1We studied the effects of seed dispersal mode and seed mass on the migration patterns of woody and herbaceous forest species in an artificial pine forest band growing on a former sand dune. Seven sites in the artificial forest, at least 44 years old, were selected at different distances from an adjacent natural forest (0.1–17.4 km).
2Both the species richness and the abundance of forest species decreased with increasing distance from the natural forest, indicating that the migration of forest species is limited by seed dispersal. Plants using different seed dispersal modes showed differences in migration rate.
3Ingested and adhesive species migrated into the artificial forest with the most success. In contrast, almost all the species utilizing other dispersal mechanisms (wind, hoarding or no dispersal mechanism) migrated only into sites near to the natural forest. This is likely to be due to low dispersal capacities. Ant-dispersed species were not found at all in the artificial forest.
4Migration distances were calculated for 43 species with a frequency of ≥ 5% in at least one site in the artificial forest. Distances were based on the occurrence of the individual of each species furthest from the natural forest, and on the maximum abundance of that species in the artificial forest. The migration distances of the species did not correlate with their seed mass.
5The dispersal efficiency is an important factor in migration of forest species on a landscape scale, and the migration ability is affected by dispersal mode rather than seed mass.
During the 20th century, poleward and upward shifts of species ranges have occurred in response to climate change across a broad spectrum of taxonomic groups and geographical locations (Walther et al. 2002). For species already threatened by local and global environmental changes, ongoing climate change is likely to increase the risk of extinction, unless species are able to migrate to habitats that become suitable (Grabherr et al. 1994; McCarty 2001) and such problems will be exacerbated by widespread loss and fragmentation of habitats (Walther et al. 2002). It is therefore necessary to examine what determines whether plant species are capable of migrating to newly suitable sites before they become extinct on their original sites.
The mechanism of seed dispersal is most likely to be an important factor in determining the migration ability of a specific plant species (Willson & Traveset 2000). Dispersal limitation is very important in plant dynamics at local and landscape scales (Bullock et al. 2002). Several studies have suggested that the migration patterns of forest species across ancient-recent forest ecotones on a local scale (within several hundred metres) were limited by dispersal (e.g. Matlack 1994; Brunet & von Oheimb 1998; Bossuyt et al. 1999; Brunet et al. 2000). However, it is unclear whether the plant migration pattern on a local scale can be applied on a landscape scale. Willson (1993) suggested that different seed adaptations might influence both the location of the dispersal peak, i.e where the majority of seeds are deposited, generally relatively close to the parental plant, and the slope and shape of the tail, which lead to a few seeds being dispersed over longer distances. It is generally believed that species that disperse their seeds through ingestion or adhesion, mechanisms appropriate for long-distance seed transfer, migrate the most effectively into new habitats (Dzwonko & Loster 1992; Dzwonko 1993; Matlack 1994).
It is, however, difficult to evaluate the migration ability of individual plant species. For example, if we survey migration into a site that already contains rhizomes, bulbs or seeds buried below ground, it is impossible to determine whether the occurrence of a species is due to existing vegetative material or new plants arising from migration. Some studies have found that the vegetation composition of recent forests was dependent more on soil conditions, light or the influence of dominant species than on the modes of species dispersal (e.g. Dzwonko & Gawroński 1994; Dupré & Ehrlén 2002). Therefore, in order to examine the effects of seed dispersal modes and seed size on the migration ability of plant species, it is essential that study sites are selected so as not to contain propagules and that the study sites provide homogenous habitat quality.
Our study sites are in an area of artificial pine forest that prior to plantation on dunes with a uniform sandy soil, did not contain any propagules of forest plants. Therefore, the occurrence of forest species in this site is entirely controlled by the efficiency of dispersal of their seeds or other propagules, or by clonal growth.
The aim was to examine the effects of seed dispersal on the migration of woody and herbaceous forest species on a landscape scale. We initially determined whether the number and abundance of forest species decrease with increasing distance from their source. Where this was the case, we asked whether migration ability of forest species varied with seed dispersal mode and/or seed mass.
Materials and methods
The study was conducted in a band of artificial pine (Pinus thunbergii) forest situated in the flat and low lying (10 m a.s.l.) region of Niigata Prefecture, central Japan (Fig. 1). This forest was established on a dune along the shoreline to protect residential and agricultural areas from shifting sand, has been unmanaged since planting and is typically about 200 m wide (ranging from 50 to 400 m). For the most part, it is wide enough to provide inside environments free of edge effects. Recent development has led to narrowing and fragmentation in some parts.
The west end of the artificial forest band is connected to Mt Kakuda (482 m a.s.l.), the location of the only natural forest within a radius of at least 20 km. Consequently, it is likely that the mountain area would be the migration source for all forest plants.
The forest band is not continuous and landscape elements, especially the 40-m-wide channel at 10.5 km from the natural forest, could hamper migration of short-distance dispersal species and movement of seed dispersers across the terrain.
The majority of the mountain is covered with natural deciduous forests dominated by Acer mono var. glabrum, Prunus jamasakura, P. verecunda and Quercus serrata. On the forest floor, the woody species Rhus ambigua, Viburnum dilatatum and Zanthoxylum piperitum, and the herbaceous species Aster scaber, Carex siderosticta, Disporum smilacinum, Epimedium sempervirens, Hepatica nobilis var. japonica f. magna, Liriope platyphylla and Viola grypoceras are abundant. Although some parts of the deciduous forests were coppiced before the 1940s, they have since been unmanaged. Soils are brown forest soils that are moderately moist (Shidei 1974).
Seven sites were studied in the artificial forest, at 0.1 km, 0.9 km, 3.3 km, 5.0 km, 9.5 km, 12.3 km and 17.4 km, respectively, from the natural forest (Fig. 1). The forest in each site was at least 44 years old, and healthy, closed stands were selected in order to exclude gap effects due to damage caused by pine wilt disease. We also selected a site in the adjacent natural deciduous forest, close to the border with the artificial forest.
A 10 × 125 m study plot was established in the interior area of each site, at least 50 m away from the forest edge. Each study plot was divided into 5 × 5 m zones, and we then set four 1 × 1 m quadrats regularly in each of these zone. We studied a total of 1600 quadrats (4 in each zone, 50 zones in each of seven sites in artificial forest and one in natural forest).
We recorded the species of all vascular plants less than 2 m tall occurring in each 1 × 1 m quadrat between June and September 2001 in the artificial forest and in September 2002 in the natural forest.
Based on the flora of the natural forest (Ikeno & Shirosaki 1976), typical forest interior species rarely found outside forests were identified with the aid of Kitamura & Murata (1974a,b), Kitamura et al. (1974a), Kitamura et al. (1974b) and Kitamura & Murata (1979), who provide a detailed ecological characterization of species. We excluded alien species, dune plant species, species that frequently occur in open fields or only on damp grounds, and species used as garden plants. Of the species found in the natural forest, 495 were classified as ‘typical forest species’.
We tested whether species richness and the total abundance of forest species decrease with distance from the natural forest by Pearson correlation analysis. Species richness was calculated as the total number of forest species in the quadrats within a site. Total abundance was calculated at a given site by summing the occurrences (presence in quadrat) of all forest species in the 200 quadrats.
All typical forest plant species were classified into one of eight dispersal types: ballistic (autochorous), wind (anemochorous), non-dispersed (barochorous), dispersed by food-hoarding animals (dyszoochorous), ingested (endozoochorous), adhesive (epizoochorous), ants (myrmechorous) and vegetative propagation. The wind-dispersed species were additionally divided into species with (i) spores, (ii) plumed seeds and (iii) winged seeds, as in the study of Dzwonko & Loster (1992). The type of dispersal was determined for each species on the basis of a direct inspection of propagules and available literature (Numata & Yoshizawa 1978; Asano & Kuwabara 1990). Species with multiple dispersal mechanisms were assigned to the category most likely to provide long-distance movement. Species that utilize both ant and ballistic dispersal, such as violet, were counted as ant dispersed.
We tested whether the mechanism of seed dispersal used by a plant affected the ability of the species to migrate into the artificial forest. This was done by determining the proportion of the total number of species with a given dispersal mechanism, and comparing the numbers between the natural and artificial forests using Fisher's exact test. In order to compare migration patterns for plants that utilize different dispersal mechanisms, trendlines for the relationship between the distance from the natural forest and both species richness and total abundance were fitted with SPSS 11.0 (SPSS 2001). We then calculated the distance from the natural forest where we would expect the number of plants of a ‘typical’ species using this dispersal mode to be decreased by half of the natural forest site.
To examine the effects of seed mass on migration of the forest plant species, the weight of the seeds of each species with a frequency of ≥ 5% in at least one site in the artificial forest was estimated. We examined 43 species in these analyses, excluding pteridophytes that propagate using spores. Detailed characterization of all but three species was available from Nakayama et al. (2000). For Acanthopanax spinosus and Lonicera morrowii, we weighed 100 seeds sampled from the study area, and for Pyrola japonica we employed data from the closely related species P. incarnate, based on Tateda & Isikawa (1968).
We measured the distances from the natural forest to the most distant site in the artificial forest that contained each species and to the site with the maximum abundance of these 43 species. The relationships between log-transformed seed mass and migration distance from the natural forest were examined by Pearson correlation analysis.
All statistical analyses were performed with the SPSS 11.0 (SPSS 2001).
The species richness of forest plants and total abundance in the artificial forest decreased significantly with increasing distance from the natural forest (Fig. 2).
Of the 495 species designated as ‘typical forest plants’ in the natural forest in our survey, 75 species (15%) were also encountered in the artificial forest (Table 1). The artificial forest was impoverished in species with no obvious vector, spore- and ant-dispersed species (Table 1). Ant-dispersed species were never found in the artificial forest. Plants using ballistic dispersal and clonal species were also not found in the artificial forest, but these were also infrequent in the natural forest. By contrast, the proportion of ingested species in the artificial forest was significantly higher than found in the natural forest (Table 1). In addition, the proportion of wind-dispersed species with plumed seeds, adhesive species and species dispersed by food-hoarding animals were somewhat higher in the artificial than in the natural forest, but these differences were not statistically significant (Table 1).
Table 1. The number of forest plant species in different dispersal modes (and their proportions) in natural and artificial forest. Numbers between natural and artificial forest were compared by Fisher's exact test
Natural forest (n = 495)
Artificial forest (n = 75)
Categories more frequent in the natural forest
Wind (winged seeds)
Categories more frequent in the artificial forest
Wind (plumed seeds)
The number of plant species utilizing a particular dispersal mode was negatively correlated with the distance from the natural forest (Fig. 3a–f). The estimated distances to points with half the species richness of the natural forest site for ingested, adhesive species and wind-dispersed species with plumed seeds were 32.8 km, 27.2 km and 6.4 km, respectively (Fig. 3a–c), whereas those for species dispersed by food-hoarding animals, wind (with wings) or with no obvious vector were no more than 1 km (Fig. 3d–f).
For total abundance of forest plants using a given dispersal mode, the pattern with the distance from the natural forest resembled that of species richness, with the exception of adhesive species (Fig. 3g–l).
There was no significant relationship between seed mass and migration distance based on either the farthest individual (Pearson r = −0.10; P = 0.516; n = 43) or maximum abundance (Pearson r = −0.05; P = 0.769; n = 43).
The artificial forest band was interrupted by some narrow paths, the channel and the other non-forest areas. Among the short-distance dispersal categories (hoarded, winged or no vector) only one species (a wind-dispersed species with winged seeds that occurred in all the sites) reached any of these barriers and their effect can be discounted.
plant distributional patterns on a landscape scale
The number and the abundance of forest species decreased with increasing distance from the natural forest (Fig. 2). Our results suggest that migration on a landscape scale is highly limited by seed dispersal. Previous studies of species distribution patterns on a local scale have found that species distribution decreased with distance from the seed source (Matlack 1994; Brunet & von Oheimb 1998; Bossuyt et al. 1999). Our results suggest that plant distributional patterns on a landscape scale are similar to those on a local scale, and that migration is limited by similar factors over any distance.
dispersal mode and migration
Matlack (1994) surveyed the migration of forest plant species across ecotones between old regrowth stands and successional stands, and reported that significant differences in rate of migration were observed among seed dispersal modes: ingested > adhesive > > wind ≥ ants ≥ none. Our results on a landscape scale are consistent with local scale findings (Matlack 1994), but there are more marked differences in migration distance from the source between species using different dispersal modes.
The ingested and adhesive species were relatively successful at migrating into the artificial forest. These results are consistent with those of previous migration studies for forest species in Europe and America (Dzwonko & Loster 1992; Dzwonko 1993; Matlack 1994). Species that disperse their seeds through animal ingestion were the most common of all dispersal modes in the artificial forest (Table 1), and most of them migrated far from the source population (Fig. 3a,g). The spread of ingested species is strongly influenced by frugivorous birds, especially Hypsipetes amaurotis (brown-eared bulbul), which are the main seed dispersers of fleshy-fruited plant species in the study area (Takanose & Kamitani 2003). Fukui (1995) estimated that H. amaurotis transported seeds at least 300 m by surveying the birds’ home ranges, as well as the distribution of tree seedlings. Many studies have confirmed that frugivorous birds are capable of dispersing seeds within a radius of several hundred metres from a fruiting tree (Howe 1977; Murray 1988; Sun et al. 1997). Migratory birds, which are not limited to a home range, can carry seeds over long distances (Cruden 1966) and a large number of frugivorous birds migrate along the coastal forest (Takanose & Kamitani 2003). Therefore, plant species that disperse their seeds through ingestion can colonize the new forest very effectively.
Species that disperse their seeds through adhesion were also successful in migrating into the artificial forest (Fig. 3b,h). Adhesive fruits are transported until they fall off naturally or are groomed off (Agnew & Flux 1970). Consequently, they have the potential to travel over long distances.
By contrast, the artificial forest was impoverished in the species with no obvious vector, spore- and ant-dispersed species, when compared with the natural forest (Table 1). Almost all these species are herbaceous plants. Forest herbs often have short dispersal distances, usually less than a few metres (e.g. Beattie & Lyons 1975; Hughes et al. 1988; Cain et al. 1998), and their establishment is limited by environmental conditions, such as former land use, soil condition and light (Dzwonko & Gawroński 1994; Honnay et al. 1999; Verheyen & Hermy 2001), so that both dispersal and establishment may limit expansion into this artificial forest.
In contrast to previous suggestions (e.g. Willson & Traveset 2000), the wind-dispersed species recorded in our study sites showed relatively slow migration. This might be because few seeds of wind-dispersed species could be dispersed over long distances (Klinkhamer et al. 1988). Furthermore, winds within dense forest canopies are relatively weak when compared with those in open landscapes (Nathan et al. 2002).
seed mass and migration
It has previously been suggested that small-seeded species are more efficient dispersers than larger seeded species with the same dispersal mode (Howe & Kerckhove 1980; Sorensen 1986; Augspurger & Franson 1987; Greene & Johnson 1993). Ehrlén & Eriksson (2000) examined the relationship between seed mass and occurrence of herbaceous species with ingested, ballistic or no dispersal mode in forest patches, and suggested that the distribution of large-seeded species was more limited by dispersal than the distribution of species with smaller seeds. Our analysis of the relationship between seed mass and migration distance across five dispersal modes (ingested, adhesive, wind, hoarding and none), however, shows no correlation. We conclude that migration in our study area is affected by dispersal mode rather than seed mass.
We are grateful to Dr P. Klinkhamer, Dr L. Haddon and two anonymous referees for their valuable comments and suggestions on an earlier draft of the manuscript. We also thank the members of the Laboratory of Silviculture, Niigata University, for their valuable discussions. This study was partly supported by JSPS KAKENHI (14560118 & 15310162).