Oecologia

, Volume 148, Issue 4, pp 573–582

Cascading life-history interactions: alternative density-dependent pathways drive recruitment dynamics in a freshwater fish

Authors

  • Rena E. Vandenbos
    • Department of Biological Sciences University of Alberta
    • School of Renewable ResourcesSelkirk College
    • Department of Biological Sciences University of Alberta
  • Shelly M. Boss
    • Department of Biological Sciences University of Alberta
Population Ecology

DOI: 10.1007/s00442-006-0410-7

Cite this article as:
Vandenbos, R.E., Tonn, W.M. & Boss, S.M. Oecologia (2006) 148: 573. doi:10.1007/s00442-006-0410-7

Abstract

Although density-dependent mechanisms in early life-history are important regulators of recruitment in many taxa, consequences of such mechanisms on other life-history stages are poorly understood. To examine interacting and cascading effects of mechanisms acting on different life-history stages, we stocked experimental ponds with fathead minnow (Pimephales promelas) at two different densities. We quantified growth and survival of the stocked fish, the eggs they produced, and the resulting offspring during their first season of life. Per-capita production and survival of eggs were inversely related to density of stocked fish; significant egg cannibalism by stocked minnows resulted in initial young-of-the-year (YOY) densities that were inversely related to adult densities. Subsequent growth and survival of YOY were then inversely related to these initial YOY densities, and survival of YOY was selective for larger fish. Because of these compensatory processes in the egg and YOY stages, treatments did not differ in YOY abundance and mean size at the end of the growing season. Because of differences in the intensity of size-selective mortality, however, variation in end-of season sizes of YOY was strongly (and inversely) related to densities of stocked fish. When mortality was severe in the egg stage (high densities of stocked fish), final YOY size distributions were more variable than when the dominant mortality was size-selective in the YOY stage (low stocked fish densities). These differences in size variation could have subsequent recruitment consequences, as overwinter survival is typically selective for YOY fish larger than a critical threshold size. Density-dependent effects on a given life stage are not independent, but will be influenced by earlier stages; alternative recruitment pathways can result when processes at earlier stages differ in magnitude or selectivity. Appreciation of these cascading effects should enhance our overall understanding of the dynamics of stage-structured populations.

Keywords

Early life-historyPopulation regulationSize-selective mortalityEgg cannibalismPimephales promelas

Introduction

Following Nicholson’s (1933) introduction of the concept and subsequent decades of debate (e.g., Andrewartha and Birch 1954; Den Boer 1968; Strong 1986), there is now general consensus that the regulation of animal populations operates through density-dependent mechanisms (Sinclair 1989; Turchin 1999; Lorenzen and Enberg 2002). Nevertheless, for most groups of animals, we still have a poor understanding of where density-dependence occurs in the life cycle, and how different density-dependent mechanisms might interact (Sinclair 1989; Hanski 1990).

For many organisms, population regulation is believed to be rooted early in life history (e.g., Miller et al. 1988; Petranka 1989). This has led in recent decades to a large increase in studies focused on these early stages (e.g., Brockelman 1969; Semlitsch and Caldwell 1982; Moksnes 2004). As a consequence of this emphasis, however, studies are often conducted in isolation of other stages (but see, e.g., Prout and McChesney 1985; Altwegg 2003).

For fishes, numerous studies have also focused on processes occurring in larval and post-larval young-of-the-year (YOY) stages (see Miller et al. 1988; Sogard 1997; Chambers and Trippel 1997). Hatch date (Cargnelli and Gross 1996), growth rate (Rice et al. 1993), and body size (Post and Prankevicius 1987) have all been shown to be important determinants of survival and cohort strength. As well, variability in growth, age, and size may also affect the intensity and selectivity of mortality in YOY fish (Crowder et al. 1992; Rice et al. 1993).

Many of these characteristics may, in turn, be affected by intra-cohort densities. Most field studies, however, can at best only estimate initial YOY densities. And, in common with those on other taxa, many studies of survival and recruitment in YOY fish largely ignore other life stages. Information on adult density is sometimes incorporated (Post and Prankevicius 1987; Tonn et al. 1994), but effects of adult density on YOY recruitment are rarely examined in detail (but see Dong and DeAngelis 1998; Post et al. 1998).

In most natural situations, life stages are not isolated from each other. For instance, YOY survival may be influenced directly and indirectly by other age-classes, via cannibalism (Polis 1981; Smith and Reay 1991), competition (Hamrin and Persson 1986; Hill 1992), and consumption of YOY predators by older conspecifics (Rettig and Mittelbach 2002). Also, although initial intra-cohort characteristics (e.g., density, timing of hatch, size-distribution) may influence which YOY survive, these characteristics can be determined by processes acting on other life-history stages (Lorenzen and Enberg 2002; Vonesh and De la Cruz 2002). Furthermore, the relative effects of one age-class on another might affect the stability of population dynamics (May et al. 1974; Cushing and Li 1992).

An early life stage that has not received much attention is the egg. Mechanisms occurring during the egg stage will affect YOY and can often be linked to characteristics of the adult population. Egg production may be affected by density-dependent growth in adults, since maturity (Andrews and Flickinger 1974) and fecundity (Wootton 1990) are often related to adult body size. Egg survival may also be related to adult density due, e.g., to a reduced ability of parents to defend eggs (Polis 1981; Smith and Reay 1991; Mappes and Kaitala 1994). However, quantifying the number and fate of eggs is often difficult, and usually only estimates are made, often rather roughly, of either egg production or survival (Fox 1994; Post et al. 1998), even though egg production and egg survival are often uncorrelated (e.g., Cole and Sadovy 1995). Until the occurrence, strength, and interactions of density-dependent mechanisms among all life-stages, including eggs, are better understood, we cannot fully understand what controls recruitment of young and the regulation of populations (Rodriguez 1989; Altwegg 2003).

We manipulated densities of fathead minnow (Pimephales promelas Rafinesque) in experimental ponds and quantified processes acting on stocked fish, the eggs that they produced, and the resulting offspring. Our objectives were to determine: (1) if YOY survival through the first growing season is influenced (directly or indirectly) by density-dependent processes acting on adults; and (2) if and how egg-stage dynamics contribute significantly to recruitment of YOY. Our ultimate goal was to assess whether an examination of density-dependent mechanisms acting and interacting during different, often under-appreciated life-stages enhances our understanding of population regulation.

Methods

Natural history

The fathead minnow is a small-bodied (ca. 5–8 cm), omnivorous fish widely distributed in North America. In the boreal region of Alberta, fathead minnows are common in small, shallow lakes and ponds lacking piscivores, and frequently occur allopatrically as the only fish species (Robinson and Tonn 1989). Two to four summers are typically required for fathead minnows to reach maturity in Alberta (Danylchuk and Tonn 2006), compared with 1–2 summers at lower latitudes (Held and Peterka 1974). Many populations are subject to periodic winterkill, but can rebound quickly (Danylchuk and Tonn 2003), suggesting that mechanisms exist that allow rapid recovery (Danylchuk and Tonn 2001). Populations in less-disturbed habitats, however, can persist at high and relatively stable densities (Danylchuk and Tonn 2003), suggesting regulation.

Males establish small (ca. 155 cm2; Flickinger 1973), aggregated nesting territories (inter-nest distances often ≤50 cm; Andrews and Flickinger 1974; Unger 1983; Jones and Paszkowski 1997) and, following spawning, guard a nest of eggs laid on the underside of plants, logs, or rocks (McMillan and Smith 1974). Fathead minnows reproduce readily in small ponds when provided with nesting substrate (e.g., floating wooden boards, Jones and Paszkowski 1997; Danylchuk and Tonn 2006); manipulating substrate availability can ensure that eggs are easy to find and enumerate. Eggs take ca. 5 days to hatch at 25°C (McMillan and Smith 1974), but require 6–12 days at ambient pond temperatures in boreal Alberta (Divino 2005; R. Vandenbos, unpublished data). Eggs have distinct developmental stages (Nagel 1976; see below), discernible to the unaided eye, which assist in tracking the fate of egg batches. Adults are fractional spawners, able to spawn more than once in a breeding season (Gale and Buynak 1982), but there is high post-reproductive mortality, especially among males (Markus 1934). Because of these life-history traits, all life-stages, including eggs, are easy to sample, observe, and use in realistic experimental settings, facilitating a study of interactions among life-stages.

Experiment

This study was conducted at the University of Alberta’s Meanook Biological Research Station (54° 37′ N, 113° 35′ W). The experiment was run for a full growing season in each of 2 years, from spring (May) to near ice-up (September–October).

Two research ponds (360 m2 surface area, 1.3 m maximum depth) were divided in half by polyvinyl curtains connected to wooden stakes that were buried into the substrate and extended well above the water level. Abundant nesting substrate (ten floating wooden boards, 0.14×1.8-m, tied to weights) was added to each pond half. The nest boards covered approximately 45% of the inshore perimeter. Prior to stocking, the ponds were raked and cleared of any other structures suitable for nesting.

We manipulated initial densities of fish by adding a 1:1:1 ratio of mature males, mature females, and juveniles, collected from nearby natural ponds, to each of the four pond halves at two density levels: 1 and 4 fish·m−2. These represent average and high densities, respectively, seen in nearby natural populations (Danylchuk and Tonn 2003). In a preliminary experiment, these densities resulted in differences in growth, but not survival, of stocked fish. There were duplicates of each treatment in each year, with each divided pond hosting both treatments.

Mature males (≥70 mm total length (TL)) and females (≥63 mm TL) were identified by secondary sexual characteristics, including tubercles on males and ovipositors on females (Danylchuk and Tonn 2001). Juveniles (two size-classes: 40–49 mm and 50–56 mm TL) were defined as smaller fish that lacked secondary sexual characteristics. Fish were marked with a subcutaneous injection of fluorescent elastomer (Northwest Marine Technology Inc., Shaw Island, WA) (Year 1) or fin clipping (Year 2) to differentiate groups throughout the experiment.

Data collection

We monitored stocked populations throughout the season to assess growth, condition, and mortality. Monthly, we set five minnow traps overnight in each pond half; setting of traps 1 day after stocking showed that this level of sampling could catch up to 90% of the stocked fish present (R. Vandenbos and W. Tonn, unpublished data). Captured fish were counted, identified by their mark, measured (TL), and returned to the ponds.

Throughout the spawning season, we examined nest boards daily to quantify the production and fate of eggs. For every egg batch, we recorded its date of appearance, mapped its location, and determined the number, stage of development, and state (healthy, diseased, depredated) of all eggs. Based on previous observations, we assigned eggs into three developmental stages (Nagel 1976). Newly laid eggs were fairly opaque, containing primarily yolk and lacking eye development in the embryo. In the second stage, eyes were visible and eggs were about half full of yolk. In the final stage, eggs were strikingly gold-tinted (from retinal development) and hatched quickly (within 24–48 h).

To quantify egg production, we identified any egg batches with eggs in the first stage of development that had appeared since the previous day. The total numbers of eggs in these batches were determined daily by counting up to 200 eggs, determining the area that this number of eggs occupied, and extrapolating to the area of the entire batch (similar to Forsgren et al. 1996). Any eggs laid in a dispersed manner were counted individually. Therefore, we were able to keep track of individual batches and could quantify the total number of eggs laid in each pond half every day.

We monitored the fate of egg batches from the time they were laid to the time they disappeared. Diseased eggs became white and opaque or showed evidence of fungal hyphae and lingered for many days before disappearing. Healthy egg batches that disappeared suddenly before the third stage of development could not have hatched and were recorded as depredated. Eggs that disappeared after reaching the third developmental stage were recorded as hatching successfully.

In Year 2, we sampled the successfully hatched YOY in five equally spaced 2.5-m horizontal net hauls, with nets (0.4 m diameter, 1 mm mesh) pulled at a speed of approximately 0.75 m/s, ca. 1 month after hatching began. This allowed us to determine mid-season size distributions of YOY fish in each pond half and to monitor their densities. All YOY caught were preserved in 95% ethanol.

In the fall of both years, ponds were drained to ca. 30 cm depth and fish were censused by removal. Adults were removed by 3 days of minnow trapping (prior to draining) and seining (following draining). YOY fish were removed with shore-to-shore seine hauls. Seining continued until two hauls in a row produced fewer than five fish. All fish removed were preserved in 95% ethanol. These fish were later counted, weighed, and measured (TL) to determine recruitment rates, final sizes, and conditions (mass/length3·100) of YOY fish. Ice-up occurred within 2 weeks of removal of fish in both years.

Analyses

Because treatments were paired within ponds, paired t tests were used for comparisons of average or total measures (survival of stocked adult males, total and per capita egg production, number and proportion of eggs surviving to hatch, YOY final abundance, survival, final size and condition, and coefficient of variation of final YOY sizes). All proportion data were arcsine square-root transformed. Mid-season length distributions of YOY, measured only in the second year, were compared between treatments within each pond separately (n=2) using Kolmogorov–Smirnov two-sample tests. In Year 2, one pond flooded in early August, shortly after the mid-season sampling, mixing fish from the two treatments. This reduced the number of pond-year replicates to three for YOY survival, final length, mass and condition, but not egg production, hatching or mid-season growth (n=4). Statistical analyses used are outlined in Sokal and Rohlf (1981).

Results

Stocked fish

Density differences of stocked fish were maintained between treatments throughout the experiment in both years. Although there were the expected declines in catch rates due to post-spawning mortality, catches were always at least three times higher in the pond halves with the high stocked fish density treatment (HDS) versus the low stocked fish density treatment (LDS) during the monthly trappings. Adult male survival at the end of the season was, in fact, higher in the HDS (17±11%, mean ± SE) than in the LDS treatment (6±6%; paired t test t=2.52, df=2, P=0.04).

There were too few mature males and females remaining at the end of the season, likely due to high post-spawning mortality, to assess adult growth. The two stocked juvenile size-classes, however, grew consistently better in the LDS treatment than in the HDS treatment (Fig. 1). As a result, final lengths of juveniles were longer in LDS pond halves than in the HDS pond halves for both large (65.2±0.5 vs. 59.4±1.0 mm, respectively) and small size classes (64.6±2.1 vs. 57.7±1.0 mm, respectively).
https://static-content.springer.com/image/art%3A10.1007%2Fs00442-006-0410-7/MediaObjects/442_2006_410_Fig1_HTML.gif
Fig. 1

Summer growth (mean total length ± SE) of a large and b small juvenile fathead minnows (Pimephales promelas) stocked into two experimental ponds in 1994 and 1995 at low (LDS) and high (HDS) densities, 1 and 4 fish m−2, respectively. Data for pond 1–95 are absent for August and September due to mid-season flooding. Initial measurements were taken on May 26–28

Egg production and hatching success

Males were observed guarding territories under nest boards within 2 days of stocking and the first egg batches appeared within 1 week in both years. Egg laying and hatching tapered off mid to late July, with the last eggs seen on 4 August. The number of egg batches present at any one time peaked during the first 2 week of the reproductive season, yet there were always empty nest sites. The maximum number of batches observed per board simultaneously was 13, but 5–6 batches on a board was more usual. Surprisingly, empty nest sites were especially common in the HDS treatment, where there were usually less than 3–4 egg batches per board, especially after the first week. Consistent with this, fewer eggs were produced in HDS pond halves (26323±4634, mean ± SE) than in LDS halves (32108±4991). As a result, there was a strong density-dependent difference in per capita egg production (mean difference (within ponds)=212±33 SE; paired t test t=6.5, df=3, P=0.004), with eggs laid per stocked fish in the HDS treatment being less than 1/4 of the amount in the LDS treatment (Fig. 2a).
https://static-content.springer.com/image/art%3A10.1007%2Fs00442-006-0410-7/MediaObjects/442_2006_410_Fig2_HTML.gif
Fig. 2

Reproductive success of fathead minnows in two experimental ponds in 1994 and 1995 with low (LDS) and high (HDS) densities: a per capita egg production, b proportion of eggs laid that hatched and c. proportion of young-of-the-year (YOY) that survived from hatching until fall. See Fig. 1 for explanation of symbols

In addition to density-dependent egg production, there were dramatic differences between treatments in the fate of eggs (diseased, predated, hatched). Only 1–5% became diseased, with no obvious difference between treatments. Instead, most losses involved the sudden disappearance of entire batches, beginning within 3 days of laying and well before development could be completed. These events were clearly the result of predation. In the HDS treatment, schools of stocked fish were frequently observed mobbing nest-guarding males and chasing males off their territories. After such encounters, we always observed signs of egg predation, with partially eaten eggs and other debris being all that remained of egg batches. Consequently, survival of eggs to hatch was lower in the HDS treatment than in the LDS treatment (mean difference (within ponds) = 0.58±0.1 SE; paired t test t=5.04, df=3, P=0.007; Fig. 2b). As a result, the number of eggs that successfully hatched in HDS averaged <20% of the number in the LDS pond halves (HDS mean ± SE=4677±2143, LDS=25161±5097).

Young-of-the-year growth and survival

Mid-season net hauls of YOY yielded dramatic differences between treatments in Year 2. Not surprisingly, given the differences in the number of eggs that were hatching, many more YOY were netted in the LDS than in the HDS treatment (Fig. 3). These samples also revealed an intracohort density effect on YOY growth. In both ponds, length-frequency distributions differed between treatments (Kolmogorov–Smirnov two-sample tests Pond 1: D=0.629, P=0.001; Pond 2: D=0.68, P<0.001), with lengths of HDS YOY generally greater than lengths of LDS YOY (Fig. 3). Although we could not assess if there was a comparable mid-season density effect on condition of YOY (mass was not measured), there was a trend for condition to be greater in HDS pond halves in the fall (paired t test, t=2.23, df=2, P=0.15).
https://static-content.springer.com/image/art%3A10.1007%2Fs00442-006-0410-7/MediaObjects/442_2006_410_Fig3_HTML.gif
Fig. 3

Mid-season (July 15) length-frequency distributions of young-of-the-year (YOY) fathead minnows in Ponds 1 and 2 with low (LDS, shaded bars) and high (HDS, open bars) densities of stocked fish in 1995. Sampling effort was equal in all pond halves, therefore, sample sizes should reflect relative densities of YOY

Overall, survival of YOY from hatching to ice-up was higher in the HDS treatment (mean difference (within ponds) = 0.44±0.07 SE; paired t test t=7.14, df=2, P=0.019; Fig. 2c). As a result, numbers of YOY in the fall converged between treatments; five of the six pond halves had 1,850–5,000 YOY regardless of treatment and no treatment effects were detected (number of YOY at ice-up: HDS mean ± SE=2991±1119, LDS=2050±960; paired t test, t=0.69; df=2; P=0.56). The exception was the LDS treatment in Pond 2–95, which had very few YOY (n=200) in the fall. Interestingly, this pond half had the highest initial number of fry. Indeed, when we compared mortality of YOY between treatments using initial number of hatched YOY as a covariate, the latter explained the majority of the variability in mortality (ANCOVA: F=41.74; df=1, 3; P=0.008), relegating the effect of stocked fish density (HDS vs. LDS) to a marginal role (F=9.3; df=1, 3; P=0.06). The relationship between the number of surviving YOY versus initial number of YOY was unimodal, best fit by a quadratic equation that explained 89% of the variation in YOY survivors, indicating that total survivors decreased as initial numbers increased beyond ca. 15,000 (Fig. 4a).
https://static-content.springer.com/image/art%3A10.1007%2Fs00442-006-0410-7/MediaObjects/442_2006_410_Fig4_HTML.gif
Fig. 4

a Number of young-of-the-year (YOY) fathead minnows from a given pond-year that survived over the summer as a function of initial number of YOY that hatched. The quadratic regression equation is: Number of survivors=330.52+0.57(initial no. YOY)−1.89E-05(initial no. YOY)2; R2=0.89 (F=12.4, df=5, P=0.04). b Coefficient of variation (CV) of final (fall) lengths of young-of-the-year (YOY) fish from a given pond-year as a function of the initial number of YOY. The regression equation is: CV=35.8−5E-4 (initial no. YOY); R2=0.92 (F=56.3, df=5, P=0.001). See Fig. 1 for explanation of symbols

Despite strong mid-season differences in growth, sizes of YOY that survived to the fall had converged between HDS and LDS populations (HDS mean ± SE=21.9±1.8 mm, LDS=22.1±2.2 mm; paired t test, n=3: t=0.29, P=0.79). Similarly, there was no relationship between final mean size and initial number of YOY (Regression F=0.30; df=1, 4; P=0.61). Likely as a consequence of differences in intra-cohort density-dependent selective mortality, however, the coefficient of variation of the final mean size (TL) of YOY was strongly related to initial number of YOY (Fig. 4b). Final sizes were less variable in cohorts with higher initial densities and higher post-hatch mortality.

Discussion

The populations of fathead minnows in this study exhibited very strong density-dependent processes at a number of different life-history stages. We documented density-dependent growth of stocked fish and observed strong density-dependent effects on per capita egg production, egg survival to hatch, mid-season size structure of YOY minnows, and YOY survival. Despite initial stocking densities that differed by fourfold, interactions and interdependencies of stage-specific processes in each treatment ultimately resulted in a convergence of abundance and average length of YOY by the end of the season. Although similar, these recruitment outcomes were achieved via alternative pathways, with density-dependent processes acting most strongly on the egg stage in HDS populations and on the larval stage at low stocking densities. Because the dominant density-dependent processes differed, however, in their (size) selectivity, the two treatments produced cohorts of young that differed in body size variation.

The loss to flooding of one pond (1–95) eliminated our replicate for that pond and year for end-of-season measurements and introduced the possibility that end-of-year (but not earlier) results were being driven by either the remaining pond or the 1994 data. However, examination of the end-of season results from individual pond-years showed that treatment effects, when present, were strong and consistent, not driven by a single year or pond. Furthermore, fathead minnows stocked at nearly the same density as the LDS treatment in a subsequent year (and in different pond halves; Grant and Tonn 2002) produced mean values for number of eggs laid per stocked fish (441), proportion of eggs that hatched (0.81), and proportion of YOY surviving to ice-up (0.04) that are consistent with our LDS results and inconsistent with our HDS results.

Density-dependent effects on the stocked fish

Density-dependent mortality is common in early life-stages of fish, whereas growth and reproduction are the density-dependent traits more often observed later in life (Charnov 1986; Lorenzen and Enberg 2002); this was also the case in our experiment. Overall, per capita mortality of stocked fish was similar for HDS and LDS populations, therefore, the treatment (high and low stocked fish density) was maintained throughout the experiment. The slightly higher survival of adult males in the HDS than in the LDS treatment could well have been due to reduced reproductive investment, and therefore reduced post-reproduction mortality, in the HDS treatment (Andrews and Flickinger 1974). Growth of stocked fish was, however, strongly reduced in the HDS treatment, most likely related to exploitation competition among the stocked fish (Tonn et al. 1994). Adult fathead minnows tend to take bigger prey than juveniles (Price et al. 1991) but otherwise feed on a similar prey base (chironomids, copepods, cladocerans, and detritus), with moderate to high overlap in diets (Schoener’s (1974) diet overlap index=0.55–0.81; Janowicz 1999), suggesting the possibility for strong competitive effects among the stocked fish cohorts.

Density-dependent effects on other life-history stages

Our results indicated that the density of stocked fish also interacted, directly or indirectly, with other life-history stages to affect the recruitment pattern of YOY fish. The strong density-dependent effect on per capita egg production, for example, was likely due to greater interference at establishing and maintaining nesting territories in HDS (Danylchuk and Tonn 2001), rather than to differences in fecundity between HDS and LDS; there were no initial differences in adult sizes and condition between treatments that would influence fecundity and adults began reproducing within a week of being stocked. Interference is also consistent with the higher survival of adult males that we observed in the HDS treatment (Andrews and Flickinger 1974). We also noted that a number of egg batches in the HDS treatment only had 5–10 eggs in them (versus an average of ca. 300 for both treatments), suggesting that spawning attempts were aborted. Finally, the greater growth of stocked juveniles in LDS pond halves likely allowed some to mature and reproduce, especially later in the protracted spawning season (Danylchuk and Tonn 2006), and thus contribute to the greater per capita production of eggs in LDS.

In contrast, the availability of nesting sites was unlikely to have limited egg production in HDS pond halves. Nest territory size is very compressible in fathead minnows and many males will communally guard adjacent egg batches at good sites where nests are clustered (Andrews and Flickinger 1974). In terms of surface area, we provided more nesting substrate than the amount required by males (Flickinger 1973), even assuming that every stocked male in HDS established a nest; indeed, we observed empty nest sites every day.

Greater per capita egg production in the LDS treatment, in fact, slightly overcompensated for differences in the density of stocked fish. Although the once-a-day sampling regime might have missed some batches of eggs produced and eaten in the same day, it was relatively rare to observe signs of a depredated nest (e.g., partially eaten eggs, egg debris) where we had not observed an egg batch the day before.

There was also a strong interaction between stocked fish density and egg survival. Most, if not all, of the difference in egg survival was likely due to cannibalism by stocked fish. There were no other vertebrate predators in the ponds, and daily observations and monthly minnow trapping revealed no obvious differences in the numbers of potential invertebrate predators, e.g., dytiscid beetles and odonate larvae (R. Vandenbos and W. Tonn, personal observation). Instead, the large numbers of conspecific egg predators in high-density ponds appeared to frequently overwhelm guarding males (Hyatt and Ringler 1989; R. Vandenbos and W. Tonn, personal observation). Although males defended their nests more actively in the high-density treatment (R. Vandenbos, personal observation), reductions in male condition due to the increased (and continual) defence activities could have limited the ability to successfully sustain nest defence (Unger 1983). Finally, because food was likely more limiting in the HDS treatment, eggs likely became a valued food item (Polis 1981).

Initial post-hatch differences in density of YOY fish were still present during the mid-season sampling, which also had indicated intra-cohort density-dependent early growth of YOY. By the end of the season, however, both mean size and density of YOY had converged between treatments. Clearly, YOY in LDS treatments experienced higher mortality after mid-season and this density-dependent mortality was selective based on size (Post and Prankevicius 1987) or growth rate (Rice et al. 1993). Not only had final mean sizes converged, but variability of the final size distributions was inversely related to mortality rate of YOY during this life-history stage.

Adult fathead minnows can cannibalize YOY in a size-limited manner in the laboratory (Vandenbos 1996); the intra-cohort density-dependent growth of YOY at mid-season meant that the slower growing YOY in LDS would have been more vulnerable to this source of mortality than the HDS YOY. Starvation was also a likely source of size-selective mortality (Frank and Leggett 1994); the density-dependent growth that YOY fish were experiencing is often an indication of poor food availability (e.g., Fox et al. 1989). Indeed, nutrient enrichment (leading to increased food availability) increased the number of YOY fathead minnows surviving to the end of the season by more than fivefold in another study (Grant and Tonn 2002). Clearly, the two mechanisms are not mutually exclusive, as small fish in poor condition are more susceptible to predation, as well as to starvation (Sogard 1997). Either way, intra-cohort density-dependent competition appeared to be ultimately behind the large difference in YOY mortality between treatments.

The convergence of final average lengths of the YOY, despite the mid-season differences, was, in turn, a consequence of the density-dependent, size-selective mortality of YOY in LDS, which also produced the strong intra-cohort density-dependent variability in the final size distributions (see also Elliott 1990). This difference in size variation at ice-up could have significant recruitment consequences. An important source of mortality for YOY fish is the overwinter period, especially for populations like fathead minnows in Alberta, near the northern limit of the species’ range (Shuter and Post 1990; Munch et al. 2003). Overwinter mortality can result from depletion of energy reserves (Post and Evans 1989) and lowered osmoregulatory ability (Johnson and Evans 1996), both of which are influenced by body size and condition (with smaller individuals faring worse), and by the length and severity of the winter (Sogard 1997). Indeed, strong size-selective mortality was displayed by YOY allowed to overwinter in these same experimental ponds in another experiment (Grant and Tonn 2002). Although YOY around 20 mm in length comprised a substantial proportion of the cohort in the fall (similar to our study, where mean length was ca. 22 mm), most survivors in the spring were >29 mm, and none were <20 mm. Thus, size-selective overwinter survival should have different consequences for populations with the same mean size but different variation in size; similar consequences of variation have been suggested with size-dependent predation (Rice et al. 1993).

The specific minimum threshold size needed to survive the winter period will differ from 1 year to the next, reflecting differences in the length or severity of the winters (see also Danylchuk and Tonn 2003). Whether cohorts with broad or narrow size distributions would display greater recruitment would thus depend on the mean size that the cohort achieved in the fall, and the threshold size required for overwinter survival.

Conclusions

Although it is commonly accepted that processes acting during the first year of life are of critical importance to recruitment in fish (or other stage-structured) populations (Sinclair 1989), there are actually several life-stages during the first year, e.g., eggs, larvae, post-larval YOY, each of which can have stage-specific processes that may be important directly or through interactions with other stages and cohorts. Therefore, it should be rewarding to understand what those stage-specific processes are and how density-dependent processes can interact among stages to drive recruitment patterns.

In our experiment, variability in end-of-summer YOY size distribution was ultimately related to the life-history stage at which the dominant density-dependent processes occurred, and whether or not those processes were size-selective. In the HDS treatment, the dominant density-dependent processes occurred during the egg stage, with little or no size-selectivity; this led indirectly to a final YOY size distribution that was more variable. In the LDS treatment, the dominant process(es) occurred during the YOY stage, resulting in strong size-selective mortality; this, in turn, produced a YOY cohort that was less variable with respect to size.

Although the processes acting on different life stages can be distinct, the stages themselves are not independent. Outcomes of processes that act, or fail to act, during one stage will ‘cascade’ to subsequent stages, in a manner analogous to interactions among trophic levels in a food chain (Carpenter et al. 1985), by contributing to the traits of subsequent stages and to the biotic environment in which these later stages must operate (Fig. 5). Such cascading processes can continue to shape populations for years (Hamrin and Persson 1986; Sanderson et al. 1999).
https://static-content.springer.com/image/art%3A10.1007%2Fs00442-006-0410-7/MediaObjects/442_2006_410_Fig5_HTML.gif
Fig. 5

Flow diagram illustrating the alternative outcomes of cascading density-dependent processes important to different life-history stages of fathead minnows, which interact to affect YOY recruitment in populations with: a low density, and b high density of older age-classes. Sizes of boxes represent relative densities at a given stage; width of horizontal (no fill) arrows reflects the importance of processes that influence the transition of individuals from one stage to the next. Width of vertical (black fill) arrows reflects the magnitude of the per capita rates represented by the arrows. Processes highlighted in bold are size-selective. Dashed lines indicate uncertain outcomes. FYOY=YOY in the fall; J2 = juveniles at the beginning of the second growing season

The relative strengths of processes that act and interact during this cascade of life-stages, and the different outcomes that these processes can produce, indicate that a focus on a single life-history stage can restrict our understanding of recruitment and population dynamics. To better understand population regulation in species with stage-structured life histories, we need to consider not only the mechanisms that act during individual life-history stages, but how stages interact.

Acknowledgements

We gratefully acknowledge M. Janowicz, S. Grant, K. Vladicka, and L. Rempel for field assistance, and S. Boutin, M. Mangel, E. Marschall, C. Osenberg, C. Paszkowski, J. Post, J. Roland, B. Wilson, and an anonymous reviewer for constructive comments and advice. This research was generously supported by grants to REV from the Electric Power Research Institute, Canadian Circumpolar Institute, and Alberta Conservation Association (Biodiversity Challenge Grant), a capital equipment grant to Meanook Biological Research Station, and by a Research Grant from the Natural Sciences and Engineering Research Council of Canada (NSERC) to WMT. REV was supported by an NSERC Post-Graduate Fellowship, assistantships from the Department of Biological Sciences, and professional development funding from Selkirk College. The research was conducted under an animal care permit issued by the Biosciences Animal Policy and Welfare Committee, University of Alberta, and collection and research permits issued by Alberta Fish and Wildlife.

Copyright information

© Springer-Verlag 2006