes.

In fish populations where size-dependent mechanisms affect survival and recruitment, fish hatched early in the season often have a survival advantage over those hatched later. For largemouth bass, early hatches typically produce individuals that more successfully exploit available prey, grow faster, reach a larger size by fall, and experience higher overwinter survival (Adams and DeAngelis 1987; Miranda and Hubbard 1994; Phillips et al. 1995; Ludsin and DeVries 1997; but for variable overwinter survival, see Garvey et al. 1998). Evidence from northern latitudes suggests that early bluegill hatches also can have higher survival and contribute more to fall recruitment than later hatches (Beard 1982; Cargnelli and Gross 1996; but for a contrasting result in sunfish, see Garvey et al. 2002). In contrast, we found that bluegills spawned later in the season consistently had higher relative survival to fall than did early hatches. Differences among studies may have resulted from latitudinal effects such as differences in length of spawning season, length of growing season, and winter severity (Fullerton et al. 2000); variation in abiotic factors such as temperature and weather conditions during early development (Kramer and Smith 1962; Rice et al. 1987b; Mooij et al. 1994); or variation in the vulnerability of larvae and juveniles to predators (Litvak and Leggett 1992; Monteleone and Houde 1992). Protracted spawning in bluegills may be an evolutionary adaptation to variable environments that increases the chances for successful reproduction and long-term population fitness (Garvey et al. 2002). Optimal spawning times result from complex tradeoffs between environmental conditions in spring, growth, and overwinter survival that may vary with latitude. Our results emphasize the importance that predation may have on these life history tradeoffs.

Our estimates of consumption by largemouth bass suggest that predation played an important role in the differential survival between early and late spawned bluegill cohorts, predation losses being greater for cohorts hatched earlier in the year. However, the earliest bluegill hatches were underrepresented in the distribution of consumed bluegills. These early hatches could have experienced high mortality as limnetic larvae before they became vulnerable to largemouth bass predators, but this seems unlikely given that our mortality estimates for limnetic larvae did not differ between early and late cohorts in any year. In addition, the larval bluegills in Ridge Lake grew well compared with those in other waters (Breck 1993; Welker et al. 1994; Cargnelli and Gross 1996; Partridge and DeVries 1999), and zooplankton densities in the lake during spring and summer (>50 organisms/L), as determined from samples collected in 1990–1991 (Wahl et al. 1996), indicate that limnetic prey resources probably were not limited. Consumption of the early-hatched bluegills may have been low because we assumed no mortality of largemouth bass during the growing season and used fall population estimates as a measure of predator population size for all monthly consumption estimates. However, anglers in Ridge Lake harvested relatively few largemouth bass, based on creel census results, and total mortality appeared low, based on bass densities encountered in sequential years. Nevertheless, actual predator density and corresponding numbers of bluegill consumed were probably higher early in the year because some mortality would occur during the growing season.

Predation mortality in young fishes is generally thought to be inversely related to growth (Werner and Gilliam 1984; Post and Prankevicius 1987; Rice et al. 1987a). Larger and older individuals experience lower predation risk due to ontogeny in their behavior and physiology (Blaxter 1986; Webb and Weihs 1986; Miller et al. 1988; Williams et al. 1996). We observed faster growth and higher relative survival in the late bluegill hatches compared with early hatches. These differences appear due more to timing and duration of predator–prey interactions than to size-related predator avoidance by the later hatches. Vulnerability to predation has been described by a dome-shaped function, where small and large fishes of a cohort experience the lowest susceptibility (Cowan and Houde 1993; Paradis et al. 1996; Kim and DeVries 2001). High predation losses suffered by early bluegill hatches may have occurred as these fish grew to a size that increased vulnerability to largemouth bass predators (Howick and O'Brien 1983). Predation risk may have been high for early bluegill hatches because they were the first fish prey available to largemouth bass, particularly the younger ages. Furthermore, early hatches made the ontogenic shift from limnetic to littoral areas of the lake (Werner 1967) at a time when littoral-dwelling age-0 largemouth bass were abundant and large enough to consume them. Because the growing season for late bluegill hatches was relatively short, these cohorts may not have attained a size preferred by most largemouth bass predators; in addition, the early hatches, which spent a longer time in the lake with predators, experienced greater opportunities to be preyed upon. Regardless of the underlying mechanisms, our results support the suggestion by Garvey et al. (2002) that strong stage-dependent predation may alter the predictions of early life history paradigms, wherein early hatching date and faster growth result in higher survival of bluegills and other fishes (Adams et al. 1982; Post and Prankevicius 1987; Rice et al. 1987b; Miller et al. 1988; Cargnelli and Gross 1996; Ludsin and DeVries 1997; Garvey et al. 1998).

The abundance and early growth patterns of age-0 largemouth bass may influence the relative degree to which predation influences the recruitment of age-0 bluegills. We found that predation losses were high for juvenile bluegills after they migrated to the littoral zone and that age-0 largemouth bass consistently consumed more bluegills than older fish. Hence, age-0 largemouth bass had the greatest effect on bluegill survival. However, high consumption of age-0 bluegills by age-1 and older largemouth bass in some years suggested that older fish also could be an important source of mortality for young bluegills, particularly when alternative prey numbers are low. In years when largemouth bass spawning success is low or reduced growth precludes age-0 fish from achieving piscivory (Miller and Storck 1984; Adams and DeAngelis 1987), older largemouth bass may have a greater effect on recruitment of young bluegills than age-0 predators.

Predation by largemouth bass did not explain the high mortality we observed for limnetic bluegill larvae. Bluegill abundance was reduced by 87–97% between small (5–6 mm) and large (10–11 mm) larvae, whereas losses from predation by largemouth bass could account for no more than 13% of the total mortality. The low occurrence of limnetic bluegill prey in diets of largemouth bass is not surprising given the tendency for these predators to live and feed in littoral habitats. Predators that feed in open-water habitat, such as adult bluegills (Werner and Hall 1988), walleyes, and black crappies, may have influenced survival of limnetic bluegill larvae more than largemouth bass predation. Other unmeasured factors such as invertebrate predators (de Lafontaine and Leggett 1988; McGovern and Olney 1988), emigration from the lake during high-water events (Walburg 1971; Santucci and Wahl 1993), and losses related to larval condition and health (e.g., starvation, competition, and disease; Toetz 1966; Werner and Blaxter 1980; Adams et al. 1982; Adams and DeAngelis 1987; Welker et al. 1994) also may have influenced survival of limnetic larvae.

First-year overwinter survival is frequently suggested as important to the recruitment success of freshwater fishes (Oliver et al. 1979; Adams et al. 1982; Post and Evans 1989; Miranda and Hubbard 1994; Ludsin and DeVries 1997). Our results are consistent with these conclusions because we showed high losses of juvenile bluegills over the winter. The phenomenon of size-dependent overwinter mortality has received much recent attention in both experimental and descriptive studies (Oliver et al. 1979; Toneys and Coble 1979; Adams et al. 1982; Post and Evans 1989; Miranda and Hubbard 1994; Ludsin and DeVries 1997; Garvey et al. 1998; Fullerton et al. 2000). When size-selective mechanisms occur during winter, then early spawn date, rapid summer growth, and large size by fall increase winter survivorship (Post and Evans 1989; Miranda and Hubbard 1994; Cargnelli and Gross 1996; Ludsin and DeVries 1997). In these situations cohort size structure is not determined until spring because small individuals die over the winter at a higher rate than larger ones. When winter mortality is not size-dependent, as in our study, bluegill size structure will be set by fall, and high winter mortality acts only to lower cohort strength.

Energy depletion due to starvation and predation are most often implicated as the mechanisms underlying size-selective overwinter mortality in young fishes (Adams et al. 1982; Post and Evans 1989; Miranda and Hubbard 1994; Cargnelli and Gross 1996; Ludsin and DeVries 1997; Garvey et al. 1998; Jonas and Wahl 1998). All sizes of age-0 bluegills grew during winter in our study, suggesting that starvation was not a factor influencing winter survival. In contrast, predation was a likely factor reducing bluegill survival overwinter. We documented bluegill prey in stomachs of largemouth bass during the winter period (late fall and early spring) in all years. Estimates of the number of juveniles consumed by largemouth bass during late fall (16 September to 31 October) ranged from 8% to 48% of fall juveniles. During the growing season, when predators and prey interact through a series of size-related stages (Werner and Gilliam 1984), growth rates and composition of the predator population determine whether predation mortality is size-selective (Post and Prankevicius 1987). In our study the absence of size-selective overwinter mortality, when predation losses likely were high, suggests that predator–prey interactions during winter may function differently than during the growing season.

Our results and those of Toneys and Coble (1979) suggest that size-dependent overwinter mortality probably is not a common phenomenon for bluegill populations. In contrast, Cargnelli and Gross (1996) interpreted high relative abundance of larger early bluegill hatches in spring samples from Lake Opinicon, Canada, as being indicative of strong size-selective overwinter mortality of smaller, late-hatching individuals. It is unclear from this work whether size-dependent mortality occurred over the winter or during the summer and fall because juveniles were not sampled during fall. Recent work in Lake Opinicon also indicated size-dependent overwinter mortality for sunfishes (Garvey et al. 2002), but the combining of bluegill and pumpkinseed populations may have confounded this finding. Alternatively, differences among studies may indicate the potential for variability in size-dependent overwinter mortality, as shown for other species of young fishes across latitudes (Garvey et al. 1998; Fullerton et al. 2000) or systems and years within the same latitude (Post and Evans 1989). Additional research (as in Garvey et al. 1998 and Fullerton et al. 2000 for bass) examining winter behavior, size-related interactions of predators and prey, and the effects of abiotic factors on predator–prey relations (e.g., cover, temperature, day length, winter duration) is needed if we are to better understand the effects of winter on survival and recruitment in fishes such as bluegills.

Acknowledgments

T. Storck provided valuable guidance and inspiration during the early stages of this study. We thank J. Mick, D. Burkett, and G. Tickachek for coordinating activities with the Illinois Department of Natural Resources. K. Brewer, D. Coates, J. Corcoran, J. Finck, R. Mauk, M. Mounce, T. Patterson, and S. Shasteen assisted with field collections and laboratory analyses. The manuscript was greatly improved by the reviews of D. Shoup, W. Knotek, two anonymous reviewers, and the Aquatic Ecology Discussion Group at the Kaskaskia Biological Station. This study was supported in part by the Illinois Department of Natural Resources through Federal Aid in Sport Fish Restoration Project F-51-R. The Illinois Natural History Survey and Max McGraw Wildlife Foundation provided additional support.