Introduction
Seed dormancy and dispersal have often been considered alternative mechanisms allowing sedentary organisms to persist in heterogeneous, dynamic landscapes (Venable and Brown, Reference Venable and Brown1988). The guild of gap-dependent forest herbs allows a test of this hypothesis. Whereas most temperate-zone forest herb species show little seed dormancy and only occasionally appear in the soil seed bank (Matlack and Good, Reference Matlack and Good1990; Hyatt and Casper, Reference Hyatt and Casper2000; Leckie et al., Reference Leckie, Vellend, Bell, Waterway and Lechowicz2000), a subset of forest herbs does appear in seed banks. These species are generally recognizable by their low tolerance of shade, small seeds, effective dispersal and opportunistic use of forest gaps (Jankowska-Blaszczuk and Grubb, Reference Jankowska-Blaszczuk and Grubb1997; Hyatt and Casper, Reference Hyatt and Casper2000). Canopy gaps are isolated and ephemeral microhabitats, placing severe limitations on the life histories of light-demanding species (Reader and Bricker, Reference Reader and Bricker1992). A light-requiring forest herb may benefit from long-range seed dispersal, allowing new gaps to be colonized as existing gaps close by succession (Gadgil, Reference Gadgil1971). Alternatively, dormancy in the soil may allow populations to persist between successive gaps at a single site, accomplishing dispersal in time rather than space (Hyatt and Casper, Reference Hyatt and Casper2000).
Seed dormancy may serve to increase reproductive success of a species in an uncertain environment by distributing emergence of seedlings over several growing seasons, thereby buffering the failure of reproduction or germination in occasional poor years (Rees, Reference Rees1993; McPeek and Kalisz, Reference McPeek and Kalisz1998). Dormancy will be favoured when the risk of mortality as a seedling exceeds the risk of mortality of seeds in the soil. Mature seeds can germinate quickly in response to environmental cues (enforced dormancy; Harper, Reference Harper1977) or may occupy a deeper state of dormancy in which they are insensitive to environmental cues until exposed to a specific trigger condition (induced dormancy). Both dormancy types may occur within a single cohort of seeds, which can be considered a way of spreading risk in an unpredictable environment analogous to spatial dispersal (‘bet-hedging’ sensu Venable and Brown, Reference Venable and Brown1988).
In this paper we examine seed longevity and seasonal changes in dormancy state in the perennial forest herb Ageratina altissima to test the complementarity of dormancy and dispersal. This species prefers isolated disturbed sites but dispersal ability appears to be limited. If seed dormancy is an evolutionary alternative to dispersal between treefall gaps, then we would expect A. altissima to have long-term persistence in the soil and a low proportion of germinable seeds.
Ageratina altissima (L.) King and H. Rob. var. altissima (Asteraceae; formerly Eupatorium rugosum, hereafter ‘Ageratina’) is a short-lived perennial herb native to deciduous forests of eastern North America, and common in forests of the upper Midwest. Ageratina prefers partially shaded moist or mesic microhabitats, typically occurring in forest edges, natural treefall gaps and burned areas (Clewell and Wooton, Reference Clewell and Wooten1971; Nuzzo et al., Reference Nuzzo, McClain and Strole1996; Landenberger and Ostergren, Reference Landenberger and Ostergren2002).
The species blooms in late summer and early autumn, potentially producing thousands of seeds per plant in well-lit sites. Seeds have the achene–pappus morphology characteristic of the Asteraceae, suggesting wind dispersal. However, a rapid rate of descent in free air, low release height and a preference for forest sites (which have little air movement near the ground) imply that the dispersal range is limited (Matlack, Reference Matlack1987). Seeds are released in mid–late autumn. At maturity they are in a state of induced dormancy which can be broken by high temperatures, decreasing the likelihood of autumn or winter germination (Lau and Robinson, Reference Lau and Robinson2010). Most germination occurs in early–mid spring (Walck et al., Reference Walck, Baskin and Baskin1997). If germination has not occurred, seeds may re-enter the induced dormancy state (Donohue, Reference Donohue2005). Ageratina has been observed in forest seed banks (Leckie et al., Reference Leckie, Vellend, Bell, Waterway and Lechowicz2000; C.D.H., pers. obs.), although it's not clear whether the seeds were in long-term dormancy or recently produced.
Materials and methods
Seed longevity and dormancy condition were examined experimentally by burying seeds in forest sites and recovering them at regular intervals, following the method of Sarukhán (Reference Sarukhán1974). To obtain large numbers of uniformly healthy seeds, we used commercially available material from outside our region. Seeds were produced in a common garden in southern Minnesota in autumn 2009 from parent genotypes collected in a single wild population in central Illinois, USA (Prairie Moon, Winona, Minnesota, USA). They were buried at study sites in south-eastern Ohio, USA, in the following year. Although the experimental seeds were probably genetically distinct from populations in the study area, south-eastern Ohio is similar to central Illinois in rainfall, temperature, day length and natural vegetation types. The weedy (sometimes invasive) character of the species suggests it is not sensitive to minor geographic differences. Thus, seed origin is unlikely to have affected dormancy behaviour substantially.
Seeds were stitched into 10 × 10 cm nylon mesh bags, which were placed in deciduous forest sites near Athens, Ohio, in July 2010 (i.e. after the period of natural seedling emergence). One hundred seeds were included in each of 120 bags. Bags were placed on the soil surface under a thin litter layer to mimic natural seed deposition. To test for environmental control of dormancy, seed bags were placed on both north- and south-facing slopes. Above-ground vegetation differs strongly between north- and south-facing slopes in the study area, reflecting variation in moisture, temperature and day length (Olivero and Hix, Reference Olivero and Hix1998). Thus, slope aspect can be considered a conservative test for environmental control of dormancy. Bags were recovered at 2-month intervals for 24 months. The complete design included 5 replicate sites × 2 slope aspects × 12 recovery dates = 120 bags, with 100 seeds in each bag.
Recovered seeds were tested for germinability on moist filter paper in the laboratory. Seeds were tested at room temperature (18–21 and °C) under fluorescent lighting (c. 12/12 h photocycle) for 2 weeks, after which germination had ceased in all samples. Seeds that did not germinate were tested for viability by visually inspecting the embryo for firmness and a yellow or green colour. Although a tetrazolium test is the preferred method of assessing viability (Huebner, Reference Huebner2011), such tests were not used here because seeds screened for germinability had already imbibed water and did not readily absorb the tetrazolium solution. The visual and chemical methods agreed strongly in trials on subsets of seeds, so visual examination was considered an acceptable measure of viability.
Differences between recovery dates are assumed to represent changes that would occur in a single sample through time (a ‘chronosequence’, sensu Pickett, Reference Pickett and Likens1989). The number of living seeds (germinable and non-germinable) plotted across recovery dates can be interpreted as a survivorship curve. However, because sampling was destructive and each sampling date represents a separate group of seeds, we do not know the fate of individual seeds. Uncertainty about dates of mortality leads to intractable problems of right-censoring (Crawley, Reference Crawley2007), so conventional survival analysis could not be used. Instead, we compared the number of surviving seeds using a general linear model set up as an analysis of covariance (package GLM), with aspect as a categorical variable and recovery date (number of months in the soil) as a covariate. Recovery date was log-transformed to improve normality.
The analysis was repeated with germinable seeds, defined as the proportion of all living seeds germinating at each recovery date; an arcsin square-root transformation was applied as appropriate to proportional data. Normality and homogeneity of variance were assessed with Shapiro–Wilk and Fligner–Killen tests. Finally, longevity in the seed bank was estimated empirically by regressing surviving seed numbers on recovery date and extrapolating to the time in months when seed number reached zero. All procedures were conducted in the R statistical environment (R Core Team, 2013).
Results
Seed viability declined rapidly, with only 39–50% of seeds surviving to spring of the second growing season (May 2012; Fig. 1), and no seeds surviving to the July sampling date. The best model predicting seed survival included only date (F date = 75.81 with 94 df, P < 0.0001, R 2 adjusted = 0.441). The Shapiro–Wilk and Fligner–Killen statistics were not significant (P > 0.05), indicating that residuals were normally distributed and variance was homogeneous among recovery dates. Slope aspect did not significantly influence seed longevity. Regression estimated a maximum survival of 33.5 months on south-facing slopes (R 2 = 0.46) and 36.5 months on north-facing slopes (R 2 = 0.57).
Seed viability and dormancy condition of Ageratina altissima at ten sites in south-eastern Ohio, USA, as a proportion of the total number of seeds buried. North- and south-facing slopes. The ‘Dead’ section of the graph (white) represents the proportion of seeds that disappeared, presumably decayed. The ‘Germinable’ (shaded) area indicates seeds recovered from buried bags, which subsequently germinated in laboratory trials. ‘Non-germinable’ seeds (diagonal hatching) appeared to be viable but did not germinate in the laboratory. Means and standard errors are plotted. Natural seedling emergence occurs in March and April.
The proportion of germinable seeds varied seasonally, with high germinability corresponding to periods of natural seedling emergence (Fig. 1). Ageratina showed a maximum of 71 and 84% germinability in March 2011 and 2012 on north-facing slopes, but only 0–21% germinability between September and November (both years; and see January 2011). The transformed date term was significant in explaining Ageratina germinability (F date = 21.6 with 92 df, P < 0.0001) but had little predictive power (R 2 adjusted = 0.18), consistent with the episodic pattern of germination. Again, the slope-aspect term was not significant.
Discussion
Ageratina seed longevity was short relative to estimates for the co-occurring Rubus allegheniensis and Phytolacca americana, perennial forest herbs which are similarly gap-dependent (Hyatt and Casper, Reference Hyatt and Casper2000). Consistent with short seed longevity and severe mortality in the soil (Rees, Reference Rees1993), high proportions of Ageratina seeds were germinable at some sample dates. A period of dormancy in the soil potentially allows Ageratina to colonize short-lived forest microsites if such sites occur frequently enough, substituting dispersal in time for dispersal in space. Populations appear to be vulnerable, however, because our projected longevity is substantially less than gap return-times estimated for deciduous forests of eastern North America (estimated return interval 20–30 years at herb level; based on Runkle, Reference Runkle, Pickett and White1985; Barden, Reference Barden1989).
It is possible that the buried-seed method underestimated the longevity of seeds (Van Mourik et al., Reference Van Mourik, Stomph and Murdoch2005). However, Ageratina would have difficulty surviving between canopy gaps even if seed longevity was considerably longer than reported here, so it appears that dormancy is not providing long-term dispersal between gaps. We infer that reports of abundance in the soil seed bank (Leckie et al., Reference Leckie, Vellend, Bell, Waterway and Lechowicz2000) and rapid germination following disturbance (Nuzzo et al., Reference Nuzzo, McClain and Strole1996) reflect frequent reproduction rather than longevity in the soil. Consistent reproduction and reliable vegetative survival may explain Ageratina's persistence in forest sites better than long-term soil dormancy (e.g. Carey and Watkinson, Reference Carey and Watkinson1993). Ageratina perennates vegetatively as a rhizome, giving populations a degree of stability independent of seed dormancy (although it's not clear whether vegetative longevity is sufficient to survive between canopy gaps). Thus, any analysis of life history and environmental heterogeneity in forest herbs should include aspects of vegetative growth and longevity in addition to seed traits.
Dormancy also provides a short-term benefit, potentially allowing a fine-grained response to climatic variation between and within years (Honek et al., Reference Honek, Martinkova and Jarosik1999; Gibson et al., Reference Gibson, Spyreas and Benedict2002). Ageratina showed little autumn and winter germination, consistent with a stratification requirement demonstrated in the laboratory (Walck et al., Reference Walck, Baskin and Baskin1997; Lau and Robinson, Reference Lau and Robinson2010), thus avoiding weather unsuitable for seedling establishment. In spring and early summer, the period of natural germination, a large portion of seeds switched to a germinable state, potentially allowing opportunistic establishment in response to short-term environmental cues. The small fraction remaining in the induced state would provide a hedge against seedling failure in unsuitable years, allowing the possibility of survival to a third growing season (Pake and Venable, Reference Pake and Venable1996; Gibson et al., Reference Gibson, Spyreas and Benedict2002). Ungerminated seeds returned to an induced state in mid summer, allowing Ageratina to avoid autumn germination and greatly increasing the number of induced seeds. We infer that opportunities for germination in the autumn are not reliable predictors of seedling survival (Densmore, Reference Densmore1979; Walck et al., Reference Walck, Baskin and Baskin1997). Return to the induced state is presumably cued by environmental variation, e.g. high summer temperatures (Allen and Meyer, Reference Allen and Meyer1998; Roman et al., Reference Roman, Murphy and Swanton2000). Laboratory trials have shown the proportion germinating to be dependent upon light, temperature and thermoperiod (Walck et al., Reference Walck, Baskin and Baskin1997), consistent with our interpretation of climatic control in natural populations.
In conclusion, a strict dispersal/dormancy trade-off is probably an oversimplification of the spatial ecology of this species. With relatively small seeds, weak spatial dispersal and short soil longevity, Ageratina has few of the seed traits that theoretically compensate for spatial and temporal heterogeneity (Venable and Brown, Reference Venable and Brown1988). The large germinable fraction and seasonal variation described here imply that seed dormancy in Ageratina is an adaptation to short-term variation in climate rather than long-term patterns of site disturbance.
Acknowledgements
We are grateful to Carol and Jerry Baskin, who made helpful suggestions at early stages of this work. Marion Holmes assisted in the field work. Two anonymous reviewers made helpful comments. We thank the landowners, public and private, who allowed us to use their woodlands.
Financial support
This work was supported under Joint Venture Agreement 10-JV-11242303-014 from the USDA Forest Service.
Conflicts of interest
None.
