Rivers as a potential dispersing agent of the invasive tree Acacia dealbata

The silver wattle Acacia dealbata is a fast-growing tree from Australia that has become naturalised in different regions of the world, attaining invasive status in most of them. In Chile, A. dealbata reaches large abundances along banks and floodplains of invaded fluvial systems, suggesting that rivers may act as a vector for seed dispersal. As hydrochory has not been documented previously in this species, the aim of this study is to evaluate the potential for water dispersal of seeds of this invasive tree along rivers. Seed samples from rivers were collected at three sites along two A. dealbata-invaded rivers within the Cachapoal basin, central Chile. Number of seeds collected was contrasted versus hydraulic and local conditions with RDA. Seed buoyancy and sedimentation velocity were determined and compared between sites with an ANCOVA. Finally, the probability of seed germination after long periods of immersion in water was assessed, simulating transport conditions in the flow. Germination results were tested with a GLM. Results indicate that increasing abundance of A. dealbata seeds in the flow is related to the level of turbulence of the flow. Seeds display high floatability but their sedimentation velocity is high when they do sink. Finally, silver wattle seeds can germinate after long periods (many weeks) of immersion in water; however, their probability of germination depends to a large extent on whether seeds are scarified or not. Based on the evidence collected, we suggest that the seeds of A. dealbata have the necessary traits to be dispersed by rivers, this being the first research testing this hypothesis. The success of hydrochory of A. dealbata would depend on river flow turbulence, and whether there are natural mechanisms for scarifying the seeds either before or during transport. The proposed methodology can be used to assess river hydrochory for any tree species.


Background
The silver wattle Acacia dealbata (Link 1822) is a tree native to southeastern Australia, occurring from Tasmania to the north of New South Wales, at elevations between 50 and 1000 m.a.s.l. [1]. This species has been declared as invasive in regions across five different continents [2], including southern Europe [1], South Africa increasing the potential impact of this species on invaded river corridors. Previously documented dispersal mechanisms for A. dealbata include primary dispersion related to seed fall by gravity, as well as the effects of wind [22]. However, and even though its habitat range is very broad [23][24][25][26], there are multiple reports of widespread distribution of A. dealbata along riparian corridors. These include rivers in the species' native range [27][28][29], as well as locations in southern Europe, such as Portugal [30], Spain [26,31,32], and Italy [33], as well as in South Africa [34][35][36], and Chile [37]. This distribution pattern, which is shared with some other Australian acacias, coupled with anecdotical observation of seeds in rivers has prompted some researchers to suggest hydrochory for many Australian trees in the genus Acacia L. [38][39][40], in a generic way. But this hypothesis has not been tested, except for the recent discovery of hydrochory in Acacia stenophylla [41], a similar species which also has an ample riparian distribution and lacks a previous hydrochory background. All this evidence suggests that the seeds of A. dealbata may well be dispersed by rivers.
Following Schupp et al. [42], the effectiveness of seed dispersal mechanisms can be evaluated as the product of two factors: quantity (numbers of seeds dispersed) and quality of dispersal (probability that a dispersed seed produces an adult). In the case of fluvial dispersion, sequential fulfilment of the following conditions would be needed: 1) the seed reaches the river channel (or a nearby location), 2) it is transported by the flow, and 3) it germinates after depositing in the floodplain [43][44][45]. These three steps have been broadly evaluated in the current literature on hydrochory [19,46] and may be checked against A. dealbata's phenology and morphology. For example, some researchers have shown that seed buoyancy is a trait that determines dispersal effectiveness and is an adaptation for hydrochory [44,[47][48][49][50], while others suggest that seed buoyancy is not an important trait as regards to dispersion by rivers, as sunken seeds may also be transported in the water column (bythisochory; [45,[51][52][53][54]). Furthermore, Thompson et al. [55] describe a new type of hydrochory mediated by overland flow, where surface runoff generated during intense storms transports seeds downslope. Acacia dealbata forms lasting seed banks [56], allowing seeds to be transported much later than the seed fall period. However, it is not clear how frequent this transport would be, or what the potential travel distances and precise transport mechanisms (floating or in the water column) are. It is relevant to note that anemochory and hydrochory have been documented acting as sequential dispersion steps [49,53,57], suggesting that seed morphologies adapted to wind dispersion may be also be indirectly optimal for water dispersion. If so, this would improve the likelihood of dispersion of A. dealbata by rivers [22].
An important consideration in this discussion is the germination capacity of the seeds upon deposition on the floodplain, after of immersion. Seeds of other riparian species have been recorded to germinate after being submerged for many days [48,53,58,59], in some cases with increasing germination probability [50,60,61]. The seeds of some riverine species are also able to germinate while floating in water [59,60,62]. All these traits are proposed as adaptative for hydrochory. However, with some exceptions [53,59], previous experiments to evaluate seed germination after submergence or while floating were all conducted in still water, highlighting the need to assess the role of river flow, which is always turbulent [63]. To the best of our knowledge, there are no previously reported assessments about the potential germination of A. dealbata's seeds after floodplain deposition. Given that these seeds require scarification in order to germinate, because dormancy is controlled by their seed coat [56,64,65], any assessment of the role of turbulence should concurrently examine seed scarification effects on the dispersal and germination process of A. dealbata. No other dormancy mechanism has been reported for A. dealbata, but other species in the Fabaceae family show two dormancy types: physiological, by controlling the production of gibberellins (GA) that stimulate germination; and physical, by waterproofing the seed coat, thus avoiding water absorption [66], as for A. dealbata.
Our aim is to assess the potential for hydrochory in this highly invasive species, using both field and laboratory experimental approaches to analyse the feasibility of fluvial dispersion of A. dealbata seeds. To evaluate the magnitude of seed transport, we conduct in situ measurements of the numbers of seeds transported by a river, at different reaches. To assess potential mechanisms of fluvial transport, we use field data as well as laboratory assessments of seed buoyancy and sedimentation velocity, in order to establish whether seeds will float or sink, as well as those hydraulic conditions that favour transport. Finally, to evaluate germination after deposition, we determine the probability of seed germination after long periods of water immersion in the laboratory, considering the effect of initial weight, scarification, and absence/ presence of turbulence.

Study area
Three sites with abundant presence of A. dealbata (centered about 34º25' Lat S, 71º04' Long W) were sampled; two in the Claro River and one at Zamorano Creek, all within the Cachapoal river basin, located in Chile's Central Valley, about 115 km SSE from Santiago ( Fig. 1), an area with sub-humid Mediterranean climate [67]. Both rivers are braided-meandering (i.e., wandering) gravelbed streams, with a median particle size in the cobble range [68]. Even though the closely located Andes range peaks at altitudes between 5000 and 6000 m.a.s.l., the hydrological regime of both rivers is markedly pluvial (i.e., rain-fed), as they drain lower-elevation Andean foothills. Mean flow maxima occur in the months between June and September, in the Austral winter, when there is an increased frequency and magnitude of storm events.

Sampling methods
At each one of the three sites, a study plot of 15 m (perpendicular to the shoreline) by 100 m (parallel to the shoreline) was randomly selected within a larger riparian patch of A. dealbata. All sampling was conducted on January 24 and 31, 2015, during a low-water period, within the seed-release period. All trees inside each plot were georeferenced, measuring their DBH (diameter at breast height). The height and age of each individual was then estimated from its DBH by applying a forest model for A. dealbata in Chile [69]. At each site, we obtained mature seeds directly from slightly closed pods in A.
dealbata trees, before any interaction with potential dispersal vectors. These seeds were then used to perform the experimental assessments of seed buoyancy, sedimentation velocity, and germination. In order to estimate the seed bank and degree of scarification, we collected all seeds found in 15 soil plots of size 30 × 30 cm 2 , carrying out a stratified sampling within the larger 1500 m 2 study plot. To assess scarification in river-submerged seeds, we also collected sediment samples within the wet channel, immediately besides each sampling plot.

Fluvial transport of seeds and hydraulic conditions
At each one of the three sampling sites, fluvial seed transport was quantified by placing capture nets within the wet channel (i.e., in the flow), close to the shoreline. Next to each 100 m-long sampled area, seed transport was measured at five sampling points in the channel, separated from each other by 25 m. A 50 cm-diameter circular net with a mesh opening of 2.3 mm was maintained for 30 min at each point, following the methodology proposed by Kehr et al. [70]. Net dimensions and material were chosen in order to retain both seeds and pods of A. dealbata but were large enough to avoid any backflow issues. The net was submerged to 60% of its diameter (i.e., 30 cm into the flow), thus collecting a combined sample of the material drifting in the water column as well as that floating at the surface. Flow depth (y) was recorded with a surveying rod at each sampling point, and current velocity was measured at five different depths over the vertical with a digital flowmeter (precision: 0.1 m/s): as close to the bottom as possible, at depths of 75%, 50%, and 25% of the total flow depth, and as close to the surface as possible. With these values, the mean flow velocity (V m ) at each measurement point was determined following a modification of the method proposed by Charlton [71]. Then, multiplying net sampling (i.e., submerged) area, sampling time, and flow velocity, we estimated sampled flow volume and seed density at each point. With the measured flow depth y and computed mean velocity V m at each sampling point, we estimated the Froude (Fr) and Reynolds (Re) dimensionless numbers, as in Merrit and Wohl [72]: where g is the standard acceleration due to gravity, µ and ρ are the dynamic viscosity and density of river water, respectively, and y is flow depth. The Froude number (Fr) quantifies the importance of inertial vs. gravitational forces in shallow flows; it is computed as the ratio of the mean flow velocity to the celerity (i.e., wave velocity) of a small disturbance. A Fr > 1 indicates rapid (or supercritical) flow, whereby disturbances cannot travel upstream, while Fr < 1 reflects a tranquil (subcritical) flow, which is by far the most common case in natural, alluvial channel.
On the other hand, the Reynolds number (Re) indexes the level of turbulence in an open-channel flow, being proportional to the ratio between inertial and viscous forces. The dynamic viscosity (µ) was obtained from a viscosity/temperature table for water [73], while water temperature was estimated by averaging data from two nearby gaging stations ( Fig. 1) operated by Chile's General Water Directorate (Dirección General de Aguas, DGA).
For each sampling point, the number of seeds and fruit structures of A. dealbata were determined by inspecting net contents. Captured seeds were subsequently used in the scarification assessments, together with the submerged seeds that were sampled from the river sediment (details in the next section).
The seed rain in the vicinity of each net was indexed as a function of the abundance and mean DBH of adult A. dealbata individuals located within different areas of influence, considering radii of 15 m, 30 m, and 45 m from each point where a net was placed. In order to relate the abundance of A. dealbata seeds captured at each sampling point in the river (response variable) to the hydraulic conditions at, and abundance of adult trees around the point (predictor variables), a Redundancy Analysis (RDA) was performed [74]. To determine the significance of each predictor variable in the RDA, we performed a permutation analysis (1000 replicates), checking if the explanation of each predictor variable on the response variable was different from that obtained by chance.

Buoyancy and sedimentation velocity of seeds
The buoyancy index of both seeds and seed pods, as well as the sedimentation velocity of the seeds were determined in the laboratory. The seeds and pods utilized in these experiments were randomly selected from those collected from A. dealbata´s trees at each sampling site. The buoyancy index is the average time a sample stays on the surface of quiet water [75], while the sedimentation (or settling) velocity is defined as the mean terminal velocity at which a seed, initially placed just under the free surface, falls through a column of quiet water [53].
The buoyancy index was estimated for both seeds and pods (10 each per site, 60 in total). For these experiments, each individual seed or pod was carefully placed on the free surface of a container filled with water and left to float, until it either sank or the experiment ended (after 30 days). The time of buoyancy of the seeds was compared with those established by Romell [76] cited in Danvind and Nilsson [75], who qualified as "good floaters" those seeds that manage to float for more than two days under ideal conditions (still water, no wind or any other perturbation). To generate his index, Romell [76] floated seeds of different species, obtaining this threshold [77].
The sedimentation velocity was measured only for seeds (10 per site, 30 in total); in laboratory conditions at ≈ 20 °C, seeds were carefully dropped just below the surface of a clear, 40 cm graduated cylinder filled with water, recording the time needed for travelling a known distance. To evaluate the role of morphological traits and seed's site of origin on sedimentation velocity, we used ANCOVA considering seed weight, length, and circularity index as covariates and site origin as a way [78]. The length and circularity indices were estimated using function coo_scalars from Momocs package [79], employing the photo of each seed to obtain its outline and then compute the different indices. The circularity index, computed as P 2 /A (where P and A respectively are the perimeter and the surface area of the seed as seen from above, when lying flat), is a shape factor that allows to compare any shape against a perfect circle; the minimum result indicates a perfect circle, while increasing values reflect more asymmetrical, elongated shapes. There was no need to include other shape indices in the ANCOVA, as they all are highly correlated.
To obtain a first estimation of the potential distance that seeds can travel downstream once they sink, we borrowed a method used in environmental engineering for the design of sedimentation (settling) tanks. By applying a vectorial composition of the sedimentation velocity in the vertical and the mean flow velocity (as measured in the field) in the horizontal, with the field-measured flow depth as an end for the displacement, we computed the distance to deposition, assuming steady-state and uniform flow conditions for sedimentation equal to those at each measurement point. As the actual flow in rivers is turbulent, instantaneous, vertical velocity fluctuations will keep the seeds in suspension for longer, so this calculation only gives a lower bound for transport distance, under the above assumptions. Moreover, downstream variations in channel shape, and corresponding flow adjustments, will either increase or decrease the travelled distance; for example, if the measurement was taken within a riffle, but a pool lies immediately downstream, settling might happen much earlier.

Germination experiments
We conducted two laboratory experiments to simulate the germination of A. dealbata seeds during and after fluvial transport, considering two factors: 1) Seed condition: "scarified" or "unscarified", and 2) Immersion time: evaluating germination effects for 15, 35, and 55 days immersed in water (Graphic summary in Fig. S1). All experiments were conducted inside germination chambers keeping temperature at 20° C.
The first experiment, 'germination in water, ' was designed to determine germination and survival of seeds (scarified or unscarified) when transported in the flow. For this, 30 seeds per treatment (10 from each site, 5 scarified, and 5 without scarification) were immersed in water for either 15, 35, or 55 days, while kept in constant motion, emulating river transport. As a control, 30 seeds (10 per site) were immersed in water at the same temperature but without movement, again including scarified and unscarified seeds (5 seeds per site of each), for 30 days. The response variable is the proportion of viable seedlings at the end of each experiment (survival rate), but we also quantified hollow seeds and rotten seedlings. The germination experiment in water was performed with a motor-powered device specifically designed to keep seeds in constant motion in a water bath (Fig. S2), attempting to simulate fluvial transport.
The second experiment, 'germination over a substrate, ' was designed to determine the probability of germination of A. dealbata once a river-transported seed is deposited in the floodplain. This experiment was carried out on vermiculite substrates in Petri dishes, with one seed per dish, irrigated ad libitum (every one or two days). We utilized 15 seeds per treatment (5 per site, all unscarified) which had been previously submerged for 15, 35, and 55 days, under motion. Then, of those seeds that did complete the immersion experiment (i.e., did not rot or ended up hollow-without cotyledon), half were scarified before putting them all to germinate (n = 98). For estimating the germination rate of seeds not subjected to fluvial transport, we made a control experiment with 30 seeds (10 per site; 15 scarified and 15 unscarified), collected from the floodplain seed bank (i.e., that never fell into the water), germinating them over the same substrate, without previous immersion. Germination was counted every 2 days over a 30 day-long period.
Since the seeds of this species germinate in autumn, the photoperiod was set to alternating 12-h periods of light and shade, to simulate a natural light environment. The scarified seeds were obtained by damaging the coat with a scalpel, at that end of the seed opposite to the location of the embryo. A seed was considered to have germinated when it exhibited both the radicle and hypocotyl.
For both experiments, results were evaluated using a GLM analysis for binomial distribution (1: seed germinated, 0: seed without germination) [80]. A two-way analysis was performed, comparing the relationship between days of immersion and condition of the seed (scarified or unscarified). In order to make a-posteriori comparisons, a permutation test was generated, which compared whether the differences in probability of germination between groups are similar to those expected by chance. To avoid Type I error, the obtained values were later corrected by an FDR (False Discovery Rate) analysis, following Benjamini & Hochberg's (1995) formula. All statistical analyses and database processing were performed in software R v.3.2.3 [82]. All raw data and the R script to replicate statistical analyses and figures are given in supplemental material.

Results
Overall, mean seed weight was 7.3 mg and mean length was 4.3 mm. At the Claro 1 and Claro 2 sampling sites, the canopy of adult A. dealbata trees extended beyond the shoreline, partly covering the water surface, whereas at the Zamorano site, the closest trees were located up to 15 m inland from the shoreline (Fig. 1) (Table S1).

Evidence of dispersion by rivers in A. dealbata
Seeds were collected in the flow at all three sites sampled (Claro 1 mean: 0.8 seeds/net -0.008 seeds/m 3 ; Claro 2 mean: 8.2 seeds/net -0.047 seeds/m 3 ; Zamorano mean: 0.4 seeds/net -0.036 seeds/m 3 , Table S1). Despite observing a few pods drifting in the flow and lying on the riverbed, no pods were captured by the nets. The density of seed banks was much higher at Claro sites (Claro 1 mean: 10,600 seeds/m 2 ; Claro 2 mean: 5,760 seeds/m 2 ) than at Zamorano (mean: 187 seeds/m 2 ).
The RDA results indicate that only the Reynolds number is significantly related to the number of A. dealbata seeds collected in the flow ( Table 1). The observed relationship between both variables is positive (Linear regression: R 2 = 0.363, p = 0.008), thus, the higher the turbulence of the river flow, the higher the number of seeds captured (Fig. S3).
The buoyancy time of seeds and pods was greater than 30 days in 94% and 100% of cases, respectively, which qualifies the seed propagules of A. dealbata as "good floaters" [76,77]. The minimum distance that a seed would travel while sinking, under the flow conditions at each sampled site (but without turbulence) varies with the flow velocity and depth. The lowest value was 0.49 m (Zamorano), while the maximum was 6.07 m (Claro 2) (Table S1).

Germination in water
In the 'germination in water' experiments, the survival probability of scarified seeds was very high in all treatments, with a value of about 0.95 for the 'still water, ' '15 days, ' and '35 days' treatments, and around 0.55 in the case of the '55 days' treatment ( Fig. 2). In contrast, the germination rate of unscarified seeds in water did not exceed 16%. Rotten and hollow seeds occurred for both the scarified and unscarified cases, with a higher proportion of rotten than hollow seeds (Table 3). GLM analysis did not detect significant interactions between immersion time and seed condition (scarified/unscarified), but each factor did show significant results (Table 4). In the case of immersion time, a-posteriori pairwise analyses did not reflect any significant differences.

Germination over substrate
For the 'germination over a substrate' experiment, as in the 'germination in water' experiment, scarified seeds germinated at a much higher proportion than the unscarified ones (Table 3). GLM analysis suggests a significant interaction between immersion time and seed condition    Table 3 Summary of results for the (A) 'Germination in water' and (B) 'Germination over substrate' experiments. Each seed was individually assessed after completion of each experiment, according to the following categories: "Germ", seeds that germinated during the experiment and were viable at the end; "Rot", rotten seeds at the end of the experiment; "Hollow", seeds with only tegument (no embryo) at the end of the experiment. "Control still W" is a result from a control experiment, where seeds were germinated in water without motion. "w/o W" is a result from a control experiment in which seeds were germinated without being immersed in water. More details in Fig. S1 Scarified Not scarified (scarified/unscarified) ( Table 4). A-posteriori analysis of paired comparisons suggests that the treatments 'scarified seeds/35 days immersed' and 'scarified seeds/55 days immersed' differed significantly from all treatments with unscarified seeds. It needs to be emphasized that, while the probability of germination decreases with immersion time in the case of unscarified seeds, it actually increases for scarified seeds (Fig. 2).

Discussion
Studies on riparian communities commonly assess the existence of hydrochory by evaluating seed floatability and/or fluvial transport [47-49, 57, 70]. On the other hand, studies that assess all the necessary steps for effective river dispersal, also considering the impact of river transport on germination rates, are rather scarce [e.g., 56,57]. Our study evaluates hydrochory as a potential, previously unreported dispersal mechanism in A. dealbata, a species of great relevance due to its invasiveness [1,10,24,26,83]. In general, both our field and laboratory evidence strongly suggest that A. dealbata seeds interact with fluvial flow, allowing rivers to act as a dispersal vector. We provide field evidence that seeds are transported by river flow and, in addition, the germination experiments show that A. dealbata seeds are able to germinate after 15, 35, and 55 days of immersion. In what follows, we discuss how the potential relationships between the recorded seed traits and the riverscape may determine the effectiveness of river dispersal in this species. We also discuss the potential implications of riverine dispersal for invasiveness, and whether the traits recorded in A. dealbata would determine an adaptation for hydrochory.

Transport by rivers
Comparing our results with previous studies, we document that A. dealbata seeds have longer average flotation periods than those reported for other riparian species [47,48,57,60,62]. Regarding sedimentation velocity of seeds, we only found data for the cottonwood Populus trichocarpa, an anemochorous/hydrochorous species that displays a mean sedimentation velocity of 0.018 m/s [53], one-third of that estimated for A. dealbata. Finally, observed seed density in the flow (number of seeds per unit volume of river water) was much lower than that recorded in some previous studies [70,84,85], but in the same order of magnitude to the densities documented by Brown and Chenoweth [86] and Meier [53]. Note though that these comparisons are not very relevant, as these authors worked with different species or groups of species; furthermore, seed density should strongly depend on the timing of seed fall interacting with the concurrent hydrologic/hydraulic conditions (such as river discharge, which locally determines flow depth and velocity, and thus the Reynolds number).
In comparison to other species in the literature [47,76] A. dealbata´s seeds can be classified as "good floaters", based on the flotation periods we observed. It should be noted though that this was mostly due to surface tension effects, so it is likely that when exposed to turbulent flow-as found in all rivers, the seeds would quickly sink. This notion is supported by the relatively high sedimentation velocity obtained in our experiments. The results of RDA, which relates numbers of seeds in the flow to turbulence (as indexed by the Reynolds number), suggest that A. dealbata's seeds may be dispersed by rivers through mechanisms similar to those responsible for sediment transport, i.e., in suspension/saltation [87], where seeds are kept in suspension within the water columnwith or without episodic bed contact -due to turbulence [19]. While studies that have tested this mechanism in detail are scarce, our results provide evidence that suggests such dispersal method. It should be noted that, even though the relationship between the number of captured seeds and the Reynolds number is indeed driven mostly by a single high value at one of the sites, the upper envelope to the data points clearly displays a regularly increasing trend (Fig. S3).
Regarding the frequency of river transport of A. dealbata´s seeds, and the fact that the density of seeds was low, we need to consider three aspects: First, we sampled towards the end of the seed fall period [1]; second, both rivers were under low-flow conditions, with close to minimum capacities for transporting seeds; third, seed bank densities at our study sites are very high compared to other species [88,89], only comparable to the seed banks for other Acacia species [90][91][92]. This suggests that A. dealbata's seeds may be transported both during low-flow conditions (Austral summer, corresponding to the seed fall period), when seeds directly fall into the river, as well as during high-flow months (Austral winter), entrained from the seed bank by overbank flows, during flooding events.
Hyslop and Trowsdale [45] provide a conceptual model which depicts how interactions between flow stage, river geomorphic diversity, and seed phenology, influence hydrochorous seed dispersal and deposition. According to our results, this model suggests that seeds on the floodplain (in the seed bank) may germinate, be remobilized, or drift downstream, depending on their position and the occurrence of flood events. Seeds that fall directly into the flow may be transported or not, depending on the river's diversity of hydrogeomorphic conditions. In reaches with lower turbulence levels (i.e., pools during low discharges), it is likely that most seeds will sink, given their relatively high sedimentation velocity, only to be mobilized later, during floods. On the contrary, in reaches with higher turbulence (e.g., riffles), seeds may be transported downstream as soon as they fall.
Finally, since the weight of A. dealbata's seeds relates positively to their sedimentation velocity, the question arises as to whether there will be a selection of seed weights by the flow, with lighter seeds travelling longer downstream. Comparing seed weights in this work with those reported in international databases for A. dealbata [93][94][95], it is observed that our seeds have the lowest recorded weights (mean of 7 mg for sampled sites, versus 11 mg mean in other databases). This could reflect either a selection of small seed sizes in riparian sites, or else could simply be a characteristic feature of this species in Chile.

Post-transport germination
In the 'germination in water' experiment, the highest germination rate (~ 95%) occurred in scarified seeds, with water motion being irrelevant. On the other hand, scarified seeds that were never immersed ('germination over a substrate' experiment control) only displayed a germination rate of around 50%. Together, these results show how relevant both water imbibition and scarification are for germination, confirming the role of the physical dormancy mechanism for this species [64,65]. The importance of water imbibition to seed germination process has been broadly recorded [96][97][98][99]. It allows for the resumption of normal seed metabolic levels and promotes mechanisms to repair the damage occurred during drying. Furthermore, different studies have shown how seeds increase water imbibition after scarification, validating our results [99][100][101][102].
In the case of the 'germination over a substrate' results, seed immersed 35 and 55 days in water, subsequently scarified, and then germinated over a substrate, showed higher germination rates than scarified seeds that were either never immersed or were just immersed for 15 days (P 15d movW in Fig. 2F). The corresponding germination rates of unscarified seeds were close to 0 in most cases, except for the treatment "15 days water in motion-germinated over substrate" (P 15d movW), where the mean germination rate was 0.22. Both results can be explained by the possible participation of a second, physiological dormancy mechanism, besides physical dormancy due to the coat. If both mechanisms are active within the first 15 days after immersion, but physiological dormancy dominates over physical dormancy during this period, then scarification would not be as important in triggering germination; in relative terms, this would cause lower germination of scarified seeds and higher germination of unscarified seeds, as compared to the longer (35 and 55 days) treatments. On the other hand, if some stimulus or process diminished or altogether stopped physiological dormancy at some moment between 15 and 35 days of immersion, it would explain the increased germination rates after scarification, in the 35 and 55-day treatments, as well as the decreased germination of unscarified seeds. In this way, our results could be explained by the differing temporal dynamics of two dormancy mechanisms. Dessi et al. [99] observed an increase in germination rates of non-scarified seeds at 25 °C in A. dealbata and Acacia mearnsii, which is a 5° higher temperature than in our experiment, suggesting than the second dormancy mechanism could be related to water temperature. However, the metabolic controls of the different germination phases and their relationship with dormancy are still not fully understood [97,103]. Further studies are needed for A. dealbata, in order to confirm or refute our proposed mechanism.
An increase in germination rates after immersion was reported by Kowarik and Säumel [50] for Ailanthus altissima, an anemochorous species that recorded higher germination rates after three days immersed. Lopez  [62] recorded that species such as Pterocarpus sp. and Pterocarpus officinalis (from tropical, seasonally flooded forests) showed a peak in germination after 40 days in water, while other species like Pentaclethra macroloba (also from flooded forests) and Gustavia superba (an upland tree) showed decreasing germination rates when increasing immersion times. However, none of these studies considered water motion, to simulate river transport. To our knowledge, only two studies have considered this factor. Meier [53] found that there is no effect of water movement on the seed germination rates of the cottonwood Populus trichocarpa, even though hypocotyl and radicle lengths were significantly shorter in the motion treatment, as compared with the control. Rouifed et al. [60] showed that after four days in moving water, germination rate increased from 0 to 80% in Fallopia x bohemica, an anemochorous species. Summarizing, the relationship between germination rates and the duration of immersion seems to vary on a species-by-species basis, without a clear pattern.
Given the discussion about transport mechanisms in rivers and considering the results of our germination experiments, we conclude that the successful germination of a seed that has been transported by the river (effective dispersal) depends on two milestones: firstly, that the seed is scarified, and secondly, that it is deposited on the banks or floodplain, after being transported. Whether these happen, and the order in which they occur may well determine the probability of germination success for the seed. For example, if a seed is scarified during transport by the flow, our results indicate that it would probably germinate in the water; whether it succeeds or not would then depend on stochastic fluvial processes, as it would need to deposit on the floodplain within some window of time, or else sink and lose viability or be washed downstream. In turn, if a seed managed to deposit on the floodplain, it could either initiate a new dormancy process in the seed bank, or else it would probably germinate if scarified. These examples show the dynamism and complexity of the different processes, which will vary depending on the characteristics of the river (including continuously changing river discharge, and thus hydraulics), plant phenology, and seed characteristics.

A new dispersion vector for Acacia dealbata
This study provides strong evidence suggesting that dispersion and colonisation of A. dealbata along rivers is feasible, leading us to consider it as a new dispersal vector for this species. Whether our results provide evidence of adaptation to hydrochory by A. dealbata or they just reflect an exaptation is an open question. According to some authors, the principal trait associated with hydrochory is phenology [44,51,104], but for others it is diaspore floatability [16,47]. However, recent literature has shown how some seed phenotypes that were typically associated with a single dispersal mechanism due to their traits (e.g., "if it floats, it is hydrochorous"; "if it has winged structures, it is anemochorous"), could very well use other dispersal mechanisms for which they are apparently not adapted [49,50,105,106].
In this context, we highlight the study of Planchuelo et al. [57], who show how morphological adaptations for wind dispersal may also be optimal for water dispersion. Their description fits A. dealbata traits that were previously described when suggesting wind dispersion for this species [22]. Thus, traits such as a low mass seed, light pods with seeds attached, or a high surface area for seeds and pods may yield positive dispersal both for wind and water transport in A. dealbata, suggesting that there may be a selective pressure to simultaneously improve both. In this context, Sádlo et al. [107] propose a new classification of plant dispersal strategies, classifying species as "Phragmites type" when they display both hydrochory and anemochory (as we are proposing for A. dealbata). Finally, it is important to note that even though our study contributes evidence, it does not fully resolve the question about "hydrochory or not?" in A. dealbata, particularly considering that the discussion about the relationship between seed traits and dispersal mechanisms is in full swing.
In general, the literature associates successful germination of A. dealbata with the recovery from forest fires (since heat scarifies the seeds) [26,31], leading to displacement of competitors due to both allelopathy [108] and the fact that adults have fast seed production, reproducing after their second year [109]. In this sense, the possibility of seed dispersion by fluvial transport was generically suggested by other authors for a range of Australian acacias [38][39][40] but has never been tested. To the best of our knowledge this study is the first record of hydrochory in A. dealbata. This would increase the list of traits explaining this species' high invasiveness, experimentally corroborating its success as an invasive species, accounting for its enormous abundances along river banks and floodplains.
The invasiveness of A. dealbata in central Chile has been broadly researched, allowing us to expand our discussion, pointing out potential impacts that hydrochory in this species could have in our study area. Following a climatic niche evaluation, it has been estimated that several areas with suitable conditions remain to be colonised by this species in Chile [10], with researchers suggesting that dispersal limitations of A. dealbata have restricted its spread. Because of increasing land-use change in riparian ecosystems [110], high fragmentation of rivers in central Chile [111], and the short length of Chilean basins, we would not expect long dispersal processes in Chile via hydrochory. By contrast, the ability of this species to colonise areas developing high-density patches that exclude native species via allelopathy [24,112,113], as well as the high seed-bank densities observed, suggest a "step-by-step" dispersal process that could be broadened by fluvial transport, extending dispersal areas downstream from site scale to local scale [114]. In this way, considering available habitats to colonise, we predict that the bigger impacts of hydrochory in Chile should occurs at the mesoscale.
Finally, significative differences have been documented in seed traits [115] and seed germination rate [116,117] within populations of different Acacia species, including A. dealbata; furthermore, Chilean populations of A. dealbata originated by several introduction events, followed by admixture [118]. This background suggests high phenotypic and genetic variation for this species, which would impact over its seed traits, and thus how generalizable our findings are to other populations (e.g., in Australia). However, the overwhelming presence of A. dealbata in riparian ecosystems suggests that A. dealbata's seeds should be able to interact with rivers as a dispersal agent in most of the area colonised by this species.

Conclusions
Our results, together with evidence from previous investigations, suggest that dispersal by rivers is a viable strategy for A. dealbata, highlighting the role of scarification for its germination success, and the importance of hydraulics in the transport and viability of the seeds. Future investigations should target the actual impacts of hydrochory on the population dynamics of A. dealbata by using methods specifically aimed at estimating whether river transport of seeds favours the reproduction and dispersion of this highly invasive species.
Abbreviations RDA: Redundancy Analysis; GLM: General Lineal Model; ANCOVA: Analysis of covariance; Treat: Experimental treatment, days of immersion in water; Cond: Seed condition for experiment, scarified or unscarified; Still W: Experimental treatment, seed germination in still water; Mov W 15d: Experimental treatment, seed germination in motion water for 15 days; Mov W 35d: Experimental treatment, seed germination in motion water for 35 days; Mov W 55d: Experimental treatment, seed germination in motion water for 55 days; w/o W: Experimental treatment, seed germination without include water; P 15d movW: Experimental treatment, seed germination after 15 days in motion water; P 35d movW: Experimental treatment, seed germination after 35 days in motion water; P 55d movW: Experimental treatment, seed germination after 55 days in motion water.