1Long-distance dispersal (LDD) is important in plants of dynamic and ephemeral habitats. For plants of dynamic wetland habitats, waterfowl are generally considered to be important LDD vectors. However, in comparison to the internal (endozoochorous) dispersal of terrestrial plants by birds, endozoochorous dispersal of wetland plants by waterfowl has received little attention. We quantified the capacity for endozoochorous dispersal of wetland plants by waterfowl and identified the mechanisms underlying successful dispersal, by comparing the dispersal capacities of a large number of wetland plant species.
2We selected 23 common plant species from dynamic wetland habitats and measured their seed characteristics. We fed seeds of all species to mallards (Anas platyrhynchos), a common and highly omnivorous duck species, and quantified seed gut survival, gut passage speed and subsequent germination. We then used a simple model to calculate seed dispersal distances.
3In total 21 of the 23 species can be dispersed by mallards, with intact seed retrieval and subsequent successful germination of up to 32% of the ingested seeds. The species that pass fastest through the digestive tract of the mallards are retrieved in the greatest numbers (up to 54%) and germinate best (up to 87%). These are the species with the smallest seeds. Seed coat thickness plays only a minor role in determining intact passage through the mallard gut, but determines if ingestion enhances or reduces germination in comparison to control seeds.
4Model calculations estimate that whereas the largest seeds can hardly be dispersed by mallards, most seeds can be dispersed up to 780 km, and the smallest seeds up to 3000 km, by mallards during migration.
5Synthesis. This study demonstrates the mechanism underlying successful endozoochorous dispersal of wetland plant seeds by mallards: small seed size promotes rapid, and hence intact and viable, passage through the mallard gut. Mallards can disperse wetland plant seeds of all but the largest-seeded species successfully in relatively large numbers (up to 32% of ingested seeds) over long distances (up to thousands of kilometres) and are therefore important dispersal vectors.
In plants, dispersal of diaspores is crucial for gene flow and colonization. Dispersal is especially important in plants of dynamic and ephemeral habitats such as shorelines, river and ditch banks, and terrestrialization phases of open water. Accordingly, plants from such dynamic wetland habitats may be expected to have evolved a high capacity for dispersal. Plants from these habitats have several dispersal vectors at their disposal to achieve dispersal on a landscape or even larger scale (long-distance dispersal (LDD)): dispersal by water, by wind and by animals moving over long distances. Dispersal by water transports diaspores to other wet sites, thus dispersing them ‘directionally’ (sensu Howe & Smallwood 1982) to sites with a high probability of being suitable habitat. However, water dispersal is limited to areas connected by surface water to the seed source, and, in moving waters, areas downstream of the seed source (Boedeltje et al. 2003, 2004). Wind dispersal does not have these spatial limitations, as wind transports diaspores to a wide variety of sites in all directions, but dispersal by wind is not directed specifically to suitable habitat (Soons et al. 2005; Soons 2006). Dispersal by waterfowl is intermediate between water and wind, as waterfowl are likely to deposit diaspores in or near water, but may also drop diaspores during flight or other activities in non-wetland areas.
In their reviews, Charalambidou and Santamaria (2002) and Figuerola and Green (2002) conclude that a relatively low number of experimental studies have been carried out on internal (endozoochorous) dispersal of seeds by waterfowl. They call for more research, focussing on comparisons of groups of dispersed plant species, especially to assess if they possess characteristics that favour endozoochorous transport, and on research analysing the different steps of the dispersal process (among others, the distinction between gut survival and subsequent germination). Recently, quantitative studies on the survival of plant seeds ingested by ducks have been carried out for Potamogeton pectinatus (Santamaria 2002; Charalambidou et al. 2005), Ruppia maritima (Figuerola et al. 2002; Charalambidou et al. 2003b) and Sparganium emersum and Sagittaria sagittifolia (Pollux et al. 2005), but each study has been carried out under different conditions so that it is difficult to generalize results. A larger and more comparative study on a wide range of plant species is therefore needed to quantify seed characteristics determining the potential for LDD by waterfowl.
To study the potential contribution of waterfowl to LDD of wetland plants we selected 23 common plants from dynamic wetland habitats, fed their seeds to mallards (Anas platyrhynchos), a common and highly omnivorous duck species (Cramp & Simmons 1977), and quantified their gut survival and subsequent germination. We used this large number of species, representing a wide range of seed traits, to study the mechanisms underlying successful endozoochorous dispersal and identify adaptations to dispersal by waterfowl. Specifically, we tested the following hypotheses:
1Mallards have the potential to disperse wetland plant seeds: seeds survive passage through the digestive tract intact and germinate afterwards when deposited in wetland conditions.
2Intact survival through the digestive tract and subsequent germination are dependent on the retention time in the digestive system of the duck: seeds that are retained shorter have higher survival and germination.
3Intact survival through the digestive tract and subsequent germination are dependent on seed traits of the wetland species:
3a. Species with smaller seeds have higher survival and germination.
3b.Species with a thicker seed coat have higher survival and germination.
4Wetland species with small seeds and thick seed coats are specifically adapted to endozoochorous dispersal and in these species passage through the digestive tract enhances seed germination, while in other species germination remains unaltered or decreases.
We selected 23 common wetland plant species (Table 1) that represent a wide range of seed traits, with seed sizes ranging from 0.4 (Juncus effusus) to 9 mm length (Iris pseudacorus). All species are common in the Netherlands and grow in lake and ditch shorelines and terrestrialization zones in open water. They are typical wetland species and occur in the habitat of many different species of waterfowl, especially mallards. Ripe seeds of all species were collected in autumn during the seed-setting season, from three populations per species. Seeds were air-dried and stored in glass vials in the dark at 4 °C until the start of the feeding and germination experiments in the following spring. We applied this stratification to break seed dormancy and enhance germination, as earlier experiments showed that stratification increased germination in many of our studied species (M. Soons et al. unpublished data). We stratified all seeds used in all experiments to keep results comparable.
Table 1. Overview of the 23 wetland plant species and their seed dispersal characteristics: retrieval percentage, ratio of the germination percentage of the retrieved seeds to the germination percentage of the control seeds, and germination expressed as percentage of ingested seeds. Species nomenclature follows Van der Meijden (2005), with former species names (Van der Meijden 1990) in parentheses
Time interval (h)
Retrieval (% of ingested)
Germination (ratio to control seeds)
Germination (% of ingested)
Ratio significantly different from 1 (P < 0.05).
Ratio not tested (N = 1 for the retrieved or control seeds).
Ratio must be higher than 1, because the ratio of the germination percentage of the ingested seeds to the germination percentage of the control seeds is already higher than one.
To assess if seeds survive the digestive tract of mallards, we fed seeds of each of the 23 wetland species to mallards. We selected mallards as single representative waterfowl species, as differences between duck species in ingestion studies have consistently been small or insignificant (Santamaria 2002; Charalambidou et al. 2003a,b; Pollux et al. 2005). During 8 weeks, 12 captive adult male mallards were each fed 100 seeds of one plant species per week. Each plant species was fed four times during the experiment, resulting in four replicates of 100 seeds per plant species (except for the large seeds of I. pseudacorus, four replicates of 50 seeds, and for Nuphar lutea, four replicates of 20 seeds). The plant species were semi-randomly assigned to the weeks and mallards so that a plant species was never fed more than once per week or more than once per mallard. Prior to feeding seeds were embedded in a bread-dough pellet and the whole pellet was fed, ensuring that all seeds were ingested. Four replicates of a plain bread-dough treatment (no seeds) were also included in the randomisation.
Outside the experiments, the mallards were kept in a free-range on a standard diet (Scharrelkorrel, Havens). After being fed the seed pellets, each mallard was kept in an individual metabolic chamber (0.6 × 0.5 × 0.5 m) with a mesh floor for 48 h. Underneath each metabolic chamber their faeces were collected in plastic trays. The metabolic chambers were located in close proximity to each other so that visual and acoustic contact between the mallards was possible. During the 48 h in the chambers the mallards had ad libitum access to water and standard diet. After 48 h the birds were released into the free-range again.
Faeces were collected at 5, 10 and 48 h after force-feeding of the seeds. Collected faeces were stored in the dark at 4 °C for 3 days. Each faeces sample was then sieved for seeds using a series of soil sieves with decreasing mesh width (ranging from 2 to 0.2 mm) and tap water. All retrieved intact seeds were counted and stored in tap water in glass vials in the dark at 4 °C for 5 weeks until the start of the germination experiment. Seeds of J. effusus, Lythrum salicaria and Typha latifolia were too small to be distinguished visually from the smallest faeces particles. Therefore, faeces samples of J. effusus, L. salicaria, T. latifolia and the plain treatment were sieved only to separate larger material from the smallest particles, after which the sample containing the smallest faeces particles and the seeds was stored in glass bottles in the dark at 4 °C until the start of the germination experiment.
the germination experiment
To assess if seeds are able to germinate after intact passage through the digestive tract, we carried out a germination experiment. The seeds that were retrieved intact from the mallard faeces were removed from their storage and placed in the germination experiment at the same time as control seeds. The control seeds were not fed to mallards but were counted and stored in tap water in glass vials in the dark at 4 °C for 6 weeks until the start of the germination experiment. Four replicates of 20 seeds per plant species (for I. pseudacorus and N. lutea four replicates of 10 seeds) were used for the control treatments.
Retrieved seeds, faeces samples of the small-seeded J. effusus, L. salicaria, T. latifolia and the plain treatments, and control seeds were placed to germinate in a glasshouse. Each replicate was placed to germinate in an individual plastic flower pot filled with sterilized sand. The flower pots were randomly positioned in larger trays filled with water. The water level was equal to the surface level of the sand in the pots, as earlier germination experiments indicated that this was the optimal hydrological condition for the germination of the selected wetland plant species (Soons et al. unpublished data). Trays were covered with transparent foil, with slits for ventilation along the sides, to increase air humidity. Other conditions in the glasshouse were also optimized for germination, with a light : dark regime of 16 : 8 h and 24 °C : 12 °C.
Every week the number of germinated seeds was counted and seedlings were removed. The germination experiment lasted 10 weeks (until germination no longer occurred). No seeds germinated from the plain treatment, as expected, and therefore this treatment was not included in the data analyses.
seed trait measurements
For each species 20 seeds were randomly selected for seed trait measurements. Seed mass was measured for each seed separately on a Mettler Toledo mx45 scale (accuracy 0.001 mg). For three species with very small seeds (J. effusus, Hypericum tetrapterum, and L. salicaria) 10 groups of 20 seeds were measured from which individual seed mass was calculated. Seed length, width and height of each seed were measured using paired callipers (accuracy 0.05 mm). Seed volume was calculated from the seed length, width and height measurements by assuming the closest matching geometrical shape (i.e. ellipsoid, cylinder, pyramid or irregular prism). Seed density was calculated from seed mass and volume. For each species five seeds were randomly selected for seed coat thickness measurements. These seeds were stored in 50% alcohol at 4 °C for 1 week and then sliced with a razor. The thickness of the seed coat was measured on the slices using a dissecting microscope connected to a computer (accuracy 0.01 µm). The thickness was measured at five different locations per slice and then averaged.
From the resulting seed retrieval and germination data, we calculated four seed dispersal characteristics: percentage of seeds retrieved, ratio of the germination percentage of the retrieved seeds to the germination percentage of the control seeds, germination percentage of the ingested seeds (germination expressed as percentage of the number of ingested seeds), and speed of retrieval (percentage of the retrieved seeds that was retrieved in the first time interval). We tested if the ratio of the germination percentage of the retrieved seeds to the germination percentage of the control seeds was different from 1 by using paired t-tests for each time interval of 0–5, 5–10 and 10–48 h (relating each set of retrieved seeds to the corresponding set of control seeds in the germination experiment) and a t-test with test value 1 for the total amounts of germinated seeds per species. We related the seed dispersal characteristics to the measured seed traits by multiple stepwise regression analysis, with the dispersal characteristics as dependent variables and the seed traits as independent variables. To test for relationships among the dependent and among the independent variables we carried out correlation analyses, using Spearman's rank correlation coefficients as not all variables were normally distributed. All percentages were arcsine-transformed before analyses. Seed mass, volume and coat thickness were log10-transformed to attain normality of residuals after regressions.
seed dispersal modelling
Now the most basic parameters are known, models on the (long-distance) dispersal of wetland plant seeds by mallards can be developed. The most basic estimate of the seed dispersal potential of a wetland plant transported by mallards is a relation of the travelling speed of a mallard to the probability of retrieval and successful germination of a seed over time. Assuming the travelling speed of a mallard to be the average observed flight speed, 78 km h−1 (Clausen et al. 2002), we can estimate the maximum seed dispersal potential as follows. For each time interval t of 1 h, we calculated the dispersal distance assuming a unidirectional flight at 78 km h−1. For each t, we then related the distance to the corresponding probability of successful germination of ingested seeds in wetland conditions, calculated by dividing the measured germination percentages in the 0–5, 5–10 and 10–48 h time intervals by the length of these intervals. We carried out these calculations for t = 0–38, as the mallard flight speed cannot be maintained for more than 38 h (Clausen et al. 2002).
Seeds of all species except N. lutea and Rumex hydrolapathum were retrieved from the mallard faeces, with retrieval percentages ranging from 0% to 54% (Table 1). Retrieved seeds of J. effusus, L. salicaria and T. latifolia that could not be sieved and counted due to their small size were also present in the faeces as demonstrated by their germination from the faeces samples in the germination experiment. In all species most seeds were retrieved within the first 10 h after feeding, with no or very few seeds after 10 h (Table 1).
Germination percentages of retrieved seeds ranged from 0% to 78%, with almost always a decrease in germination percentage with increasing retention time in the mallard. In most species these germination percentages did not differ from the germination percentages of the control seeds (Table 1). In the cases of Berula erecta, Epilobium palustre, I. pseudacorus, Lysimachia vulgaris and Phragmites australis, the germination rates of the retrieved seeds were significantly lower than the germination rates of the control seeds, while retrieved seeds of S. erectum had significantly higher germination percentages than control seeds (Table 1). The difference in germination between the retrieved seeds and the control seeds was mostly very small, particularly in the first time interval (0–5 h).
In general, 0–32% of the ingested seeds were retrieved and germinated (Table 1). This proportion decreased with increasing retention time, from 0–27% after 0–5 h to 0–1% after 10–48 h (Table 1). These percentages are relatively modest, which is mainly caused by low retrieval percentages. Still, internal seed transport and subsequent germination were observed for 19 of the 23 plant species.
Nonparametric correlation analyses revealed that retrieval percentage, germination percentage and retrieval speed were highly positively correlated (Table 2). This shows that the seeds that were retrieved in greater numbers also germinated better and proceeded more rapidly through the mallards’ digestive system. Retrieval percentage, germination percentage and retrieval speed were all best explained by seed volume in stepwise multiple regression analyses (Table 4), even while there were strong correlations among seed traits (Table 3). This result shows that smaller seeds are retrieved in greater numbers, germinate better and pass faster through the digestive tract (Fig. 1). Unexpectedly, seed coat thickness had almost no effect on retrieval or germination (Table 4). Seed coat thickness did explain effects of gut passage on the germination of seeds compared to control seeds: in species with thick seed coats, gut passage enhanced germination, while in species with thin seed coats germination was reduced. Seed volume and coat thickness data are given in Appendix S1 in Supplementary material. Overall, these findings identify the mechanism of successful internal dispersal by ducks: the species with the smallest seed volume pass the fastest and in the greatest numbers through the digestive tract and are most viable afterwards.
Table 2. Spearman's correlations for the seed dispersal characteristics show that seed retrieval, germination and retrieval speed are highly correlated to each other
Table 4. Stepwise multiple regression analyses of the seed dispersal characteristics on the seed traits (seed mass, volume, density and coat thickness) show that seed volume is the main determinant of seed retrieval, germination and retrieval speed, while seed coat thickness determines the ratio of germination and plays a minor role in seed retrieval
Significant independent variables (seed traits) in the regression model
Retrieval (% of ingested)
1. Seed volume (2. Seed coat)
Negative relationship (Positive relationship)
R2 change = 0.25 (R2 change = 0.15)
P = 0.024 (P = 0.057)
Germination (ratio of retrieved to control seeds)
1. Seed coat
R2 change = 0.43
P = 0.003
Germination (% of ingested seeds)
1. Seed volume
R2 change = 0.22
P = 0.025
Retrieval speed (% seeds in first time interval)
1. Seed volume
R2 change = 0.37
P = 0.008
Table 3. Spearman's correlations for the seed traits show that seed mass, volume, density and coat thickness are highly correlated to each other
The next step is to use our simple dispersal model to estimate effects on dispersal distributions. As usual in seed shadows, the resulting dispersal distributions are strongly skewed, with a decreasing probability of dispersal over longer distances (Fig. 2). Most (19 out of 23) species show relatively high probabilities of dispersal of ingested seeds over short to landscape scale distances, with 10 species showing a probability of > 0.01 of dispersal up to c. 350 km. Also LDD is highly possible for most of the species. Most species have a probability of > 0.0001 to be dispersed up to c. 780 km, and about half of the species (12 out of 23) have a probability of > 0.001 to be dispersed over such distances. In addition, five of the species can be dispersed in low probabilities over distances > 780 km, up to a maximum of 3000 km. The species that are dispersed with the highest probabilities and over the longest distances are the species with the smaller seeds.
Discussion and conclusions
We assessed the potential for endozoochorous seed dispersal by mallards of 23 common wetland plant species with a wide range of seed characteristics. At least 19 of the 23 plant species can be dispersed endozoochorously by mallards. One species (S. erectum) seems to be especially adapted to endozoochorous dispersal, as it exhibited enhanced germination following transport through the mallard digestive tract. Thirteen species, the majority, had unaltered germination following endozoochorous dispersal by mallards when compared to control seeds, while five species can be dispersed endozoochorously but show reduced germination afterwards. The species that have the highest potential to be dispersed, based on their high (> 25%) germination percentages of ingested seeds, are H. tetrapterum, Lycopus europaeus, Mentha aquatica and Silene flos-cuculi. Only two species (N. lutea and R. hydrolapathum) cannot be dispersed endozoochorously by mallards, as also found for N. lutea by Smits et al. (1989). The percentages of successful germination after internal dispersal found in this study (0–32%) are comparable to, but covering a wider range than, viability percentages found in eight other wetland species (1–16%) by Mueller and van der Valk (2002).
On the dispersal of the remaining two species, P. pectinatus and S. sagittifolia, we are unable to draw firm conclusions. These species were retrieved intact from mallard faeces, as in previous studies (Santamaria et al. 2002; Charalambidou et al. 2005; Pollux et al. 2005), but in contrast to these studies no germination was found. This was probably due to the germination conditions, because the control seeds of these species also did not germinate. Germination conditions were equal for all species and selected to be optimal for the majority of species (see Methods), most of which occur under more variable or somewhat drier conditions than the submerged P. pectinatus and S. sagittifolia. Consequently, the conditions used in the germination experiment may have been too dry for them. Given the germination of these species following internal transport by mallards in previous studies (Santamaria et al. 2002; Charalambidou et al. 2005; Pollux et al. 2005), there is evidence that these species can be successfully dispersed endozoochorously by mallards.
Our study shows that the plant species with the smallest seed sizes have the highest potential to be dispersed by mallards, a result also found for seed lengths (instead of seed volumes) by Mueller and van der Valk (2002). We show that the species with the smallest seeds have the highest and fastest seed retrieval and the highest germination after gut passage, indicating that the mechanism for successful seed dispersal by mallards is that smaller seeds travel through the digestive tract faster, are deposited in larger numbers and are more viable. Seed coat thickness plays only a minor role, and is of importance mostly in one species that appears to be adapted specifically to endozoochorous dispersal, S. erectum. This species has large seeds and slow and relatively low retrieval and germination, but seeds germinate much better after passage through the mallards’ digestive tract than in a control situation, probably due to breaking or removal of the thick seed coat, as also found in Sparganium emersum (Pollux et al. 2005).
Our model calculations estimate how far seeds of the wetland species might be dispersed by mallards in migratory flight. These estimates represent maximum dispersal distances that seeds may reach via endozoochorous transport by mallards, as they assume that the mallards depart directly after seed ingestion and fly continuously in one direction at average flight speed until the seeds are dropped. Such conditions occur only during long migratory flights, the mechanism for reaching maximum dispersal distances via transport by ducks (Figuerola & Green 2002; Green et al. 2002; but see Clausen et al. 2002). These maximum dispersal distances are long. Our calculations show that most species have relatively high probabilities of dispersal of ingested seeds over short to landscape scale distances, while about half of the species have a relatively high probability to be dispersed over distances up to 780 km, and five species have the potential to be dispersed in low probabilities over distances > 780 km, up to a maximum of 3000 km. Our estimate of 3000 km as the upper limit of dispersal distances is likely to be an overestimation, as seeds may not remain in the mallard for the full time interval required to cover such a distance (38 h). Nevertheless, our findings show that mallards have the potential to contribute to the LDD of wetland plants, and have the potential to disperse wetland seeds over longer distances than wind (up to 6 km during very heavy storms, Soons et al. 2004b, or 4 km under more frequently occurring conditions, Soons 2006), fish (up to 27 km, Pollux et al. 2007) and water (e.g. the EU's longest river, the Danube, is 2860 km).
Although ducks are known to be able to cover long distances during flight (56 ringed ducks in Spain travelled a mean distance of 384 km, range of 59–1801 km, within 1 week, Green & Figuerola 2005; one individual ringed teal travelled 1285 km in < 24 h, Clausen et al. 2002), their movement pattern will normally consist of shorter flights. Ducks often fly tens of kilometres between feeding and roosting sites and show high regional mobility, especially during winter, which is also the seed dispersal season (Green et al. 2002). This behaviour will result in significantly shorter dispersal distances than the maximum estimates from our model. These distances may be in any direction, as the movement pattern of ducks is often nomadic rather than directional (Figuerola & Green 2002). The majority of seeds are thus likely to be dispersed over landscape scale distances, and be much more in line with the distances reached by wind and fish. More elaborate and more realistic models including the daily movement patterns of birds, such as the models recently developed by Spiegel and Nathan (2007) and Levey et al. (2008), will allow the calculation of complete dispersal kernels for wetland species. The collection of detailed spatial movement data for mallards will be an important next step towards this goal.
We conclude that, in line with our first hypothesis, mallards have the potential to disperse wetland plant seeds endozoochorously. For 21 out of the 23 common wetland plant species used in this study, seeds survived passage through the digestive tract and germinated afterwards when deposited in wetland conditions. This indicates that a wide variety of wetland plants can be dispersed by mallards. As expected from our second hypothesis, survival through the digestive tract and subsequent germination are dependent on the retention time in the digestive system of the duck: seeds that are retained shorter have higher survival and germination. Ingestion by mallards does not hamper seed germination in most (14 out of 23) species, especially after short retention times. In line with our third hypothesis, survival through the digestive tract and subsequent germination are dependent on seed traits of the wetland species. Species with smaller seeds are dispersed in the largest numbers and germinate most successfully afterwards. Seed volume is the trait that matters most, while seed coat thickness plays only a minor role in gut survival. As stated in our final hypothesis, passage through the mallard digestive tract enhances seed germination compared to control seeds in species with a thick seed coat (but not small seed size). However, our findings show that the ability to be successfully endozoochorously dispersed by mallards is widespread among wetland plants, and not limited to species with thick seed coats.
Overall, endozoochorous transport by mallards appears to be a widespread, successful dispersal mechanism among wetland plants. Mallards can disperse wetland seeds over a wide range of distances, including LDD, with the potential to disperse relatively high numbers of seeds between 0 and 350 km during flight, and the potential to disperse small-seeded species over more than 780 km, up to 3000 km. Mallards are thus potentially (very) long-distance dispersers of wetland plant seeds. This shows how an individual seed trait, seed volume, can have a large effect on the landscape scale and larger scale species dynamics of wetland plants.
This paper was funded under the European Union Sixth Framework Programme integrated Project Euro-limpacs (GOCE-CT-2003-505540). The authors thank Bart Pollux for discussion of the feeding experiment, Anna Traveset for discussion of the seed coat measurements, Mariet Hefting for general discussions and James Bullock, Iris Charalambidou and an anonymous referee for helpful suggestions for improving the manuscript. This is publication 4235 of the Netherlands Institute of Ecology (NIOO-KNAW).