Introduction

Seed dispersal is one of the most important life-history traits that influence plant fitness, because it determines the locations at which seedlings establish (Howe and Smallwood 1982; Nathan and Muller-Landau 2000; Lopez 2001; Wenny 2001; Fenner and Thompson 2005). A variety of dispersal vectors, including wind, mammals, birds, invertebrates, and water, provide plants with many advantages, including (1) escape from specialist herbivores or pathogens near the parent plant (Janzen 1970), (2) colonization of suitable sites whose existence is unpredictable in space and time (Brokaw and Busing 2000), and (3) directed recruitment to “safe sites” where seeds can successfully germinate and seedling establishment is disproportionately high (Howe and Smallwood 1982).

The adaptive significance of each mode of seed dispersal has been explored for a variety of plant species. However, there is increasing evidence that few plant species have specialized to a single dispersal agent. In contrast, the seeds of many plants are dispersed in two or more phases, with different dispersal agents involved in each phase (diplochory; see Vander Wall and Longland 2004). Phase one is the initial mode of seed dispersal, resulting in dispersal away from the parent plant, whereas phase two involves subsequent movement by another agent (Wenny 2001; Vander Wall and Longland 2004). If the two phases enhance recruitment to safe sites, directed dispersal may be more common than previously assumed (Wenny 2001; Vander Wall and Longland 2004; Hampe 2004; Fenner and Thompson 2005). Recently, Wenny (2001) and Vander Wall and Longland (2004) reviewed the adaptive significance of diplochory in several syndromes (wind dispersal and scatter-hoarding animals, bird dispersal and water dispersal, frugivore dispersal and ant dispersal). Little evidence was available for the wind and water dispersal system (but see Schneider and Scharitz 1988; Telenius and Torstensson 1989; Merritt and Wohl 2002).

In riparian zones of temperate and boreal regions, the seeds of most anemochorous (wind dispersed) tree species, such as willow, cottonwood, alder, elm, and ash, have hairs, wings, or wing-like structures, and are able to remain on the water surface, probably because of sinking resistance or surface tension. Seeds disseminated into a river or lake may be secondarily dispersed downstream by water flow (Schneider and Scharitz 1988; Thebaud and Debussche 1991; Anderson et al. 2000; Jansson et al. 2005; Merritt and Wohl 2006). If so, it remains to be determined whether the morphological adaptations are specialized to one dispersal mode or both.

The seeds of willows (Salix species) are light and enveloped by soft cottony hairs in a ring structure. When dry, the cottony hairs spread out so that many seeds cluster together in woolly masses. Because the cottony hairs allow the tiny seeds to gain altitude even in very gentle convective air currents, the hairs are considered morphologically adaptive attachments for anemochory (Niiyama 1990; Karrenberg et al. 2002). Long-distance dispersal of a great number of small seeds may be an adaptive trait for anemochorous pioneer species, increasing the probability of colonization to safe sites when such sites are not predictable in space or time (Howe and Smallwood 1982; Venable and Brown 1988). However, Gage and Cooper (2005) clearly showed that few willow seeds move far from their parent plants; they reported an approximately 90% reduction in seed rain density at more than 200 m from parent plants. Furthermore, tree species in the Salicaceae (such as willow and cottonwood) have much stricter environmental requirements for seed germination and seedling establishment than most other pioneer tree species. Willows usually establish on finely textured substrates, such as sand or silt, with higher water availability, conditions created by large disturbances at both the canopy and soil levels after the removal of litter and vegetation cover by flooding, landslides, fire, or tree falls (Yanai and Kikuzawa 1991; Johnson 2000; Karrenberg et al. 2002; Woods and Cooper 2005). Moreover, willows usually release seeds for a short period of time (2–4 weeks), and the longevity of their seeds is extremely short (less than one month; McLeod and McPherson 1973; Niiyama 1990; Karrenberg et al. 2002; Gage and Cooper 2005), resulting in a transient seed bank (Thompson et al. 1997). These seed traits may not enable willows to bridge unfavorable times and spaces by spreading the risk of death (Venable and Brown 1988; Seiwa 1997; Boedeltje et al. 2004), resulting in a lower probability of colonizing safe sites (see also Johnson 2000). However, if seed release is synchronous with the hydrologic regime, such as snow-melt flow or monsoon-driven flow, the likelihood of seeds encountering suitable habitats increases (e.g., Karrenberg et al. 2002; Meritt and Wohl 2006). Since safe sites are usually species-specific, species whose safe sites are rare in space and time should exhibit more efficient seed dispersal than species with common safe sites (Fenner and Thompson 2005). In this study, we hypothesized that the cottony hairs of willows non-randomly disperse the small and vulnerable seeds to sites well-suited for establishment and growth (directed dispersal hypothesis; see Howe and Smallwood 1982; Wenny 2001; Vander Wall 2004).

To test this hypothesis, we performed laboratory and field experiments to compare several recruitment-related traits, including buoyancy, germination, and trapping at favorable microsites, in seeds of the riparian willows Salix sachalinensis and S. integra with and without cottony hairs. First, we tested whether willow seeds with cottony hairs are selectively trapped at microsites suitable for seedling establishment, as compared to seeds without cottony hairs. Although seeds are able to colonize disturbed sites by wind, willow seeds can be blown to any substrate, such as sites with coarse sediments and low water availability. Thus, seed-trapping experiments were conducted with three typical substrates: one suitable microsite (wet sand) and two unsuitable microsites (water, dry sand). Second, we evaluated whether seeds are released from the cottony hairs and germinate only when they have arrived at suitable microsites, such as those with wet sand. In contrast, if the seeds were released from the cottony hairs when they are on water or dry ground, they would fail to establish. Thus, we investigated the timing of seed release from the cottony hairs and of seed germination on three different substrates: wet sand, water, and dry sand. Third, we evaluated whether the cottony hairs facilitated floating. If the cottony hairs do not facilitate floating, the seeds would be quickly submerged in water and would fail to establish. If the seeds floated, they could be transported downstream to a riverbank or lakeshore by water movement, increasing the probability of recruitment to safe sites, because mesic sites with fine sediments are usually abundant along riverbanks and lakeshores.

Materials and methods

Species

Salix sachalinensis Fr. Schmidt and S. integra Thunb. are deciduous, broad-leaved, dioecious trees that are common in cool temperate regions of Japan. They reach maximum heights of about 20 and 6 m, respectively (Niiyama 1990; Seiwa et al. 2006; Ueno et al. 2007). Salix sachalinensis inhabits riparian forests (Niiyama 1990; Seiwa et al. 2006; Ueno et al. 2007), whereas S. integra inhabits a wide range of mesic sites, from along rivers to steep hill slopes (Seiwa et al. 2006; Tozawa et al. personal observation). Female S. sachalinensis and S. integra disperse seeds during 1–2 week periods in mid-May and early June, respectively. In both species, several seeds are dispersed together, with the cottony hairs forming a single mass. The number of seeds per cluster (capsule) and seed wet mass (mean ± SE) are 4.80 ± 0.22 (n = 45) and 0.127 mg (n = 100) for S. sachalinensis and 4.46 ± 0.20 (n = 45) and 0.079 mg (n = 100) for S. integra, respectively. In riparian forests, both species also show vegetative propagation, although the frequency is very low (Tozawa et al. unpublished data).

Seed germination

We evaluated the most suitable microsites for seed germination for each species in the laboratory. Seed germination was observed in 9.0-cm diameter Petri dishes filled with dry sand, wet sand, or water. For each species, seeds without cottony hairs were placed on the substrate surface, and the number of seeds that germinated was counted daily until no germination was observed for at least 2 days in any of the treatments. The Petri dishes were incubated in an alternating temperature regime with a constant photoperiod (16 h light, 25°C; 8 h dark, 10°C). The percent germination in the three substrates was compared after 7 days. Each treatment contained five replicates of 50 seeds per species. Prior to the experiment, the dry sand was exposed to sunlight for more than 4 days, resulting in very little moisture in the sand, and wet sand was prepared by mixing dry sand with water at a 10:3 ratio, to saturation. The Petri dishes of wet sand were watered frequently during the germination trials.

Seed trapping

To test whether seeds with cottony hairs are selectively trapped at suitable germination microsites, the number of seeds trapped in the three substrates (dry sand, wet sand, and the water surface) was compared under both field and laboratory conditions.

Field experiment

The experiment was conducted on 25, 27, and 30 May 2001 in a riparian forest along the Ikusazawa River, a tributary of the Eai River, in Miyagi Prefecture in northeastern Japan (38°20′ N, 140°45′ E; 400 m above sea level). Petri dishes were filled with one of the three substrates and randomly set on the ground beneath each of three Salix sachalinensis adult individuals. Each day, 15 Petri dishes (five per substrate) were placed at 09:00 and collected at 18:00, after which the number of seeds in each dish was counted. A total of 45 Petri dishes (15 Petri dishes per tree) were used. The dominant species at the site were Salix integra and S. sachalinensis. Salix subfragilis, S. gracilistyla, and S. hukaoana also occurred sporadically in the study area. During the experiment, mainly S. sachalinensis was dispersing seeds.

Laboratory experiment

Plastic trays (12.5 × 12.5 × 3.5 cm) were filled with dry sand, wet sand, or water. Eighteen trays (six per material) were randomly placed in the experimental box (Fig. 1). Approximately 60 one-year-old shoots with mature seeds were collected from five randomly selected S. sachalinensis and S. integra individuals. Shoots were collected from S. sachalinensis on 13 May 2003 from a riparian forest along the Ikusazawa River (location given above) and from S. integra on 23 May 2003 from a riparian forest along the Eai River (38°44´ N, 140°43´ E; 140 m above sea level). On the day following shoot collection, three one-year-old shoots that had opened capsules were set at the entrance (27 × 24 cm) of the experimental box (Fig. 1). The shoots were exposed to wind from an electric fan for 30 min, which disseminated the seeds into the box. This manipulation was repeated ten times for a total of 30 shoots per species. After confirming that no seeds were floating in the air of the box, the fan was turned on for 30 s to distribute the seeds around the box. Then, the fan was turned off for 1 min to allow floating seeds to settle. This procedure (30 s on, 1 min off) was repeated three times. The number of seeds trapped in the water was counted. The number of seeds trapped in both wet and dry sand was estimated by dividing the number of seedlings by the percent germination in that substrate, because the trapped seeds were not clearly visible. In dry sand, the number of seedlings was measured after supplying water. For each species, the experiment was replicated five times and a total of 150 one-year-old shoots were used.

Fig. 1
figure 1

Apparatus used for the laboratory experiment. Plastic trays (12.5 × 12.5 × 3.5 cm) were filled with dry sand, wet sand, or water and randomly placed in the experimental box. For both S. sachalinensis and S. integra, three one-year-old shoots with open capsules were set at the entrance (27 × 24 cm) of the box. The shoots were exposed to wind from an electric fan, disseminating the seeds into the box. The procedure and replicates are described in detail in the text

Full size image

Floating ability

To evaluate to what extent cottony hairs facilitate the floating of seeds, the duration of seeds with and without cottony hairs floating on the water surface was compared in the laboratory. The floating period was defined as the length of time between the placement of seeds on the water surface and the seeds sinking. Furthermore, to investigate how long the cottony hairs prevent seed germination in water, the timing of seed germination in seeds with and without cottony hairs was compared. Seed floating and germination measurements were conducted in 9-cm Petri dishes at 2-h intervals for the first day, at 6–18 h intervals between days 2 and 6, and at 1–3 day intervals between days 7 and 15. The experiments were conducted under an alternating temperature regime with a constant photoperiod (16 h light at 25°C; 8 h dark at 10°C). Each treatment contained five replicates of 50 seeds per species

Seed release and germination on wet sand

To investigate if cottony hairs prevent seeds from germinating at unfavorable microsites by allowing the release of seeds only when they arrive at a suitable microsite, such as wet sand, the timing of germination of seeds with and without cottony hairs was compared after placing the seeds on wet sand in Petri dishes in the laboratory. Each treatment consisted of five replicates of 50 seeds of each species. The length of the experiment, and the experimental conditions (temperature and photoperiod) were identical to those of the floating experiment.

Data analysis

In the field seed-trapping experiment, differences in the number of seeds trapped among the three substrates were analyzed using a repeated measures analysis of variance (ANOVA), with the three investigation dates as repeated measures and the substrate type (nested within tree species) as a main effect. In the laboratory seed-trapping experiment, differences in the number of seeds trapped among the three substrates were analyzed using one-way ANOVA. Comparisons among all pairs were performed using Tukey-Kramer HSD tests when the substrate type had a significant effect. To test for differences in percent germination between seeds with and without cottony hairs in the water and wet-sand germination experiments, multivariate analysis of variance (MANOVA) was used with investigation time as a repeated measure and the presence of cottony hairs as a main effect. Data were ln- or arcsine- (for % values) transformed when necessary to meet the assumptions of normality. All statistical analyses were performed in JMP version 3.1 (SAS Institute Inc., Cary, NC, USA).

Results

Germination of seeds without cottony hairs

In S. sachalinensis and S. integra, the percent germination of seeds without cottony hairs was high in both wet sand (76.0 ± 6.5 and 88.0 ± 2.0%, respectively) and water (84.4 ± 2.2 and 94.8 ± 0.9%, respectively), whereas no seeds germinated in dry sand.

Seed trapping

In the field experiments, more seeds were trapped in both water and wet sand than in dry sand (repeated-measures ANOVA; time, F 2,35 = 39.4, P < 0.01; substrate type, F 2,36 = 475.0, P < 0.01; time × substrate, F 4,72 = 3.7, P < 0.01; Fig. 2). In the laboratory experiments, more seeds were trapped in both water and wet sand than in dry sand for both S. sachalinensis (one-way ANOVA; F 2,87 = 143.0, P < 0.01; Fig. 3a) and S. integra (F 2,87 = 57.5, P < 0.01; Fig. 3b). Most seeds trapped in water had cottony hairs and floated on the surface.

Fig. 2
figure 2

Number of seeds trapped at different microsites (dry sand, wet sand, and water) in the field on 25, 27, and 30 May 2001. The values are means ± SE (n = 15). Solid and open circles indicate seeds with and without cottony hairs, respectively

Full size image
Fig. 3
figure 3

Number of seeds trapped in dry sand, wet sand, and water in laboratory experiments with Salix sachalinensis (a) and S. integra (b)

Full size image

Floating ability and seed germination in water

In both S. sachalinensis and S. integra, all seeds without cottony hairs sank within a few seconds after placement on the water surface. In contrast, 59.4, 17.6, and 2.4% of S. sachalinensis seeds with cottony hairs and 74.0, 68.0, and 62.0% of those of S. integra floated at least 2, 6, and 15 days, respectively (Fig. 4a, b).

Fig. 4
figure 4

Number of seeds of Salix sachalinensis (a) and S. integra (b) with and without cottony hairs that floated on the water surface, and underwater germination rates of seeds of Salix sachalinensis (c) and S. integra (d) with and without cottony hairs. Solid and open circles indicate seeds with and without cottony hairs, respectively

Full size image

In S. sachalinensis, the underwater germination of seeds with cottony hairs was slightly later than that of seeds without cottony hairs (Fig. 4c), although the total seed germination percentage did not differ significantly (MANOVA; seed type, F 1,8 = 7.10, P = 0.03; time, F 20,160 = 1317, P < 0.01; time × seed type, F 20,160 = 9.40, P < 0.01). In S. integra, the germination of seeds with cottony hairs was delayed, and only 36.4% germinated, as compared to the 87.6% germination rate of those without cottony hairs (MANOVA; seed type, F 1,8 = 116.0, P < 0.01; time, F 22,176 = 204, P < 0.01; time × seed type, F 22,176 = 46.0, P < 0.01; Fig. 4d).

Seed release and germination on wet sand

When S. sachalinensis seeds with cottony hairs were placed on wet sand, most seeds were released from the cottony hairs within a few minutes. As a result, seeds both with and without cottony hairs germinated within 2 days; there was little difference in the time to germination between seed types (MANOVA; seed type, F 1,8 = 3.81, P = 0.087; time, F 20,160 = 469, P < 0. 01; seed type × time, F 20,160 = 2.76, P < 0.01; Fig. 5a). In S. integra, seeds with cottony hairs showed delayed germination and a lower germination rate as compared to seeds without cottony hairs (MANOVA; seed type, F 1,8 = 187, P < 0.01; time, F 22,176 = 321, P < 0.01; seed type × time, F 22,176 = 100, P < 0.01; Fig. 5b), because approximately 70% of the seeds were not released from the cottony hairs within 4 days.

Fig. 5
figure 5

Germination on wet sand of seeds of Salix sachalinensis (a) and S. integra (b) with and without cottony hairs. Solid and open circles indicate seeds with and without cottony hairs, respectively

Full size image

Discussion

Role of cottony hairs in secondary dispersal by wind

In field and laboratory seed-trapping experiments, few seeds were trapped in dry sand, but many seeds of both willow species were trapped in water and wet sand. In the laboratory experiments, seeds with cottony hairs rolled over the surface of dry sand and were repeatedly blown up into the air, before landing on either wet sand or water. Once seeds were trapped on water or wet sand, they were not transported to another microsite. These traits, together with the strong water requirement for germination, suggest that the cottony hairs play an important role in preventing the deposition of seeds on unfavorable xeric microsites such as dry sand, and selectively recruit seeds to mesic microsites that are suitable for germination, such as wet sand and water.

In all of the experiments performed, only soil with a small particle size (sand) was used, although great variation in soil particle size is typically found in riparian zones (Nilsson et al. 2002; Shin and Nakamura 2005). Gage and Cooper (2005) clearly showed that moisture is important for trapping willow seeds in terrain without microtopography, but less important in terrain with microtopography. Keddy and Constabell (1986) found that the soil water content drastically decreases with increasing soil particle size, regardless of the water level under field conditions. These results suggest that even though willow seeds colonize coarse sediments (gravel, rocks) in relatively xeric conditions, seeds with cottony hairs may be subsequently dispersed more readily to soils with fine particles and mesic conditions.

Furthermore, the cottony hairs became wet and stuck to the sand immediately after the seeds were placed on wet sand, and the seeds were released from the hairs within a few minutes. As a result, seed germination occurred promptly in both willow species, regardless of the presence of cottony hairs, although the time to release in S. integra seeds was somewhat longer. These results strongly indicate that seeds are released from cottony hairs only when they have reached a suitable germination microsite, such as wet and fine sediments, although only soil with a small particle size (sand) was used in this study. Such non-random secondary dispersal by wind to mesic microsites is advantageous for the establishment of riparian willow seedlings because mesic microsites are very rare in space and time, even in highly disturbed environments subject to flooding and disturbance through erosion and the deposition of fine sediments.

In willows, cottony hairs are considered useful anemochorous attachments that increase the probability of colonization of suitable habitats by randomly dispersing a large number of small and vulnerable seeds far from the mother tree (e.g., van der Pijl 1982; Niiyama 1990; Karrenberg et al. 2002). The results of this experiment clearly reveal an additional and important role of the cottony hairs: the non-random secondary dispersal of seeds by wind to the microsites most favorable for seedling establishment. To confirm the importance of secondary dispersal in nature, it is worthwhile to estimate the probability that the seeds are dispersed into suitable microsites where secondary dispersal comes into play. Further detailed study is needed to estimate this probability.

Role of cottony hairs in secondary dispersal by water

In the floating experiment, approximately 18% of S. sachalinensis and 68% of S. integra seeds with cottony hairs floated on the water surface for at least 6 days. The duration of the floating period was comparable to that of seeds of other riparian tree species adapted to hydrochory (water dispersal) in other regions (Andersson et al. 2000; Lopez 2001). In contrast, most seeds without cottony hairs sank within a few minutes after landing on the water. These traits indicate that the cottony hairs on willow seeds are morphologically or mechanically well adapted to both anemochory and hydrochory.

The abundance and distribution of riparian plants largely depend on the floating abilities of their seeds (long-distance dispersal; Nilsson et al. 1991; Johansson et al. 1996) and the availability of suitable habitats (Danvind and Nilsson 1997; Johanson et al. 2005). Riparian zones are complex and dynamic systems characterized by turbulent and tranquil reaches that create different fluvial environments (Nilsson et al. 2002; Shin and Nakamura 2005). In particular, tranquil reaches or secondary channels have weaker currents that facilitate the deposition of small particles such as sand, silt, clay, and fine-particulate organic matter (Nilsson et al. 2002; Shin and Nakamura 2005). Finer-textured substrates usually favor the regeneration of small-seeded species (Keddy and Constabel 1986; Nilsson et al. 2002), particularly for willows (McLeod and McPherson 1973; Niiyama 1990; Shin and Nakamura 2005). Therefore, the probability of willow seeds reaching safe sites is higher along tranquil reaches of rivers than along turbulent reaches, where floating ability may be diminished due to swift currents (Staniforth and Cavers 1976; Danvind and Nilsson 1997). In riparian zones, if willow seeds with cottony hairs are dispersed into water in areas with weaker currents or in lakes, then recruitment to suitable microsites would be enhanced, because floating seeds are usually transported to riverbanks or lakeshores by waves or running water. As a result, seedlings of willow species are frequently observed along riversides, particularly in areas with weak currents (Robertson and Augspurger 1999; Shin and Nakamura 2005). These traits also suggest that cottony hairs are adaptive for non-random secondary seed dispersal by water to microsites suitable for seedling establishment.

Most seeds without cottony hairs sank into the water, and most of the viable seeds germinated promptly underwater. In most areas with rivers and lakes, seedlings that germinate on the bottom may establish in only a narrow area of extremely shallow water because of the initial short stature of seedlings (less than 3 mm; Seiwa, unpublished data), which leads to the immersion of photosynthetic tissue. In flooded environments, however, germination in water is not always detrimental to seedling establishment. Because the emerging phase that is suitable for establishment is temporally reduced, early germination in water can help maximize the use of the short emerging period, promoting seedling establishment. Even though the seeds are released from the cottony hairs promptly after landing on water, the success of seedling establishment may not be diminished in flooding environments. These results suggest that the roles of cottony hairs are much more complex, particularly in flooded environments, and further experimental study simulating spatially and temporally heterogeneous conditions is needed.

Directed dispersal in diplochorous willows

The two willow species studied exhibit diplochorous seed dispersal with two dispersal phases. The first (wind dispersal) and second (wind or water dispersal) phases of diplochory offer different benefits to willows, as suggested for other diplochorous species (reviewed by Wenny 2001; Vander Wall 2004). Phase one usually facilitates the colonization of habitat patches far from parent plants, whereas phase two moves seeds to discrete and predictable microsites where seedling establishment is disproportionally high. These diplochorous seed-dispersal traits suggest that cottony hairs function in directed dispersal for riparian willows. In non-riparian willow species, which have similar seed morphology (cottony hairs), further comparative studies are needed among the willow species in a wide range of habitats to understand the adaptive significances of cottony hairs in ecological and evolutional contexts.