Across all subestuaries, we generally found a similar pattern in the community composition of the seedbank. The seed banks in most cases were not distinct by plot types (C, R, N), and instead showed a moderate amount of overlap across plot types and years (as indicated by somewhat overlapping ellipses in Fig. 2). Exceptions to these patterns did occur, however, as we will describe below in our detailed discussion of the individual subestuary responses (Table 1).
In the Nanjemoy, prior to herbicide treatment (2011), the seedbank in the native reference wetland differed from the Phragmites removal and control plots, yet there was overlap in seedbank composition between all three plots (Table 2; Fig. 2). In 2012 and 2013, following herbicide treatments, the seedbank in the reference plot remained distinct, and there was little differentiation between the removal and control plots (Table 3; Fig. 2). Pairwise perMANOVA results showed that seedbanks in control and reference plots were marginally not significantly different (α = 0.05) during 2011, but were significantly different during subsequent years. Meanwhile, the intact control plot and the herbicide-treated removal plot initially differed significantly but converged following herbicide treatments. During all years of the study, seedbank composition was significantly different between reference wetlands and herbicide-treated plots (Table 2).
In 2011, the seedbanks of all three plots in the Patapsco subestuary differed from each other (Fig. 2), but the native reference plot was more similar to the Phragmites removal plot than either were to the Phragmites control plot (Table 2; Fig. 2). Following the first herbicide treatment, the seedbank of the Phragmites removal plot differed from the control and reference plot so much that there was no overlap between plots (Fig. 2). The native reference clustered with the Phragmites control plot (Table 2; Fig. 2). In the second year after herbicide application (2013), there was more overlap between all three treatments (Table 2; Fig. 2). All pairwise comparisons of seedbank composition between plots were significantly different in each of the 3 years (Table 3).
In the Severn River, the three plots clustered individually with minimal overlap, indicating different initial seedbank compositions (Fig. 2). Following the initial herbicide treatment, those differences became more pronounced (Fig. 2). By 2013, the plot from which Phragmites had been removed had a more distinct seedbank, as seen in the distance between clusters, from intact control and native reference plots (Fig. 2). The three plots were significantly different from each other in all pairwise comparisons over all 3 years (Table 3).
Prior to herbicide treatment, the native reference seedbank in the St Leonard samples was distinct from either of the two Phragmites dominated sites (removal and control) and exhibited a spatially uniform distribution of seeds within the plot, as seen in the tight NMDS plot clustering (Fig. 2). The control and herbicide-treated Phragmites plot seedbanks were similar, but neither were as spatially uniform as the native reference plot (Fig. 2). Following the first spraying treatment, the seedbank at the native reference plot remained distinct from the Phragmites plot. The Phragmites removal and control plot seedbanks remained similar, but not as evenly as prior to herbicide application (Fig. 2). By 2013, the Phragmites removal and control plots developed similar and homogeneous seedbanks (Fig. 2). The distinctness of the native reference plot, which moved farther from the Phragmites control and removal plots in each year, was validated by perMANOVA results, which showed that the reference wetland was significantly different from the Phragmites-removal wetland each year (Table 3).
The seedbanks of the plots in the Wicomico subestuary were similar prior to herbicide application, as indicated by the significant overlap in the NMDS plots (Fig. 2). Differences between Phragmites-removal and control treatments emerge by 2013, though there were still similarities between plots (Fig. 2). The reference plot was significantly different from control and removal plots in 2011, according to perMANOVA results, but there were no significant differences between plots during in subsequent years (Table 2).
The functional composition of the seed banks generally reflected unique compositions of each subestuary, rather than distinct differences across years and plot types (C, N, R; Fig. 3). Based on seedling emergence counts, seed quantity varies broadly across years in nearly all subestuaries. Nonetheless, some interesting year and plot type trends within a subestuary occurred and are discussed in greater detail below.
In the Nanjemoy, the functional vegetation types (forb, graminoid, and woody) showed similar compositional trends (no significant plot effects across years) in during all 3 years of the study: increasing germination over time, especially among forbs and graminoids (Fig. 3). The number of woody plant species seedlings were low in all years, however, and the number germinating decreased over time. There was no significant plot effect on the number of seedlings in the emergence of species in the functional groupings in any year (Fig. 3).
Prior to spraying in the Patapsco River, the seedlings that emerged in the functional groups were similar in the two Phragmites-dominated plots: forb seedlings were more prevalent than graminoids, which, in turn germinated in higher numbers compared to woody species. The native reference plot had fewer graminoids seedlings relative to forbs and woody vegetation. This overall pattern of emergence continued in 2012 (Fig. 3). Woody vegetation emergence was significantly higher in the reference plot in years 2011 and 2013, which was the only significant effect of plot on functional vegetation (Fig. 3). Within the native reference plot, woody plant germination was higher than graminoid, but not forbs in all 3 years. In all treatments, the germination of forbs was exceptionally high in 2013. Overall, graminoid germination was low in all three treatments for each year. In the removal and reference plots, the emergence of forbs increased by an order of magnitude between 2011 and 2012, while the number of forbs did not increase in the controls until 2013 (Fig. 3). Typha spp. and Pluchea odorata were the dominant forbs across all plots.
During 2011, the functional groups in the Severn River removal plot resembled the native reference more than the Phragmites control, in that both seedbanks were dominated by forbs. The reference plot initially had higher emergence of woody species and lower graminoid emergence than the two Phragmites dominated plots (Fig. 3). In years 2012 and 2013, forb germination in the control and reference plots was significantly greater than in the removal; during that period, the removal plot had significantly higher graminoid emergence (Figs. 3 and 4).
Examining the functional group diversity within seedbanks in the St Leonard River, plants that emerged from soils in the native reference plot was heavily dominated by forbs during the entire study. The most common forb in the reference seedbank was A. cannabina. In the reference plot, few graminoid and woody species emerged from the samples in all years, and the graminoid germination was significantly different than the other plots in each year (Fig. 3). In Phragmites removal and control plots, 2012 graminoid emergence was higher than forb emergence, a shift from other years.
In the initial, pre-spraying sampling of the Wicomico River Phragmites plot, the three plots showed very similar patterns in emergence between the functional groups. In all three plots over all 3 years, forbs dominated the seedbank, followed by graminoids, while woody vegetation was a minor component (Fig. 3). This pattern persists in 2012, with the exception of greater graminoid germination in the control plot, while the reference and removal plots remained similar to 2011. The only significant difference in functional vegetation between plots was between the control and removal plots in 2013. Despite being low relative to forb germination, there was greater graminoid emergence in the Wicomico than other subestuaries. The graminoids were dominated by Schoenoplectus robustus, Schoenoplectus acutus, and Spartina cynosuroides.
The seedbank in each subestuary differed in the composition of species in guilds, and generally did not vary by year or cover type. Exceptions to this overall pattern are discussed in detail below. Even though the relative proportions of guilds in the seedbank were consistent within subestuaries (Fig. 4), the numbers of seeds representing each guild often varied by several orders of magnitude.
The Nanjemoy seedbank plant guild (native annual, native perennial, and introduced perennial) germination showed anecdotal variation between years. In the Phragmites control and removal plots, native annual emergence was greater than native perennial emergence in 2011, while they were equal in the native reference plot. In 2012, the Phragmites removal plot plant guild composition resembled the pre-treatment (2011) pattern, while in the control and reference plots native annual germination was marginally higher than native perennial and both exceeded germination of introduced perennials. The only significant difference in emergence between plots was in 2013, when the removal plot had significantly lower introduced perennial germination than the reference plot (Fig. 4).
Introduced perennial germination was the distinguishing feature of the functional guild analysis of the Patapsco subestuary. In 2011, introduced perennial species dominated the Phragmites removal plot. The native reference had higher germination of native perennials than any other guild, though introduced perennial germination was high as well (Fig. 4). In 2012, after spraying, the removal plot had similar plant guild responses in the control plot (Fig. 4). The control and reference plots both had high emergence of introduced perennial, which was significantly different from the removal plot in years 2012 and 2013. The removal had highest germination of native annuals in 2013, while the reference plot had a high germination rate of native perennials (Fig. 4). The introduced perennial guild increases paralleled the forbs pattern seen in the function analysis and was driven by the prevalence of Typha spp. The elevated presence of native perennials in the reference site was likely due to the relatively high germination of woody species Iva frutescens, Baccharis halmifolia, and Hibiscus moscheutos.
Prior to spraying in the Severn River, all three plots had high numbers of seeds germinating in the native perennial and introduced perennial guilds. Native annual emergence was significantly higher in removal and reference plots than the Phragmites control plot. In 2012, the number of plants that emerged in each of the guilds changed dramatically with the number of native annuals highest in the control plot, followed by the reference, while absent in the removal. All three plots had high to moderate numbers of introduced perennial species that germinated in 2013. The control and reference plots had significantly higher rates of native annual germination than the removal (Fig. 4). The prominence of introduced perennial in seedbank composition in all years and plots was driven by two species: Phragmites and Typha spp. The native annual guild, prevalent in alternating plots over the study, was largely comprised of P. odorata. The forb functional group was also likely dominated by P. odorata as well. The less common native perennial guild was comprised mostly of Spartina patens and Schoenoplectus robustus.
The dominance of A. cannabina in the St Leonard reference wetland was reflected in the species richness and plant guild assembly as well: native annuals were the most prevalent group. In all years, the control and removal seedbanks were significantly different and comprised primarily of introduced perennial and native perennial (Fig. 4). Common native perennial species in this wetland included Schoeplectus robustus and Eupatorium altissima. The control and removal plots both had increased emergence of both perennial guilds in years 2012 and 2013 (Fig. 4).
As with the functional types, in 2011, the Wicomico seedbanks had similar plant guild patterns across all plots (Fig. 4). The only change in 2012 was increased emergence of native perennial species from soils collected at the control plot. During 2013, the number of native annual species that germinated rebounded to 2011 levels. None of the guilds were significantly different between plots in any year (Fig. 4). Across all plots and years, the native perennial guild is a prominent component of the seedbank composition. The most prevalent forbs (the dominant functional group) in the Wicomico were the native annual species P. odorata and A. cannabina. Common native perennial species included the graminoids mentioned above, as well as the native species Lythrum lineare and Polygonum punctatum, and the introduced species Typha spp. and Polygonum hydropiperoides.
The number of Phragmites seeds that germinated from the seed bank was generally a small proportion of the total seeds in each seed bank, across plot types and years, regardless of subestuary (Fig. 5). The year-to-year variation in the total number of plants that emerged from the seedbanks was quite large and did not seem to generally vary in any discernable pattern (e.g., pre- vs. post-Phragmites removal) across subestuaries (Fig. 2). The within subestuary patterns in total seed germination seemed to reflect unique seed bank properties specific to the subestuaries, as described below.
Phragmites germination was a small fraction of the total germination in the Nanjemoy and only significantly different between the reference and the two Phragmites plots in 2011 and 2012 (Table 1; Fig. 5), while total germination of all species across functional types and guilds was highest in the reference plot in all years. Species richness was highest in the reference plot in years 2011 and 2013, but not in 2012 (2011 6.7 ± 0.58; 2012 2.7 ± 0.54; 2013 4.6 ± 0.41; Fig. 6). Richness was higher in the control than the removal in years 2012 and 2013, and the latter plot did not seem to have an herbicide effect on richness (Fig. 6).
Germination of seeds from Patapsco River samples was significantly different between the plots in 2011 and 2012, but not 2013 (Table 4). The pattern of total emergence was driven by significantly higher germination from samples collected in the native reference plot (Fig. 5). Phragmites emergence was significantly lower in the reference plot than the Phragmites removal and control plots in 2011 (Fig. 5), and there were no significant differences in Phragmites germination in 2012 or 2013 (Table 1). Species richness was highest in the native reference plot, and all plots showed the lowest richness in 2013 (N plot values 2011 4.1 ± 0.42; 2012 4.5 ± 0.63; 2013 2.9 ± 0.47; Fig. 6).
Severn River Phragmites germination was significantly higher in the removal plot in 2012 and 2013 and negligible in the reference plot in all years (Table 1). Total germination was significantly different between the plots in 2011 and 2013, with the removal plot having the lowest emergence (Table 4; Fig. 5). Species richness was not significantly different in 2011, yet richness precipitously decreased in the removal plot in 2012 and 2013 (N plot values 2011 3.8 ± 0.47; 2012 3.1 ± 0.47; 2013 3.1 ± 0.38; Fig. 6).
St. Leonard total emergence was not significantly different between plots in any of the 3 years (Table 4; Fig. 5). Phragmites germination was significantly lower in the native reference plot in all years (Table 1; Fig. 5). Species richness was lowest in the reference plot in all years, and highest in the removal plot after herbicide application (N plot values 2011 1.6 ± 0.5; 2012 1.2 ± 0.54; 2013 1.1 ± 0.41; Fig. 6).
The total germination and Phragmites germination were both significantly different across plots in 2011 for the Wicomico River (Fig. 5), and Phragmites was a comparably minor component of the total seedbank composition relative to the other subestuaries. The total germination rate was lowest in the control plot in years 2011 and 2013, which also had the highest Phragmites emergence (Fig. 5). Species richness was not different between the plots (N plot values 2011 4.7 ± 0.54; 2012 1.9 ± 0.51; 2013 1.8 ± 0.35; Fig. 6), which was reflected in the guild analysis.