Germination variation facilitates the evolution of seed dormancy when coupled with seedling competition

Abstract

Fluctuating environmental conditions have consequences for the evolution of life histories because they cause fitness variance. This variance can favor risk-spreading strategies, often known as bet-hedging strategies, in which growth or reproduction is spread over time or space, with some costs, but greater certainty of success. An important example is seed dormancy in annual plants, in which some fraction of seed remains dormant at any given germination opportunity with the potential of germinating later when environmental conditions may differ. Previous theory shows that environmental variation is critical for the evolution of dormancy. However, these studies have focused on temporal variation in reproduction, ignoring the strong observed effects of environmental variation on the germination fraction, a major contributor to fitness variance. We ask what effects germination fluctuations have on selection for dormancy by adding germination fluctuations to existing density-independent (d.i.) and density-dependent (d.d.) models of annual plant dynamics, extending previous analyses by including temporally fluctuating germination. Specifically, we ask how germination variance affects selection on the temporal average germination fraction, here used to define dormancy. When present alone, or when independently varying with other fitness components, germination fluctuations do not affect selection for dormancy in the d.i. model, despite generating fitness variance because this variance contribution is not reduced by higher dormancy. Germination fluctuations have strong effects in the d.d. model, favoring dormancy when present either alone or coupled with variation affecting plant growth. This is because germination variation causes seedling density to vary, which causes variable reproduction through variable intraspecific competition. Dormancy is advantaged under variable reproduction because it creates a more convex relationship between population growth and reproduction leading to benefits from nonlinear averaging. Predictive germination, a positive statistical association between germination and growth, weakens selection for dormancy under strong competition and strengthens it when competition is weak. Our results suggest that variable germination is a potential explanation for high levels of dormancy observed in nature, with implications for life-history theory for fluctuating environments.

Introduction

It has long been appreciated that the complex life histories of many organisms may be due in part to adaptation to variation in the physical environment (Cohen, 1966, Schaffer, 1974, Roff, 1993, Koons et al., 2008). Annual organisms, which might be considered to have the simplest life histories, are often not strictly annual, because a resting stage, such as a seed, egg, or pupa may persist over several reproductive seasons (Evans and Dennehy, 2005). Thus, understanding the evolution of life-history complexity can begin with the study of why these simplest life histories are not simpler yet. Indeed, a notable case of this phenomenon is that of between-year dormancy in annual plant species, which extends the life of an individual beyond one reproductive season on average. These annual plants live in environments where conditions for germination might be favorable at some given time although only a fraction of the seed of a given annual plant species germinates (Venable and Pake, 1999, Venable, 2007). The remaining fraction stays dormant with the potential that some might germinate in the next or later seasons. Dormancy has been a puzzle because, under favorable conditions for seed production, delaying germination can only reduce fitness. Further, delayed germination increases the risk that a seed dies before germination.

One explanation for the occurrence of between-year dormancy, with theoretical and empirical support, involves temporal environmental fluctuations (Cohen, 1966, Bulmer, 1984, Ellner, 1985a, Ellner, 1987, Gremer and Venable, 2014). Growing conditions vary from year to year and, without dormancy, unfavorable years for growth lead to major fitness losses (Cohen, 1966). However, theoretical models demonstrate that dormancy reduces the possibility that all offspring encounter unfavorable conditions, i.e. increases the probability that at least some offspring encounter favorable conditions, which is adaptive even given the cost of delaying reproductive opportunities (Cohen, 1966, Bulmer, 1984, Ellner, 1985a, Ellner, 1985b). This idea has been generalized and referred to as bet-hedging, a strategy that reduces variance in fitness at the cost of reduced average fitness (Seger and Brockmann, 1987, Philippi and Seger, 1989, Simons, 2011). Models of both density-independent and density-dependent fitness predict that dormancy evolves in fluctuating environments, although greater dormancy evolves with density-dependence (Bulmer, 1984, Ellner, 1985a, Ellner, 1985b) largely due to competition’s role in enhancing the effects of environmental fluctuations that affect reproductive success (Ellner, 1987).

These previous studies assume that the effects of the changing environment on fitness are fully captured by variation in the per individual seed yield. However, seed yield is not the only fitness component that responds to environmental fluctuations. In nature, environmental conditions have major effects on a species’ germination fraction (Went, 1949, Juhren et al., 1956, Harper, 1977, Venable and Pake, 1999, Facelli et al., 2005), an important fitness component. This observation applies to very many annual plant species (Baskin and Baskin, 2014). For animals, variable emergence from various quiescent states is a well-known analogous phenomenon (Hakalahti et al., 2004, McAllan and Geiser, 2014). Variable germination in annual plants contributes to striking patterns in nature such as the spectacular wildflower blooms of the arid Southwestern USA, which vary greatly over time due to species-specific environmentally-dependent germination (Juhren et al., 1956, Bowers, 1987, Bowers, 2005). In addition to precipitation during the growing season — which clearly has a large effect on germination rates — other abiotic factors such as temperature at the time of rainfall explain if and how readily plants germinate (Venable and Pake, 1999, Huang et al., 2016). Moreover, variable germination figures prominently in the theory of annual plant coexistence by the storage effect (Chesson, 1994, Chesson et al., 2004, Holt and Chesson, 2014). The bet-hedging role of seed dormancy is an important component of the storage effect, which, within that theory, is called buffered population growth (Chesson and Huntly, 1989, Chesson and Huntly, 1997, Chesson et al., 2004).

These observations mean not only that a major component of variation in fitness is missing from previous studies, but that dormancy itself is represented in an overly simplified way. When dormancy is treated as a constant rather than a variable, potential major implications for population dynamics and species interactions are missed. Rather than characterize germination as a particular number, germination should be characterized as a function of environmental conditions, i.e. as phenotypically plastic. Because environmental conditions vary unpredictably from year to year, germination, and thus dormancy, are naturally characterized by a probability distribution. Here we ask about the evolution not of the dormancy distribution per se, but its temporal mean. The mean germination fraction is a natural interpretation of the constant germination fractions studied in previous investigations, with the variance reflecting environmental variability. Average reproductive opportunities decrease as the mean germination fraction decreases, whether germination varies temporally or is constant. By recognizing germination variance, we are better placed to understand how the evolution of dormancy may shape coexistence through the storage effect.

Our interest is in the effect of the germination variance on the evolution of the mean, but yield variation is also an undeniably major factor in nature. Hence, as in previous studies, it is integral to our investigation. We ask about the combined effects of per-capita yield variance and germination variance on the evolution of dormancy, including the effects of their correlations. When germination and yield are positively correlated, germination is said to be predictive. Predictive germination is expected to evolve when environmental cues to growing conditions are present at germination or when similar conditions affect both germination and yield (Cohen, 1967). Predictive germination has been noted empirically (Gremer et al., 2016) and it means that a species has a higher tendency to germinate when conditions will be favorable to growth, and not germinate when conditions will be unfavorable. Under such circumstances, variable germination is an adaptively plastic trait (Simons, 2011). Because adaptive plasticity has been assumed to be an alternative to bet-hedging in variable environments (Dewitt and Langerhans, 2004, Simons, 2011), it might be expected to influence the evolution of dormancy in addition to any effects of variable germination or variable seed yield alone.

Previously, plant yield was treated in models of the evolution of dormancy as a single parameter that, even in density-dependent models, acts in a density-independent way. A notable exception is Ellner (1985b), who allows that the intensity of per-capita competition might be environmentally variable. Such linkages between environmental factors and competition have potential roles in the evolution of dormancy and also have critical roles in species coexistence by the storage effect (Angert et al., 2009). To account for environmentally-dependent competition, we assume that the physical environment affects plant growth, which we refer to as vigor following Chesson et al. (2005), and that larger plants have large per-capita competitive effects (Weiner, 1990, Schwinning and Weiner, 1998). Thus, per-capita competitive effect is larger in years when plants grow larger, i.e. in years when plants have higher vigor.

Including environmental variation in both germination and plant growth, we evaluate the effects of environmental fluctuations on selection for dormancy. As mentioned earlier, temporal fluctuations in per-capita seed yield are known to favor dormancy both theoretically (Cohen, 1966, Ellner, 1987) and empirically (Philippi, 1993, Gremer and Venable, 2014), but our question relates to whether fluctuating germination favors dormancy beyond that favored by fluctuating yield. To do so, we use a model of annual plant growth in a fluctuating environment in which germination fractions vary from year to year, phenotypes differ in their temporal mean germination, and all phenotypes have fixed germination variance. Included is the potential for correlations between germination and seed yield and their effects on the evolution of dormancy in models of density-independent and density-dependent fitness. We first investigate the density-independent model to show that the evolution of dormancy depends critically on a mean–variance trade-off in per-capita seed yield. Germination variation, when independent of vigor, does not contribute to this tradeoff but does when correlated with vigor due its consequences for reproduction. We then study how this tradeoff is affected by density dependence. Because germination variation causes density dependence to vary, with flow on effects to reproduction, evolution of dormancy can be affected by germination variation even when germination is not correlated with vigor.