Allee effects within A. napellus populations
Our results confirm our hypothesis that the spatial distribution of both plants and flowers within natural populations (number, density, and isolation) influences reproductive components of A. napellus ssp. lusitanicum.
We found that floral display influenced an individual's reproductive success. As expected, inflorescences with many flowers produced more seeds per flower, and this effect of floral display size was stronger at low density. This suggests that heavily flowered individuals received more viable pollen per flower than sparsely flowered individuals, presumably because of the greater attractiveness of the former. Thus, reproductive success was at least partially determined at the plant level.
We also found that reproductive success was affected by larger scale patterns of floral distribution (local density and isolation distance). Seed production of A. napellus in open-pollinated flowers was positively correlated with local flower density and negatively correlated with isolation distance. Furthermore, the number of seeds produced by open-pollinated flowers was always lower than the number of seeds produced by hand-supplemented flowers. This set of results suggests that A. napellus suffers from within-patch Allee effects as a result of pollen limitation in low-density or isolated patches.
We also examined seed quality, germination rate and juvenile survival. The seeds produced in low-density patches were on average heavier than those produced in high-density ones. This could result from a resource allocation trade-off between seed mass and seed number (De Jong & van Noordwijk, 1992). Seed mass and local density influenced germination rate. The lower number of (heavier) seeds produced in low-density patches was not counterbalanced by a higher germination rate (Oostermeijer et al., 1994; Kéry et al., 2000; Kolb, 2005 but see Menges, 1991; Fischer et al., 2003). In fact, both germination rate and offspring survival were lower for seeds produced in low-density patches. Similar results have also been found in small populations of other plant species (Oostermeijer et al., 1994; Fischer & Matthies, 1998; Fischer et al., 2003; Kolb, 2005).
Four processes could lead to this reduction in fitness at low local densities or in isolated patches. First, local densities could reflect differences in environmental quality (the availability of water, soil nutrients or light). For example, if patches were variable in quality, we would expect to find a positive correlation between adult plant density and mean individual vigor. Secondly, pollinators of A. napellus may exhibit less constancy when other nectar-producing flowers are available; that is, the frequency of allospecific pollen transfer may increase, clogging A. napellus stigmas with pollen that does not germinate or that cannot fertilize its ovules. Alternatively, other flowering species could act as magnets and promote visitation of A. napellus. These phenomena should affect seed production but not the germination rates of the seeds produced. Thirdly, in patches where floral density is low, the frequency of self-fertilization or pollination between closely related genotypes may increase, thereby reducing seed production as a consequence of partial self-incompatibility or inbreeding depression, or seed quality as a consequence of inbreeding depression. Finally, pollinators may be less attracted to flowers when they occur at low density, such that seed production becomes highly pollen-limited.
We found no evidence suggesting that variation in patch quality – the first hypothesis – is the cause of the observed variation in reproductive success; that is, we found no relationship among the total number of flowers per inflorescence (open flowers and buds), the number of axillary stems per plant, stem height, and the number of open flowers per patch. If low-density patches were resource-limited, they should produce fewer seeds, smaller seeds, and/or seeds with lower germination rates. Contrary to this prediction, in our study, seeds were heavier in low-density patches. These results are consistent with those reported by Lamont et al. (1993), Widén (1993), Fischer & Matthies (1998) and Costin et al. (2001).
Similarly, the number of individuals of other flowering nectariferous plants – the second hypothesis – did not negatively affect seed production in A. napellus (cf. Caruso, 1999, 2001; but see Waser 1978; Campbell & Motten 1985; Galen & Gregory 1989; Jennersten & Kwak 1991; positive effects have been reported by Thomson, 1981; Moeller, 2005). We detected no significant effect within patches of the number of plants of other rewarding species on seed production per flower. This is in accordance with our field observations (unpublished results); bumblebees visiting A. napellus flowers are mostly specialists and avoid the flowers of other species. A. napellus pollen transported by pollinators is probably not significantly diluted with pollen from other plant species.
The third hypothesis for the observed reduction in seed production per flower at low floral density – an increase in selfing rate and/or in consanguineous matings – is likely to apply here: an example of a genetic Allee effect. In our experiment, the reduction of 20% in seed production between geitonogamous and outcrossed hand-pollinations suggests three possible mechanisms: a failure of self pollen to germinate or to grow effectively, early-acting inbreeding depression or late-acting maternal selection occurring between pollination and seed maturation. Moreover, the greater the distance between pollen donor and recipient, the higher the seed production per flower, indicating that pollinations between proximate and probably related plants resulted in a higher number of unviable offspring than crosses between more distant plants. At low density, plants have fewer available mates, which may result in higher rates of selfing and crosses with related individuals. This is consistent with a pattern in which pollinators are more likely to transfer pollen between closely related plants in low-density patches than in high-density patches. Indeed, when plants are scarce, pollinators tend to maximize their foraging efficiency by visiting more flowers per plant, thus increasing geitonogamous (within-plant) pollen transfer (Heinrich, 1979; De Jong et al., 1993; Klinkhamer & De Jong, 1993; Ferdy & Smithson, 2002; Ohashi & Yahara, 2002; Oddou-Muratorio et al., 2006). Relative to outcrossing, such geitonogamous selfing results in reduced offspring fitness (germination rate and juvenile survival) because of inbreeding depression.
Low seed production and poor offspring quality in low-density patches could result from inbreeding depression (biparental inbreeding) and/or maternal control over fertilization and provisioning (Stephenson & Winsor, 1986; Shaw & Waser, 1994; Charlesworth & Charlesworth, 1999; Kenta et al., 2002). Such results of maternal control have been found in Delphinium, a genus close to Aconitum (Waser et al., 1987; Waser & Price, 1991, 1993). Based on the present results, we are not able to disentangle the effects of deleterious recessives in self-pollinated flowers from the preferential allocation of maternal resources to cross-pollinated flowers. Regardless of the mechanism, however, genetic Allee effects may reduce fitness components in plants growing in low-density patches (Fischer et al., 2003; Kolb, 2005; Willi et al., 2005). Differences in pollen quality could also explain the significant positive relationship observed between seed production among within-patch hand-pollinated flowers and floral density. However, this mechanism is not sufficient to explain the magnitude of the difference between hand-supplemented pollinations and open-pollinations. Thus, plants in low-density patches appear to suffer from a decline in both the quality and the quantity of the deposited pollen.
Finally, the fourth hypothesis – that A. napellus is more pollen-limited at low density than at high density – may also contribute to the observed variation among patches in seed production. When patches of flowering spikes are quite small, pollinators are unlikely to encounter them by chance or are not sufficiently attracted to spend the energy necessary to visit them (e.g. Schulke & Waser, 2001; Kirchner et al., 2005; Cheptou & Avendaño, 2006). At very low density, some flowers produced no seed. If pollinators are less attracted by small patches, they are also less attracted by isolated patches. Indeed, we found that seed production decreased as the distance between two adjacent flowering patches increased. More precisely, the number of seeds per fruit increased with distance when patches were less than 8 meters apart, and decreased with distance for distances between 8 and 18 m. In patches isolated from others by > 18 m, seed production was very low. This suggests that reproduction in small and isolated patches of A. napellus ssp. lusitanicum is limited by the amount of pollen deposited, which constitutes an ecological Allee effect.
The third and fourth hypotheses are supported by studies performed on pollinator behaviour that showed that low-density patches received few visits but bumblebees visited more flowers per inflorescence, and inflorescences in isolated patches at very low density were not visited at all (Le Cadre, 2005).
By comparing the results of open and hand-supplemented pollinations, we can estimate the relative contributions of genetic and ecological Allee effects to the reduction of fitness components in A. napellus patches. To measure the ecological Allee effects more precisely, it would be necessary to measure pollen deposition among the stigmas of flowers in patches representing different densities and degrees of isolation.
In summary, our study highlights the importance of considering variables measured at different scales to determine factors controlling the reproductive success of a given species. It also reveals that comparisons of fitness components at the population level may be unable to detect component Allee effects such as density-dependent pollen limitation. In our study, we did not detect any significant effect of population identity on reproduction and fitness. These results underscore the value of examining directly those attributes of population structure that significantly contribute to plant reproductive success.
Effect on population viability and implications for conservation
In this study, we detected component Allee effects as defined by Stephens et al. (1999). In fact, seed production per flower (a component of individual fitness) of A. napellus is reduced at low density as a result of reductions in both pollen quantity and quality. The extent to which this impact results in a decline in population growth rate – a demographic Allee effect – depends on the life history of the species. Unfortunately, in the case of A. napellus, we have no data to support a demographic Allee effect because the populations have not been monitored for long enough (2 yr only) and because individual-based demographic censusing is not possible in this species (because of vegetative reproduction, individuals are not discernable).
Nevertheless, we deduce that seed production in A. napellus depends strongly on insect pollination and that the distribution and size of plant patches may be critical to the success of any restoration plan. Indeed, for all species, there may be a density threshold below which the extinction risk is high (Dennis, 1989; Kunin & Iwasa, 1996). However, few empirical studies have been conducted to detect the parameters of this extinction threshold. For example, Groom (1998) established a threshold number of individuals and a threshold distance between patches of Clarkia concinna (respectively, 50 individuals and 16 m of isolation). In the case of A. napellus populations, patches with < 12 flowers or isolated by > 18 m face uncertain reproductive futures. To manage populations, we could either reinforce existing patches or create relay patches (Kwak & Vervoort, 2000). Further studies of the potential effects of pollinator composition and behavior, and their spatial and temporal variation, on plant reproductive success should be investigated (Kwak et al., 1991). We propose that the construction and monitoring of experimental populations would be a useful tool for determining the existence of extinction thresholds and the population parameters (i.e. the distribution, abundance, and size of subpopulations) at which they occur. Such an approach could contribute to the identification of a long-term population management plan (Bosch & Waser, 2001; Schulke & Waser, 2001).