- Top of page
Lower population sizes in geographically peripheral populations may be the result of the less favourable environmental conditions experienced by individuals in populations at a species’ distribution limit (Brussard 1984). Such conditions are presumed to affect population densities either directly, via lower fecundity or recruitment (e.g. Pigott & Huntley 1981; García et al. 2000; Dorken & Eckert 2001; Jump & Woodward 2003), or indirectly as a result of higher year-to-year variation in fecundity or recruitment, resulting in greater interannual variation in population size (Brussard 1984; Nantel & Gagnon 1999). In general, comparisons of population dynamics at broad scales are rare for plant populations, with Nantel & Gagnon (1999) being, to our knowledge, the only comparison of dynamics of central and peripheral plant populations. For two clonal species, the perennial herb Helianthus divaricatus and the shrub Rhus aromatica, Nantel & Gagnon (1999) reported a larger interannual variation of vital rates and intrinsic growth rates in two populations at the northern limit of the species’ distribution range (southern Quebec) than in two more central (Ontario) populations of each species.
Large demographic fluctuations may reduce the effective population size, leading to population bottlenecks. These bottlenecks induce loss of genetic variance and reduce individual fitness (Hartl & Clark 1997). The presumably higher risk of extinction in peripheral populations and the potential adaptation of marginal populations to different environmental conditions have both acted to focus the attention of conservation biologists on marginal populations (e.g. Haeck & Hengeveld 1981; Hoffmann & Blows 1994; Lesica & Allendorf 1995).
The negative genetic consequences expected from large demographic fluctuations could be buffered by seed banks, which may, in turn, conserve genetic variance (Baker 1989; McCue & Holtsford 1998) and influence effective population size (Nunney 2002). In a predictable environment (as is presumed to occur at the centre of a species’ distribution), seeds should germinate as soon as conditions are favourable (Rees 1994), and lower predictability (at the periphery) would be expected to favour greater allocation to the seed bank, so that seeds remain available until the next favourable season (Rees 1994). Whether banker or non-banker morphs are favoured by selection depends on life-history traits other than the fraction of seeds allocated to the seed bank, such as fecundity (Aikio et al. 2002). Although seed banks play an important role in population dynamics (particularly in annuals, where the long-term persistence of populations completely depends on seed banks; Baskin & Baskin 2001; Aikio et al. 2002) and in evolutionary theory (e.g. Silvertown & Lovett Doust 1993; Baskin & Baskin 2001; Aikio et al. 2002), they have, to our knowledge, never been integrated into the central-marginal model.
In addition to the position of a population within its geographical range, densities or life-history traits may vary, both within and between populations, in such a way that the population effects are superimposed on the range position effects.
We directly compare local populations of the annual species Hornungia petraea Rchb. (L.) (Brassicaceae) in two contrasting regions within the species’ range, and also at the scale of populations within the two study regions. Ten populations in each study region were monitored for 3 years. We hypothesized that Italian populations (representing the centre of the species’ range) would show higher densities than German populations (representing the periphery) for both adult plant and seed bank populations. We expected demographic factors, in particular fecundity of individuals, dynamics of fecundity and annual seed bank dynamics, to contribute to these patterns. A multifactorial study design with nested factors, repeated over 3 years, was set up to partition density effects into effects of country, population and time. This study employs a statistical approach by combining two hierarchical levels and repeated measurements in one single statistical model.
- Top of page
In contrast to the expected pattern of higher population densities in central than in peripheral populations, we found that population densities of H. petraea were significantly lower in Italy than in Germany. This applied both to adult plant populations and to seed bank populations. Studies on population densities across the distribution range are quite rare for plant species and have yielded contradictory results. For example, in Lactuca serriola there was no evidence for a decline in population density towards the species’ northern distribution limit in Britain (Carter & Prince 1985; Prince et al. 1985). In contrast, Carey et al. (1995) found evidence for declining population densities towards the northern distribution limit of the annual species Vulpia ciliata ssp. ambigua in Britain. Recently, Jump & Woodward (2003) confirmed the ‘abundant centre’ hypothesis for the northerly distributed Cirsium heterophyllum, whereas Cirsium acaule, a species from southern parts of the UK, did not confirm the hypothesis. For the two species, V. ciliata and C. heterophyllum, for which the ‘abundant centre’ pattern has been confirmed (Carey et al. 1995; Jump & Woodward 2003), lower densities were explained by reduced fecundity at the species’ distribution limit (as also observed, for instance by Pigott & Huntley 1981; Eckert & Barrett 1993; García et al. 2000; Dorken & Eckert 2001). In contrast, the small difference in fecundity of H. petraea individuals between the two study regions cannot explain the differences in density. Fecundity was affected to a high degree by the year of study, which could even invert the effect of the study region. Such a high interannual variation in seed production is especially well known in annuals (e.g. Mack & Pyke 1983; Schmidt & Levin 1985; Watkinson 1990). However, the greater interannual variation in fecundity in central Italian populations was not reflected in increased interannual variation in the adult population in spring. In contrast, in the study by Nantel & Gagnon (1999) there was a larger interannual variation in vital rates and intrinsic growth rates at the distribution limit of their study species compared with central populations. As with our results, the differences reported by Nantel & Gagnon (1999) were not reflected in population densities, as would have been predicted by the central-marginal model. In both their species there were only minor differences in the densities of the plants in the two types of populations (central vs. peripheral) (D. Gagnon, personal communication).
The fact that the large fluctuations in fecundity in H. petraea populations in Italy are not reflected in adult spring populations possibly means that the seed bank populations act as a buffer against environmentally driven variation in fecundity. We had expected that interannual seed bank dynamics would be more pronounced in the study region with lower population densities, whilst seasonal seed bank dynamics would be more pronounced in the study region with higher population densities. However, there was no evidence that annual seed bank dynamics differed between the two study regions, whereas seasonal seed bank dynamics were significantly influenced by country. We found more seasonal variation in seed bank densities in the denser German populations. This corresponds with the observed population densities in the two countries and with the idea that lower seasonal seed bank dynamics must be expected under less predictable conditions (Rees 1994). In particular, the transition into the seed bank (i.e. the contrast spring/summer) showed clear differences between the two countries. In the second study year, Italian H. petraea populations even showed a significant negative trend in seed bank density. Thus, we conclude that Italian populations may have difficulties incorporating seed into the seed bank. Seeds produced by plants in Italian populations may also have a lower germination rate than seeds originating from German populations, supported by results from glasshouse experiments (C. Kluth & H. Bruelheide, unpublished data). However, many other factors, such as carry-over effects, predation, dispersal ability, interactions with the accompanying vegetation or storage conditions in the soil, may affect the seed dormancy cycle or post-dispersal seed loss (e.g. Van Tooren 1988; Cabin et al. 2000; Ehrlén 2000; García et al. 2000; Baskin & Baskin 2001).
The importance of local dispersal processes for biogeographical patterns has recently been pointed out by Hengeveld & Hemerik (2002). For H. petraea, seed loss from Italian populations might be attributed to higher emigration rates associated with higher dispersal mortality than in Germany. A possible cause might be a higher activity of seed feeding and seed dispersing ants, which are more abundant and have denser populations in the Mediterranean region than in central Europe. For example, the harvester ant genus Messor has a circum-Mediterranean distribution (C. Detrain, personal communication), with about 15 species (B. Seifert, personal communication). The only Messor species occurring in Germany, Messor structor (Latreille 1798), occurs only in a few locations and does not exist in our German study region (Seifert 1996). In contrast, in Italy Messor structor is a common harvester ant within the distribution range of H. petraea (L. Gubellini, personal communication). However, the higher seed emigration and mortality from local populations in Italy may, to some extent, be compensated for by an increased number of biotopes occupied by H. petraea (Kluth & Bruelheide 2004).
In addition to the effect of higher emigration rates, the greater variation in the interannual reproduction rate could lead to higher extinction risk for local populations in Italy. In combination with higher colonization rates, this increases interactions between the local populations, i.e. by metapopulation dynamics (Hanski & Gilpin 1991). Therefore, the differences in local population densities between Italian and German populations may be the result of different metapopulation structures. H. petraea populations are very dynamic in the Italian study region (A. Brilli-Cattarini, personal communication), where we found some extremely small populations, some consisting of only a single individual. Such ‘populations’ may represent colonization events. These findings are contrary to the existing theory that assumes an increasing tendency for species to persist as metapopulation towards their range boundary, where habitat is naturally more fragmented. (Hanski 1999; Wilson et al. 2002). So far, such patterns have only been found for relatively well-dispersed animal species, e.g. the butterflies Aricia agestis and A. artaxerxes (Wilson et al. 2002).
Using modelling approaches, Kirkpatrick & Barton (1997) showed that steep environmental gradients in space, combined with increased migration rates and gene flow, inhibit local adaptation and, in consequence, decrease a species’ total population size. Indeed, H. petraea grows in a wider variety of habitats in Italy than in Germany, and the elevational range of populations in Italy is much wider than in Germany (Kluth & Bruelheide 2004). This more heterogenous environment, in combination with higher metapopulation dynamics, could result in local maladaptation in Italian H. petraea populations. Such an interpretation would be in accordance with the larger proportion of variance in fecundity being explained by the population effect, and with a more population-specific response in different years in Germany than in Italy.
In general, the demography of H. petraea in the two studied regions conforms to the observed density patterns. However, these density patterns do not correspond with the ‘abundant centre’ hypothesis. This might have various reasons. We arbitrarily chose two sample regions within the range of H. petraea and studied three successive vegetation periods. The ‘abundant centre’ pattern is expected to be the average over all range parts over a longer time span (Hengeveld & Haeck 1982; Hengeveld 1990; Hengeveld & Hemerik 2002; Holt 2003) and might therefore not be shown in our more limited data. Nevertheless, there are indications that the pattern described here might apply to the whole range of H. petraea. Additional central populations from other parts of the more or less continuous distribution in the Mediterranean region had low densities (Sicily, C. Kluth, personal observation; Andalusia, A. Erfmeier, personal communication; Catalonia, H.-G. Stroh, personal communication), whereas peripheral and disconnected populations consistently showed particularly high densities (Rhineland-Palatinate, Bavaria, Lower Saxony, C. Kluth, personal observation; Öland H. Bruelheide, personal observations). Geographically intermediate populations (Vienna, Trentino South Tyrol, C. Kluth, personal observations) occurred at intermediate densities. The short time span of three study years does not allow conclusions to be drawn regarding the stability of the observed density patterns. Although the literature indicates a continuous existence of the studied populations in Germany for more than a century, their densities might have undergone changes. Under global change scenarios (IPCC 2001) the peripheral German populations might be the ones most favoured by global warming and their higher densities might already be the result of such a climate change. In any case, the dense populations at the periphery would contribute most to range expansion. The general importance of dense populations in range dynamics was shown for relatively dispersible animal species, both by observations (Turin 2000 quoted in Hengeveld & Hemerik 2002) and by modelling (Hengeveld & Hemerik 2002).
The data needed to clarify these relationships between range dynamics, regional metapopulation dynamics and local population dynamics is scarce not only for H. petraea but for species in general. Long-term data sets of population dynamics across distribution ranges are not available, particularly for plant species. There are either long-term studies on single local populations (e.g. Watkinson 1990; Woodward 1997; Bengtsson 2000) or short-term studies on the distribution of abundance across the range (e.g. Prince et al. 1985; Carter & Prince 1985; Carey et al. 1995; Jump & Woodward 2003). Regarding the role of dispersal or metapopulation dynamics, we expect to gain more insight by including ongoing analyses of the genetic structure of the study populations of H. petraea. These data will allow us to test the assumption of higher gene flow in central Italian populations.