Propagule deposition along river margins: linking hydrology and ecology


Correspondence author. E-mail:


  • 1Research has highlighted the importance of buoyancy for hydrochorous propagule dispersal, but recently the river bed has been identified as a significant store of viable propagules.
  • 2Over four consecutive 4-month periods, deposited propagules (predominantly seeds) and sediment were sampled at 78 river bed, bank face and bank top sites within three river reaches in two catchments. Species of deposited propagules were compared with the propagule bank and standing vegetation to identify ‘new’ species.
  • 3Forty-four percent of deposited propagule species were not present in the vegetation and the largest proportions were deposited in winter. New propagules showed a higher proportion of species that establish long-term seed banks and produce light seeds, a higher proportion of R-strategists, and a lower proportion of C-strategists than species in the vegetation.
  • 4Ordination of propagule species abundance data revealed differences in samples between reaches; bed and bank locations; and sampling periods. Within depositional samples there were significant positive correlations between average C scores and the proportion of species producing relatively heavy, short-lived seeds and scores on an ordination axis describing a gradient from channel bed to bank top.
  • 5Correlations and multiple regression models between species richness and abundance of deposited propagules (total and new), and the quantity and calibre of river-deposited sediment, demonstrated a direct link between propagule deposition and hydraulic conditions during inundation.
  • 6Synthesis. Riparian connectivity of river flows, varying hydraulic conditions and temporary storage of propagules are all complex components of hydrochorous propagule dispersal in the study reaches. Flood flows can transfer sediment particles and non-buoyant, viable propagules from the river into the riparian zone. Propagules can pass in and out of storage within the channel bed and margins through a variety of dispersal processes and can then settle out of the water in suitable depositional environments generating distinct spatial and temporal patterns in total propagule deposition. New propagules show greatest floristic similarity to propagule bank and bed samples. They are preferentially smaller, lighter and form longer term seed banks, making them particularly suitable for long distance transport over prolonged time periods and for widespread dispersal across river margins.


Plant propagules are dispersed along river corridors through a variety of natural mechanisms, including direct deposition from the parent plant and transport by wind, water and animals (Goodson et al. 2001; Fenner & Thompson 2005; Pollux et al. 2005). Dispersal often involves two or more of these mechanisms with water (hydrochory) frequently complementing other methods of dispersal (e.g. Hampe 2004; Jansson et al. 2005). Hydrochory has been shown to play a major role in transporting and depositing freshly-produced seeds along river corridors (Thebauld & Debussche 1991; Cellot et al. 1998; Merritt & Wohl 2002; Boedeltje et al. 2003a; Vogt et al. 2004; Truscott et al. 2006), in remobilizing seeds after their initial deposition (Pettit & Froend 2001; Goodson et al. 2003), and in structuring riparian plant communities (Nilsson et al. 1991; Johansson et al. 1996; Andersson et al. 2000; Goodson et al. 2002).

The effectiveness of hydrochory varies through time (Andersson & Nilsson 2002) with periods of propagule release and flooding impacting on the quantity of propagules in transport (Schneider & Sharitz 1986; Kubitzki & Ziburski 1994; Boedeltje et al. 2004). Large-scale, often seasonal, flood events are particularly instrumental in causing both physical and biological changes in riparian environments (Junk et al. 1989). Smaller flow pulses can also be important for the near-continuous hydrochorous dispersal of materials within the riparian zone (Tockner et al. 1999). Although large flood events may transport seeds particularly long distances, many plant species produce seeds that can remain buoyant for prolonged periods, owing to their low density, impermeable or air-filled seed casing or appendages (Murray 1986) and so can be transported significant distances even when river levels and flow velocities are low (Danvind & Nilsson 1997; Nilsson et al. 2002; Boedeltje et al. 2004).

Recently, the overriding importance of hydrochory for propagule dispersal and colonization of the riparian zone has been demonstrated for a newly cut river channel with initially bare river banks (Gurnell et al. 2006). Moreover, in a detailed analysis of sediment samples taken from the river beds of the relatively undisturbed study sites discussed in this paper, Gurnell et al. (2007) showed that numerous viable but non-buoyant propagules were stored on the river bed. The propagules accumulated preferentially in particular bed habitats, and the species composition of these river bed stores changed through time. These combined research results suggest that propagules have the potential to move between storage sites during hydrological events and that an integrated approach to investigating reach-scale propagule dynamics is needed, incorporating local vegetation and propagule bank sources in both the riparian zone and river channel, and recognizing the potential of longer distance propagule dispersal.

The present paper adopts such an integrated approach and applies it across three different river reaches in order to establish the species abundance of plant propagules deposited within river margins (banks and bed); to identify the degree to which these reflect species present in the local vegetation and propagule bank, and so assess the degree to which observed depositional patterns along well vegetated reaches may be explained by local or longer distance dispersal processes, particularly fluvial processes. It should be noted that although we use the term propagule throughout this paper, the methods used largely censused seeds. However, it is possible that small vegetative propagules may also have been included in our analyses.

In particular, this paper addresses the following research questions:

  • 1What is the species abundance of viable plant propagules deposited across river margins (bed and banks) and does this vary in space (between- and within-river reaches) and through time (seasonally)?
  • 2Do the deposited species differ from those already present in the vegetation and pre-existing soil propagule bank? In other words, are ‘new species’ being deposited?
  • 3Are there characteristic spatio-temporal patterns in the species abundance of new species and do these patterns differ from those for all deposited species?
  • 4To what extent can these spatio-temporal patterns of deposition be attributed to fluvial rather than other processes?


study reaches

Three perennially flowing river reaches on two rivers (Fig. 1) were selected for study. Two of the reaches (F1 and F2) were located 2 km apart on the River Frome, Dorset, UK to investigate differences in propagule dynamics between reaches located in close proximity on the same river. The third reach (TN) was located c. 250 km from the other two reaches in the headwaters of the River Tern, Shropshire, UK to investigate differences in concurrent propagule dynamics between reaches on different river systems. The reaches ranged in length from 115 (TN) to 195 m (F2), average bank-full width from 11.8 (F2) to 6.3 m (TN), and slope from 0.0045 (F1) to 0.0009 mm−1 (F2 and TN). All reaches were fringed by riparian trees and ungrazed herbaceous ground cover and had c. 1 m high banks exposed above low-flow water levels. Key differences between sites were the chalk bedrock of the Frome in comparison with the sandstone of the Tern; the higher sinuosity and morphological complexity of both channel bed and margins at F1 in comparison with the straight channels of F2 and TN; and the steep, morphologically simple bank profiles at F2 in comparison with the more irregular bank profiles of F1 and TN.

Figure 1.

Location and instrumentation of the study sites.

field methods

All three reaches were instrumented in a similar manner (Fig. 1). Water levels along each reach were monitored using three pressure transducers located at the approximate midpoint and towards the upstream and downstream ends of each reach. These instruments recorded river stage at 15 min intervals (mean of 15 1-min observations) from June 2003 to October 2004 (e.g. Fig. 2). The positions of the transducers and all sampling locations were accurately surveyed using a Leica Total Station. Linear interpolations of recorded water levels between the pressure transducer sites permitted estimation of river level time series at intervening sampling locations. From these interpolated time series, the percentage time that sampling sites were inundated (%Inund) and their elevation relative to the local mean water level (EMWL) could be estimated for any time period within the study.

Figure 2.

Water level variations at the three study sites during the four sampling periods (note that each water level record is relative to an arbitrary local datum within each reach).

Soil propagule bank samples were collected during May 2003. May is the time at which most field germination has taken place and so it is the best time to sample the persistent seed bank. Different methods were used to sample river bank and river bed propagule banks. Soil cores were obtained from two zones of the river bank, the top of the bank (‘Top’– T) and the central area of the bank face (‘Mid’– M) using a 5-cm deep × 6-cm diameter corer, with five replicate cores taken at each location and bulked together. Within each of the study reaches, 16 bulked samples were obtained, respectively, from different locations along the top of the banks and also along the bank faces. Propagule bank samples were also obtained from 16 locations on the river bed of each reach (‘Channel’–‘C’) using an Aston Sampler with a 26-cm diameter sampling area and modified to sample in water up to 1 m deep. The river bed was vigorously disturbed within the sampler to a depth of 5 cm, and samples of bed material were collected in a 100-µm mesh net. The small mesh size allowed the full range of seed sizes (typically > 212 µm, Thompson et al. 1997) to be collected. Each sampling location was marked by a post on the nearest bank top and was also fixed by detailed topographic survey to permit repeat sampling of bed surface material at the same locations for the remainder of the study.

Following the initial propagule bank sample collection, deposition of propagules was monitored along the three reaches between June 2003 and October 2004 over four 4-monthly periods: period 1 – June 2003 to September 2003; period 2 – October 2003 to January 2004; period 3 – February 2004 to May 2004; and period 4 – June 2004 to October 2004. These 4-monthly sampling periods were chosen to correspond with phases in the annual flow hydrograph and in the influence of seasonal growth of in-channel and emergent vegetation on river levels (e.g. Gurnell & Midgley 1994; Gurnell et al. 2006). Different methods were used to sample the river banks and river bed. Deposition of propagules and sediment along bank Top and Mid areas were sampled using 18 pairs of artificial turf mats with 1.5 cm bristles at a density of 8 bristles per cm2. Each mat pair comprised one 36 × 24 cm mat for sampling viable propagules and one 18 × 12 cm mat for sampling the properties of sediment deposited at the same location. Because of the ungrazed, well-vegetated nature of the river banks in all three reaches, the only significant source of sediment was from the water column during bank inundation. The mats were secured using metal pins and were distributed widely along each reach to provide good spatial sampling of deposited material. However, the steep, frequently undercut banks at F2 restricted suitable Mid sampling sites to the downstream end of the reach (Fig. 1). At the end of each sampling period all of the mat pairs were retrieved and new mat pairs were installed in the same locations. At the same time, the 16 Channel sampling sites in each reach were carefully re-located and superficial material that had accumulated on the bed since the previous collections was sampled using the Aston sampler. These sampling methods on both bed and banks are susceptible to local reworking of surrounding sediments. Such local reworking is part of propagule dynamics and is likely to affect the bed samples the most, but could also occur during inundation of the bank deposition sampling sites. In either case, the fine superficial deposits that were sampled are highly susceptible to remobilization by flow pulses and we have based our interpretations of temporal dynamics on contrasts in species composition between samples obtained from the same sites in different sampling periods.

Between 27 May and 3 June 2004 a vegetation survey was conducted along all three reaches. Each reach was subdivided into at least three subreaches. Within each the two river banks were searched independently and species abundance was recorded using the semiquantitative Domin cover scale (Kent & Coker 1992), with separate records obtained for the bank toe and channel (C), the bank face (M) and the bank top within 5 m of the bank face (T).

laboratory methods

Different methods were used to determine species abundance of viable propagules (all T, M and C propagule bank and depositional samples) and sediment characteristics (all T and M depositional samples).

Following collection from the field, the samples that had been obtained for determining species abundance of viable propagules were placed in individual sealed black plastic bags and were stored at 4 °C until they were processed.

The aggregated Top and Mid propagule bank samples were condensed by wet-sieving through a 212-µm mesh sieve. The Channel propagule bank and depositional samples were also sieved to remove excess fine silts and clays. All of these samples were then spread to an even depth (< 1 cm) in individual seedling trays previously filled with 3–4 cm sterile soil (John Innes #2).

The Top and Mid depositional (36 × 24 cm) mat samples were punctured and placed in seedling trays previously filled with water-soaked vermiculite, with further vermiculite added around the edges of each mat to ensure moisture could pass up and across the mat from below. Mats with little deposited sediment were augmented with additional sterile soil to prevent desiccation and maintain water percolation across the entire mat.

All propagule bank and depositional samples were then subjected to a 10-week greenhouse emergence trial. The seedling trays were arranged randomly in the greenhouse, watered twice daily using an automated pipe-feed system, with 16 h daylight, and were provided with controlled temperatures between 15 and 22 °C. As seeds germinated, the seedlings were identified to species and then removed. From these data the number of species and density of viable propagules per m2 sampling area were estimated for each sample. Based on information from Stace (1999), Hubbard (1992) and Rose (1981), each identified species was allocated to a group according to its general habitat requirements: Aquatic, Wetland and Terrestrial. Using information from Hodgson et al. (1995), the flowering time, plant functional type (CSR), propagule weight and seed bank type were identified for as many of the recorded species as possible. A score between 0 and 1 was allocated to C (competitor), S (stress-tolerator) and R (ruderal) components of the CSR functional type for each species (Hunt et al. 2004) to allow quantitative comparisons between samples.

The second (18 × 12 cm) depositional mat sample from each mat pair was used to determine characteristics of deposited sediment. Each mat was wet-weighed, dried at 50–60 °C and re-weighed to determine the dry weight of sediment that had been deposited. The sediment was then removed and processed to determine the weight of organic matter (WtOrg in kg m−2) and mineral sediment (WtSed in kg m−2) by loss on ignition. The particle size of the mineral sediment was determined by dry-sieving to 1 mm and then by laser-sizing to 0.2 µm and the weight of fine sediment (silt and clay, < 64 µm) was then determined (WtFines in kg m−2) as an index of mineral sediment calibre. Low river flows during the first sampling period (period 1, June to September 2003, Fig. 2) resulted in insufficient sediment deposition on the mats for analysis. For the remaining three sampling periods (periods 2–4), sufficiently large sediment samples were collected for analysis from the majority of mats.

data analysis

At the reach scale, vegetation and viable propagule data were presented graphically to illustrate contrasts in species richness and propagule abundance between reaches and sampling periods, and also contrasts in the richness of terrestrial, wetland and aquatic species, functional types and propagule traits (longevity, weight) of species present within the standing vegetation, propagule bank and depositional samples. In addition, the same characteristics were investigated for ‘new species’ that were deposited but were not present in the reach vegetation, and also for species that were not in the vegetation or soil propagule bank. Differences in the quantity and properties of all propagules in comparison with new propagules were used to isolate the impact of propagule dispersal into, rather than simply within, the study reaches, and were also used to investigate the possible relative influences of vegetation and propagule bank sources on the species composition of deposited propagules. The statistical significance of differences in propagule quantities or properties was tested using either χ2 tests or Kruskal–Wallis analysis of variance, as appropriate. Where Kruskal–Wallis anova was applied, it was followed by pairwise comparisons using Dunn's procedure with a Bonferroni-corrected significance threshold of P = 0.05. These analyses were performed using xlstat Pro-7.5.

Within reaches, similarities in the species composition of vegetation, propagule bank and depositional samples at different elevations (T, M and C) within different sampling periods (1–4) were identified using agglomerative hierarchical cluster analysis based on Sorensen's similarity index and agglomeration by weighted pair-group averages. The analysis was performed using xlstat Pro-7.5. This analysis identifies broad groupings according to similarity in species composition and so provides insights into potential links between vegetation, seed bank and depositional species composition at different locations within the river margins over the four study periods. Gradients in average species abundance of all and only new species within the propagule bank and depositional samples obtained at different elevations (T, M and C) and depositional samples obtained during different time periods (1–4) were extracted using Detrended Correspondence Analysis (DCA). In all cases propagule abundance was log-transformed, no species were down-weighted, no samples or species were excluded, detrending was by segments and the analyses were performed using canoco v4.5 (Ter Braak & Smilauer 2002). Differences in the spatial and temporal structure of propagule bank and propagule deposition of all species in comparison with species not present in the local vegetation can be identified from DCA plots and provide a basis for inferring impacts of within reach (vegetation, local remobilization of propagules by water and wind) in comparison with longer distance controls on the species abundance of deposited propagules. Significant correlations between sample scores on the DCA axes and the traits of propagules within the samples supported interpretation of the ordinations.

At the patch scale, the species richness and propagule abundance of both all and new species found in individual samples deposited on mats installed on river banks (T and M samples only) were correlated with fluvial variables (%Inund, EMWL, WtOrg, WtSed, WtFines) using xlstat Pro-7.5. These analyses established whether there was any significant fluvial control on propagule deposition. EMWL indicates the position of each mat relative to all other mats and also to the mean river surface level in each sampling period. It is also indicative of mat location relative to the local gradient in vegetation composition up the river banks. %Inund directly indicates the degree to which each mat has been inundated during a sampling period. Since inundation may not be associated with sufficiently low flow velocities and shear stresses for particle deposition, particularly of low-density organic matter and small mineral particles, three direct indicators of the depositional environment were also investigated. Within the well-vegetated, ungrazed banks of the study reaches, the only significant source of mineral sediment for bank deposition is the river. Therefore, mineral sediment characteristics, particularly the total quantity of mineral sediment deposited (WtSed) and the quantity of fine, easily transported, sediment deposited (WtFines) are direct measures of the degree to which particles had been able to settle out of the water column during bank inundation. Although organic matter can be derived locally as well as from the river, the quantity of organic matter deposited (WtOrg) is important because it contains plant propagules, and any similarities in the behaviour of the organic and mineral components of the deposited sediment indicate common (fluvial) deposition.

Multiple regression models were then estimated between species richness and log-transformed propagule abundance of the depositional samples (dependent variables) and the five fluvial variables (independent variables) using MINITAB 14. The regression model with the highest coefficient of determination (R2 adjusted for degree of freedom), which included regression coefficients that were all significantly different from zero (P < 0.05), and a maximum of one river level (%Inund, EMWL) and one sediment (WtOrg, WtSed, WtFines) variable to maintain independence of the explanatory variables, was selected to describe the dependence of the deposited propagules on fluvial processes.


reach scale

In total 104 818 viable propagules of 172 species were deposited across the banks and bed of all reaches during the study, and 153 and 116 species were identified, respectively, in the vegetation and propagule bank. χ2 goodness-of-fit tests tested the null hypothesis that the species richness of deposited samples was the same in all sampling periods within each reach. There was a significant seasonal pattern in the species richness of deposited propagules at reaches F1 (χ2 = 23.4, d.f. = 3, P < 0.01) and F2 (χ2 = 10.7, d.f. = 3, P < 0.02) with most species deposited in period 2 (October to January) followed by period 3 (February to May) (Fig. 3). Although a similar seasonal pattern in species richness also occurred at TN, it was not statistically significant (χ2 = 2.03, d.f. = 3, P > 0.05). The abundance of viable propagules was very variable between reaches and time periods with particularly high abundances at F1 during period 1 caused by the deposition of enormous quantities of propagules from one species (Veronica anagallis-aquatica) on a few bank face mats. Overall, deposited propagules were dominated by terrestrial species (59%), with sizeable numbers of wetland species (34%) and a few aquatic species (7%) (Fig. 4a). When species present in the local vegetation were excluded (Fig. 4b) and when species present in the local vegetation and propagule bank were excluded (Fig. 4c), the total number of deposited species was reduced from 172 to 75 and 41, respectively. The frequencies of species deposited, deposited but not in the vegetation, and deposited but not in the vegetation or propagule bank, did not vary significantly between time periods (χ2 = 4.7, d.f. = 6, P = 0.588).

Figure 3.

Mean number of propagules per m2 and total species found within depositional samples (T, M and C combined) from all three study reaches during periods 1–4 and on average over the entire study (periods 1–4). The error bars on the upper graph represent 1 SEM.

Figure 4.

Number of terrestrial, wetland and aquatic propagule species deposited within the three study reaches during each of the four sampling periods (1–4). (a) Total deposited species. (b) Species not present in the reach vegetation. (c) Species not present in the reach vegetation or reach soil propagule bank (Note differences in the scales of the vertical axes).

Information on species and propagule traits were obtained from Hodgson et al. (1995) for 124 vegetation, 143 deposited and 101 propagule bank species. Traits of all recorded species; those in the vegetation; depositional samples; propagule bank samples; depositional samples excluding species in the vegetation; and depositional samples excluding species in the propagule bank or vegetation were compared (Fig. 5). There was no significant difference in the frequency of species that flowered at different times between the different sample types (χ2 = 4.5, d.f. = 8, P = 0.810). In the 4-month periods investigated, an average of 60% of the recorded species flower during June to September, 10% during October to January and 30% during February to May (species were counted in all seasons during which they flower for all or part of the time). However, when average C, S and R scores were investigated for the species present within each of the sample types (Fig. 5a), there was a decrease in the average C score (from 0.49 to 0.34) and an increase in the average R score (0.36 to 0.46) of species present in the vegetation in comparison with species in the depositional samples when species in the vegetation were excluded. Kruskal–Wallis anova identified significant differences in the reach average C (d.f. = 4, P = 0.002), S (d.f. = 4, P = 0.011) and R (d.f. = 4, P = 0.002) scores for species present in the vegetation, propagule bank and depositional samples, with or without exclusion of vegetation and propagule bank species. Pairwise comparisons revealed significant differences (P < 0.05) in average C scores between vegetation and depositional samples (with or without exclusion of vegetation and propagule bank species). S scores differed significantly (P < 0.05) between all depositional samples, depositional samples excluding species present in the vegetation, and depositional samples excluding species present in the vegetation and propagule bank. R scores differed significantly (P < 0.05) between depositional samples excluding vegetation species and the standing vegetation; and between depositional samples and depositional samples excluding vegetation and propagule bank species.

Figure 5.

Differences in the frequency distribution of (a) plant functional type scores, (b) seed bank types and (c) dispersule weights of species found in different subsets of the samples from all reaches, bed and bank locations and sampling periods combined. In (c) S represents dispersules whose weights are too small to be easily determined.

The proportion of species establishing long-term seed banks increased (from 42% to 71%, Fig. 5b) when species present in the vegetation were compared with new species in the depositional samples that were not in the vegetation. There were significant differences in the frequencies of species forming transient, short and long-term seed banks between vegetation and propagule bank (χ2 = 10.5, d.f. = 2, P = 0.005) and also vegetation and depositional samples (χ2 = 6.6, d.f. = 2, P = 0.038) but there was no significant difference between propagule bank and depositional samples (χ2 = 1.9, d.f. = 2, P = 0.378). There was an increase in the proportion of species producing light (< 0.2 mg) seeds (from 22% to 61%, Fig. 5c) when species present in the vegetation were compared with new species in the depositional samples that were not in the vegetation. There were significant differences in the frequencies of species with seeds falling into different weight classes between the standing vegetation and depositional samples (χ2 = 16.3, d.f. = 5, P = 0.006) but not between the vegetation and the propagule bank (χ2 = 9.5, d.f. = 5, P = 0.090). Furthermore the subset of species within the depositional samples that were not present in the vegetation displayed significantly different frequencies by weight from the entire set of species found in these samples (χ2 = 15.7, d.f. = 5, P = 0.008).

sub reach scale

Cluster analysis of species data for Top, Mid and Channel vegetation, propagule bank and depositional samples, based on Sorensen's similarity index with clustering based on unweighted pair-group averages (Fig. 6), revealed a clear distinction between species present in the vegetation in all three reaches and all of the propagule samples. There was also a distinction between the channel and bank vegetation across all reaches. Within the propagule samples, samples from sites F1 and F2 clustered separately from TN, and then within the two individual catchments, M and T depositional and propagule bank samples clustered separately from C depositional and propagule bank samples, with a few exceptions (emboldened, italic sample labels, Fig. 6).

Figure 6.

Cluster dendrogram of species composition of vegetation (veg), soil propagule bank (pb) and depositional samples collected during different time periods (p1–p4) from the bank top (T), bank face (M) or channel (C). The similarity index was Sorensen's and clustering algorithm was unweighted group-pairs. The composition of major (separated by a vertical solid line) and more minor (separated by a dashed line) groupings are described along the horizontal axis according to the reach (F1, F2, TN), type of sample (pb, p1–p4) and sampling location (C, M, T) and the labels of samples which do not conform to these general groupings are emboldened and italicized.

Detrended correspondence analysis of average species abundance in T, M and C depositional and propagule bank samples, revealed clear spatial gradients and temporal patterns within the data (Fig. 7). DCA axis 1 separated samples drawn from F1 and F2 reaches from samples drawn from reach TN. Within these separate catchment groupings, a similar gradient was displayed from C through M to T depositional samples with respect to DCA axis 2 and a more subtle seasonal gradient between summer (periods 1 and 4), autumn-winter (period 2) and spring (period 3) samples within the two catchments with respect to axis 1. All propagule bank samples plotted close to C depositional samples, particularly those from the spring (period 3). In the species DCA plot (not illustrated), aquatic species had low scores on axis 2, illustrating their association with channel depositional samples.

Figure 7.

Plot of sample scores on the first two axes of a DCA performed on average species abundance in propagule bank samples and depositional samples. Abundances were log transformed and no species or samples were excluded from the DCA. The samples are average species abundances obtained at different levels (T, M, C) within the three study reaches (F1, F2, TN) over the four sampling periods. (a) The bank position of the samples – each sample is labelled T, M or C and the outer limit of the distributions of T, M, C depositional samples are enclosed by shaded polygons. The propagule bank samples are not linked by a polygon. (b) The sampling period of the samples – each depositional sample is labelled 1–4 and the outer limit of the distributions of 1–4 within the Frome and Tern sample subsets are enclosed by shaded polygons. The propagule bank samples are not linked by a polygon.

The sample scores on DCA axes 1 and 2 were correlated with traits of species present in the samples: average C, S and R scores; the proportion of species attributable to different seed bank types (transient, short- and long-term); and propagule weight classes (< 0.2, > 0.2–0.5, > 1–2, > 2–10 and > 10 mg). Scores on DCA axis 2 were significantly correlated with average C (r = 0.401, P = 0.006) and R (r = –0.335, P = 0.024) scores; with the proportion of species forming transient (r = 0.494, P = 0.001) and long-term (r = –0.539, P < 0.001) seed banks, and with propagule weight classes > 2–10 mg (r = 0.433, P = 0.003) and > 10 mg (r = –0.533, P ≤ 0.001). Only four species fell into the heaviest weight class, of which by far the most frequently occurring was the aquatic plant Sparganium erectum. Scores on DCA axis 1 were only significantly correlated with propagule weight classes < 2 mg (r = –0.343, P = 0.021) and > 2–10 mg (r = 0.712, P ≤ 0.001).

When all species present in the vegetation were excluded, DCA conducted on the remaining species abundance data identified strong contrasts in depositional and propagule bank samples between the three study reaches. When sample scores were plotted with respect to the first two DCA axes (not illustrated), they formed three compact clusters, one for each reach, but no other discernible spatial or temporal distinctions between the samples. Thus the analysis revealed major contrasts in the species composition of depositional samples between rather than within the study reaches, once species present in the vegetation were excluded.

patch scale

The flow regime varied greatly between study periods leading to marked differences in river bank inundation, which were consistent between reaches (Fig. 2). %Inund showed little significant correlation with deposited propagule abundance or species number and virtually no correlation with these propagule properties once they were confined to new species that were not present in the vegetation (Table 1). EMWL showed higher correlations with properties of the deposited propagules than %Inund, particularly when all propagules were considered and during sampling period 2. However, the strongest and most consistently high correlations were found between the properties of the deposited propagules and the sediment variables (Table 1). Regression analyses applied to all deposited propagules demonstrated significant dependence of both the number of species and abundance of deposited propagules on the fluvial variables, with sediment variables being particularly influential (Table 2). During period 1 when the banks were not inundated, propagule abundance and species richness were weakly associated with EMWL, showing an increase with decreasing elevation. During periods 2–4, when bank inundation occurred, species richness and abundance both showed a generally stronger dependence on fluvial variables, increasing with the quantity of sediment deposited and with decreasing elevation on the river bank. Species richness was more strongly related to the fluvial variables than propagule abundance (higher adjusted R2) and in most cases was explained most effectively by a single sediment variable.

Table 1.  Correlations between propagule species abundance and fluvial variables in the three study reaches and four sampling periods (emboldened correlations are significant at the 0.01 level and italic correlations are significant at the 0.05 level, two-tailed tests)
 Log-SPM2Log-SPM2 not in vegetationNumber of speciesNumber of species not in vegetation
Period 1
 Sample Size363636363636363636363636
Period 2
 Sample Size363638363638363638363638
Period 3
 Sample Size353434353434353434353434
Period 4
 Sample Size363537363537363537363537
4 Period mean
 Sample Size353334353334353334353334
Table 2.  Regression relationships between propagule abundance and species richness of the depositional samples (dependent variables) and fluvial independent variables (%Inund, EMWL, WtSed, WtOrg, WtFines). The listed regression relationships were those with the highest coefficients of determination (adjusted for degrees of freedom), where all of the regression coefficients were significantly different from zero (P < 0.05), and where a maximum of one hydrological (EMWL or %Inund) and one sediment (WtSed, WtOrg or WtFines) variable was included to retain independence in the explanatory variables. NS indicates that no significant regression relationship was estimated
PeriodSiteLog-SPM2 regression equationR2 (adj)No. species regression equationR2 (adj)
1Frome 14.08 – 1.79 EMWL0.38913.7 – 9.23 EMWL0.368
 Tern3.335 – 0.66 EMWL0.17310.61 – 4.3 EMWL0.159
2Frome 13.44 – 0.440 EMWL + 0.200 WtFines0.66612.4 + 0.734 WtSed0.719
 Frome 23.04 + 0.0278 WtSed0.39316.0 – 11.3 EMWL + 0.777 WtSed0.864
 Tern2.67 + 0.528 WtFines0.2988.68 + 5.67 WtFines0.375
3Frome 12.08 – 0.071 EMWL + 0.0592 WtSed0.7201.85 + 1.23 WtSed0.826
 Frome 22.59 – 1.46 EMWL + 0.648 WtFines 0.7822.21 + 16.7 WtFines 0.912
 Tern2.21 – 0.708 EMWL + 0.218 WtSed0.6503.63 + 10.5 WtOrg0.575
4Frome 12.95 – 0.021 EMWL + 0.632 WtOrg0.39614.8 – 8.13 EMWL + 0.602 WtSed0.584
 Frome 22.92 + 0.863 WtFines0.2996.85 + 18.4 WtFines0.477
 Tern2.86 + 0.290 WtOrg0.14413.1 – 8.31 EMWL0.257
4 period averageFrome 13.53 – 0.847 EMWL + 0.032 WtSed0.69810.7 – 4.92 EMWL + 0.899 WtSed0.772
 Frome 23.09 + 0.042 WtSed0.38810.40 – 4.82 EMWL + 1.05 WtSed0.904
 Tern3.08 – 0.528 EMWL + 0.443 WtOrg0.42612.0 – 8.44 EMWL0.496


In relation to our four initial research questions, we have established: (i) significant variations in the species abundance of deposited propagules between sampling periods with some contrasts between reaches; (ii) large numbers of new species in the depositional samples, particularly in winter samples; (iii) differences in deposited species abundance between catchments with clear gradients from channel bed to bank top and between summer and winter-spring; and (iv) strong associations between propagule deposition and fluvial processes.

Overall the research has revealed distinct spatio-temporal patterns of viable propagule deposition that are consistent across all study reaches and catchments. Highest deposited propagule species richness in late autumn and winter followed by spring has also been observed by Gurnell et al. (2006) in a restored reach of the River Cole, West Midlands, UK, and implicates the importance of winter high flows for remobilizing and transporting propagules, as has been shown in low-energy systems in the Netherlands (Boedeltje et al. 2004). However, propagule abundance did not show a clear, temporal trend, lacking the consistent summer peak in deposition observed by Gurnell et al. (2006) but also not matching the winter peak in hydrochorous transport demonstrated by Boedeltje et al. (2004). The lack of a consistent temporal trend in deposited propagule abundance in the present study appears to be largely a function of exceptionally heavy seed production and local deposition onto some sample mats in one study reach from one species, V. anagallis-aquatica. Exceptionally high local deposition was not a major factor amongst the early colonizing vegetation of the study by Gurnell et al. (2006) and could not have influenced the centre-channel sampling of propagules by Boedeltje et al. (2004). Boedeltje et al. (2004) attributed their winter peak in propagule abundance to high winter discharges. However, in the present study and that of Gurnell et al. (2006), sampling of deposited propagules within the riparian zone is affected by both local seed production (predominantly in summer and early autumn) and remobilization, transport and deposition of propagules by high flows (predominantly in winter).

There are a number of aspects of our research which throw new light on propagule dynamics:

viable propagules on the channel bed

A particularly novel aspect of this study has been the identification of a very significant store of viable propagules on the river bed. Whilst previous research has identified viable propagules in bed sediments within backwaters (e.g. Boedeltje et al., 2003b) and sediment dredged from former river channel beds (e.g. Henry & Amoros 1996), the present study has shown that open, active river beds can store numerous propagules and that their species abundance varies in time and space. Indeed, a more detailed analysis of the channel samples from the study reaches has identified preferential sites for propagule storage within the bed and bank toe that are strongly related to the distribution of aquatic and emergent vegetation (Gurnell et al. 2007). The following discussion considers how the propagules in this channel bed store may interact with other propagule sources to feed bank deposition.

deposited propagules of species that are not present in the standing vegetation

Our study has shown that many of the deposited propagule species were not present in the standing vegetation. Such differences have been found in a range of environments. For example, in a recent review, Hopfensperger (2007) investigated 108 studies which compared species richness of the standing vegetation and underlying seed bank, noting that wetlands displayed a level of similarity (average Sorensen index = 0.47, n = 42) intermediate between forests (average Sorensen index = 0.31, n = 34) and grasslands (average Sorensen index = 0.54, n = 55), which she attributed to varying hydrology and to the presence of both transient and persistent seed bank species.

Within wetlands, water-edge environments are particularly susceptible to hydrological and hydraulic processes of scour or deposition disturbance by water that affect both the seed bank and growing vegetation. Grelsson & Nilsson (1991) found a low floristic similarity between the vegetation and seed bank along a lakeshore (Sorensen's index = 0.25), which did not differ between bank levels but did increase with wave exposure. Capon & Brock (2006) found that the similarity between the composition of the soil seed bank and extant vegetation increased following flooding across a desert floodplain and was greatest in more frequently flooded areas of the floodplain, presumably related to propagule reworking and deposition as well as germination of pre-existing propagules. By comparing colonization in pairs of plots, one subject to flooding and deposition of hydrochores and the other unflooded, Jansson et al. (2005) demonstrated that plant dispersal by water and fluvial disturbance are important for enhancing species richness in riparian plant communities. They attributed this to the transport of buoyant seeds over long distances by flowing water and their delivery to riparian plant communities affected by flooding. The differential impact of flood scour and flood deposition was demonstrated by Combroux et al. (2001), who found no relationship between propagule bank composition and established vegetation in scour zones of a side channel, but a strong positive relationship between vegetative propagules and the established vegetation in depositional areas. They suggested that hydrochory was the major controlling mechanism, introducing new species during floods and also remobilizing propagules that had initially been dispersed by other mechanisms. These results underline the importance of floods for recruiting high numbers of species, particularly outside of the growing season (Tabacchi et al. 2005).

In our study, clear gradients in the species abundance of deposited propagules were found from channel bed, through bank face to bank top, in addition to contrasts between summer and autumn-winter-spring, but this spatio-temporal structure largely disappeared when only new species were considered. This suggests that the propagules of new species were deposited more widely and variably across the river bed and banks in comparison with species already present in the reach.

elevational structure in the deposited propagules

The elevational gradients in species abundance found in the depositional samples along all the study reaches is similar to those found in winter-deposited propagules by Goodson et al. (2003). The fact that the new deposited species lacked such a clear elevational structure suggests that the elevational pattern was probably largely driven by direct deposition from plant species growing in the riparian zone along the study reaches. However, the temporal patterns and the shift in the spatial patterns that were also present in the data set cannot be explained by this mechanism, because many of the species deposited during winter and spring were not in flower or producing seed at that time. Therefore, these propagules must have been remobilized from their initial deposition site to reach the sampling mats. Moreover, the relatively high species richness of winter and spring samples at a time of low propagule production and high river flows, suggests fluvial scour and transport of propagules from upstream river margins as well as from within the study reaches.

The species traits of propagules within the depositional samples also contrasted strongly with those of the vegetation. In relation to plant functional types, the depositional samples possessed significantly higher R and lower C scores than the vegetation. When species in the vegetation were excluded, the depositional samples had significantly higher S and R scores than when all deposited species were included. Both propagule bank and depositional samples had significantly lighter seeds than the vegetation, with a greater proportion forming long-term seed banks, lending them to longer term survival, longer distance transport and more widespread deposition across riparian zones than heavier seeds typical of the local vegetation. The similarity in species abundance between all seed bank samples and depositional samples found on the channel bed suggests that many of the propagules stored on the bed are sourced from river bank erosion and bed storage upstream. Deposited propagules also had similar mass and seed bank longevity frequency distributions to those in the soil propagule bank, providing a further indication of a common source or at least a common storage history favouring species with long-lived, light propagules.

Elevational structure in deposited propagule species traits was revealed by significant correlations with DCA axis 2, again probably reflecting an elevationally structured mix of propagules drawn from local and more distance sources. There was an increase in sample C scores and a decrease in R scores along DCA axis 2, suggesting increasing remobilization of locally-produced seeds (i.e. scores becoming more similar to the local vegetation) towards the bank top in combination with the deposition of propagules introduced from more distance sources. Conversely, the relative decrease in C scores and increase of R scores towards the bottom of the bank indicates a relative increase in propagules from the channel bed and bank propagule bank, and possibly new species from upstream sources. The up-bank gradient in total deposited propagule mass and seed bank longevity also indicates subtle interactions between different propagule sources. The proportions of deposited species with relatively heavy seeds and those forming transient seed banks were positively correlated with scores on DCA axis 2, whereas the proportion forming long-term seed banks was negatively correlated. However, the heaviest seed mass class, which largely comprised propagules from the common aquatic species S. erectum, was negatively associated with DCA axis 2, illustrating that heavy seeds produced within the channel are predominantly moved and deposited locally rather than being dispersed up onto the banks. In summary, although the species traits of the depositional samples show overall significant differences from the vegetation, there is an internal elevational structure in the samples that suggests subtle changes in the mix of propagule sources with elevation from bed to bank top.

the role of fluvial processes

All of the above observations suggest more complex sources and delivery mechanisms than previous studies and implicate the importance of fluvial rather than simply hydrological processes. We directly demonstrated the crucial role of fluvial processes through the significant, strong, positive correlations and regression models estimated between properties of the propagules deposited on the river banks and hydraulically-influenced sediment variables (WtSed, WtFines, WtOrg) and the relatively weaker associations with hydrological variables (%Inund, EMWL). Much previous research has considered the buoyancy and floating times of seeds as potential controls on their hydrochorous dispersal into river margins but research results have not been consistent. In an analysis of information from 11 Swedish rivers, Johansson et al. (1996) found that species with high diaspore floating ability were generally more frequent along river corridors, and Boedeltje et al. (2003a) showed that flotation times were important in controlling the hydrochorous dispersal period of species in low energy canal environments. However, in river (e.g. Andersson et al. 2000; Moegenburg 2002) and flume (Merritt & Wohl 2002) studies, stronger associations have been found between environmental, particularly hydraulic, factors and the dispersal/retention of propagules or propagule mimics within river margins, and Nilsson et al. (2002) proposed important interactions between propagule buoyancy and reach characteristics that were reflected in riparian plant community composition. Our results indicate that hydraulic factors are crucial to propagule dynamics in our study reaches. Despite the fact that the rivers Tern and Frome are both quite low-energy, groundwater-fed river systems, mineral sediment particles, which are of higher density than organic material, were mobilized within the river channel during floods and deposited on the banks. This demonstrates that flood flows are also capable of moving non-buoyant but viable propagules off the bed and bank toe into the riparian zone, where we have shown that they settle out of the water column in suitable depositional environments.

Finally, the fact that deposited propagules of species not present in the vegetation are preferentially smaller and lighter and form longer term seed banks than all deposited propagules and also the propagules found in the soil propagule bank, indicates the suitability of these new species for long distance transport over prolonged time periods, whether or not they remain buoyant. Such long distance propagules may be dispersed by a variety of other mechanisms as well as hydrochory and may pass temporarily in and out of storage within the channel margins and riparian zone. In short, connectivity of river flows within the riparian zone (e.g. Leyer 2006) and storage of propagules within the river channel as well as across its banks are complex but fundamental components of propagule dispersal along the study river reaches.


We thank two anonymous referees for their very helpful comments. We also thank Owen Mountford for his invaluable input to the vegetation survey. This research was funded by NERC research grant NER/T/S/2001/00930 as part of the LOCAR programme. We are also very grateful to Mr & Mrs Crook, Mr Druce, Mr & Mrs Fisher, Dr Shaw, Mr Welch, Mr R. Smith and Mr Holmes for granting permission to work on their land.