Introduce seeds of non-woody plants: freshwater wetlands
Overall effectiveness category Likely to be beneficial
Number of studies: 13
Background information and definitions
This action involves introducing seeds of emergent plants to restore/create marshes or swamps. Seeds may be collected from plants in greenhouses/laboratories, or from natural sites. They may be sown directly into the soil, scattered over the surface, or carried to suitable sites by water (e.g. after dropping them into the sea during an incoming tide).
Introduction of target vegetation might be useful in severely degraded or bare sites – which may lack remnant plants or seed banks to kick start revegetation with desirable species, and may be at risk of being taken over by undesirable species (Brown & Bedford 1997). It might also be useful in isolated wetlands, far from sources of marsh or swamp plant propagules. Seeds are easier to handle than plants, and can be a cost-effective way to introduce vegetation to large areas – but they can be more susceptible to herbivory or being washed away (e.g. Schoenholz et al. 2001).
The effects of sowing may be highly dependent on the environmental conditions in each study. Questions you might ask when interpreting the evidence include: Is the study site degraded? Where and when were seeds introduced? Was there any intervention to improve conditions before planting? What were the conditions over the duration of the study?
The scope of this action does not include sowing nurse plants; sowing submerged or floating plants; sowing to restore bogs, fens, fen meadows or peat swamp forests (see Taylor et al. 2018); or sowing facultative wetland plants in upland sites. In contrast, the scope does include sowing non-native species to conserve marshes or swamps – whilst acknowledging that this is often considered ethically unacceptable due to the risk of invasion (e.g. Ren et al. 2009).
Related actions: Directly plant whole plants; Introduce vegetation fragments; Transplant or replace blocks of vegetation; Transplant or replace wetland soil; Restore/create marshes or swamps using multiple interventions, often including planting.
Brown S.C. & Bedford B.L. (1997) Restoration of wetland vegetation with transplanted wetland soil: an experimental study. Wetlands, 17, 424–437.
Ren H., Lu H., Shen W., Huang C., Guo Q., Li Z. & Jian S. (2009) Sonneratia apetala Buch.Ham in the mangrove ecosystems of China: an invasive species or restoration species? Ecological Engineering, 35, 1243–1248.
Schoenholz S.H., James J.P., Kaminski R.M., Leopold B.D. & Ezell A.W. (2001) Afforestation of bottomland hardwoods in the Lower Mississippi Alluvial Valley: status and trends. Wetlands, 21, 602–613.
Taylor N.G., Grillas P. & Sutherland W.J. (2018) Peatland Conservation: Global Evidence for the Effects of Interventions to Conserve Peatland Vegetation. Synopses of Conservation Evidence Series. University of Cambridge, Cambridge.
Supporting evidence from individual studies
A replicated, site comparison study in 1989–1992 of 10 created wetlands in Wisconsin, USA (Reinartz & Warne 1993) reported that wetlands sown with herb (and some shrub) seeds contained a different plant community to unsown wetlands, with greater richness and diversity but similar vegetation cover. Unless specified, statistical significance was not assessed. After 1–2 years, the overall plant community composition differed between sown and unsown wetlands (data reported as a graphical analysis). Sown wetlands contained more plant species than unsown wetlands (sown: 46–56; unsown: 40–42 species/wetland), contained more native wetland plant species (sown: 25–42; unsown: 18–20 species/wetland) and had higher plant diversity (total and native wetland plants; data reported as a diversity index). However, vegetation cover did not significantly differ between sown and unsown wetlands. This was true for cover of all vegetation (sown: 40–73%; unsown: 29–67%) and cover of native wetland species only (sown: 21–59%; unsown: 6–41%). After two years, 17 of the 21 sown herb species were found in at least two of five sown wetlands. Meanwhile, the most abundant species in both types of wetland was cattail Typha spp. (sown: 13% cover; unsown: 17% cover; see original paper for data on abundance of other species). Methods: In autumn 1989 or 1990, ten areas of agricultural land (<2.2 ha) were flooded by blocking or removing drainage channels. Five were also sown with a mix of 22 wetland plant species (21 herbs, 1 shrub). When each wetland was one or two years old, all plant species were recorded and vegetation cover was estimated in twenty-five 1-m2 quadrats.Study and other actions tested
A replicated, before-and-after study in 1998–1999 in two experimental wet basins in Minnesota, USA (Green & Galatowitsch 2002) reported that 11 of 11 sown wetland herb species established. After 15 months, all 11 sown species were present as plants. The most abundant species was mannagrass Glyceria grandis (above-ground biomass: 248–681 g/m2). Total above-ground biomass was 1,915–3,079 g/m2 (including native grass-like plants: 366–1,252 g/m2; native forbs: 386–1,932 g/m2). Methods: In May 1998, seeds of 11 native sedge meadow grass-like plants and forbs were sown into sixty 1.13-m2 plots across two saturated wet basins (equal mix of all 11 species in each plot; total 1,500 viable seeds/m2). Seeds were dipped in bleach, then stored cold (4°C) and wet for 46 days before sowing. In October 1997, the plots had been levelled, enclosed in a plastic barrier, and treated with a chemical to kill all seeds in the soil. For experimental reasons, 30 plots also received seeds of invasive reed canarygrass Phalaris arundinacea (136 viable seeds/m2) and 40 plots were fertilized to simulate pollution. Vegetation was cut from one 0.5-m2 quadrat/plot in August 1999, then dried and weighed. This study used the same site as (3) and (7), but a different experimental set-up.Study and other actions tested
A replicated, before-and-after study in 1998–1999 in an experimental wet basin in Minnesota, USA (Perry & Galatowitsch 2003) reported that plots sown with porcupine sedge Carex hystericina seeds supported porcupine sedge populations after 1–2 growing seasons. In plots only sown with porcupine sedge, above-ground biomass was <1–16 g/m2 after one growing season, then 0–1,790 g/m2 after two. In plots sown with other species alongside porcupine sedge (potential nurse plants and/or an invasive grass), sedge biomass was 0–3 g/m2 after one growing season, then 0–1,130 g/m2 after two. Variation was related to elevation (less biomass in higher, drier plots) and which companion species were planted. Methods: In June 1997 and April 1998, porcupine sedge seeds were sown onto four hundred and eighty 0.25-m2 plots in an experimental wet basin (500–5,000 seeds/m2). Plots were 2–37 cm above the water level. Sedge seeds were dipped in bleach, then stored cold (4°C) and wet for eight weeks before sowing. For experimental reasons, 432 plots were also sown with also one or two other plant species. Biomass was sampled from the centre of the plots – half after one growing season, half after two – then dried and weighed. This study used the same site as (2) and (7), but a different experimental set-up.Study and other actions tested
A replicated, randomized, paired, controlled, before-and-after study in 1999–2000 in a floodplain marsh in the Northern Territory, Australia (Paynter 2004) reported that only two of five sown herb species were present after one year, and found that sowing had no significant effect on vegetation cover. After one year, the only two sown species present in any sown plots were wick grass Hymenachne acutigluma (in 1 of 35 sown plots; approximately 1% cover) and water chestnut Eleocharis dulcis (in 5 of 20 sown plots; cover not reported). Wick grass was present in 0 of 15 unsown plots. Water chestnut was present in 1 of 10 unsown plots. Sown and unsown plots had statistically similar cover of vegetation overall (approximately 76–90%), sedges and grasses (approximately 12–27%) and invasive mimosa Mimosa pigra (approximately 10–17%). Before sowing, plots destined for each treatment had similar cover of vegetation (<1%), dead mimosa stumps (5–15%) and bare mud (85–95%). Methods: In November 1999 (end of the dry season), herb seeds (collected from local wetlands) were sown into 5 x 5 m plots in a degraded floodplain marsh. Mimosa had recently been cleared from the marsh using herbicide, crushing and burning. In one area, ten sets of three plots were established. Twenty plots (two random plots/set) were sown with seeds of five mixed species, including wick grass and water chestnut (1 g/species). In the other area, five sets of four plots were established. Fifteen plots (three random plots/set) were sown with wick grass seeds (1.25 g, 5 g or 12.5 g). All other plots were not sown. Vegetation was surveyed immediately before sowing and approximately one year after (October 2000). This study used the same marsh as (5), but different experimental set-ups.Study and other actions tested
A replicated, randomized, paired, controlled, before-and-after study in 2000–2003 in a floodplain marsh in the Northern Territory, Australia (Paynter 2004) found that sowing seeds of three wetland herb species had no significant effect on their abundance or overall vegetation cover, or on germination rates of invasive mimosa Mimosa pigra. Immediately before sowing, plots had no vegetation cover. After one year, only one of three sown species (wick grass Hymenachne acutigluma) was present in sown plots. However, wick grass was present in the same proportion (17%) of sown and unsown plots. In three of three years after planting, sown and unsown plots had statistically similar cover of wick grass (sown: <2%; unsown: <2%), grasses/sedges overall (sown: 34–52%; unsown: 41–45%) and vegetation overall (sown: 40–76%; unsown: 50–76%). Finally, mimosa germination rates did not significantly differ between sown and unsown plots in any of the three years after sowing (see original paper for data). Methods: In July–September 2000 (at the end of the wet season), twelve pairs of 7.5 x 7.5 m plots were established on a degraded floodplain marsh. Mimosa had recently been cleared from the marsh using herbicide, crushing and burning. Then, one plot in each pair was sown with an equal mix of three herb species (2,667 g/ha of seeds collected from local wetlands). The other plots were not sown. Vegetation was surveyed immediately before sowing and in the following three dry seasons (July–October 2001–2003). This study used the same marsh as (4), but a different experimental set-up.Study and other actions tested
A replicated study in 2000–2004 in two wet meadows in Minnesota, USA (Reinhardt Adams & Galatowitsch 2006) reported that 15 of 26 sown herb species established. For these 15 species, some biomass was found in at least one sown plot after one and/or two growing seasons. In every plot, the total biomass of sown species was <10% of the biomass of species that had not been sown (excluding invasive reed canarygrass Phalaris arundinacea). Sown species only established in plots where reed canarygrass had been controlled with herbicide before sowing. Methods: One hundred and sixty 25-m2 plots were established across two canarygrass-invaded wet meadows. All plots were sown with a mix of grass and forb seeds in May 2001, 2002 or 2003 (further details not reported). Most (140) plots had been burned and/or sprayed with herbicide in the year(s) before sowing to control reed canarygrass. Vegetation was surveyed in August, one and two growing seasons after sowing.Study and other actions tested
A replicated, before-and-after study in 2004–2005 in two experimental wet basins in Minnesota, USA (Iannone & Galatowitsch 2008) reported that 33–61% of sown sedge meadow plant seeds germinated depending on the presence/diversity of a nurse crop, and that vegetation abundance after one growing season depended on the presence/diversity of a nurse crop, presence of an invasive plant species, sawdust addition and the outcome metric. For example, the total density of target (sown) sedge meadow species was lowest (370 shoots/m2) in plots with a high-diversity nurse crop and reed canarygrass Phararis arundinacea, but without added sawdust, and highest (1,300 shoots/m2) in plots without a nurse crop, but with reed canarygrass and added sawdust. The density of individual sown species ranged from 0 shoots/m2 (e.g. great blue lobelia Lobelia siphilitica under all conditions) to 490 shoots/m2 (prairie ironweed Vernonia fasciculata under one set of conditions). Methods: In May 2005, seeds of ten target sedge meadow species were sown onto seventy-two 1-m2 plots (total 2,250 seeds/m2) across two experimental, vegetation-free wet basins. The seeds were stored cold (4°C) and wet for four months before sowing. All plots were weeded for 10 weeks after sowing. For experimental reasons, 48 plots were also sown with a potential nurse crop (one or five species), 36 plots were sown with reed canarygrass, and 36 plots were amended with sawdust before sowing. Target vegetation was surveyed for 16 weeks after sowing. Seedlings were counted in five 100-cm2 subplots/plot. Shoot density and cover were monitored across the whole of each plot. This study used the same site as (2) and (3), but a different experimental set-up.Study and other actions tested
A replicated, controlled, site comparison study in 1999–2008 in 256 excavated ephemeral pools on one air force base in California, USA (Collinge & Ray 2009) found that plots sown with seeds of five native, pool-characteristic herb species contained a greater abundance of these species than unsown plots. In seven of seven years, the combined frequency of the five pool-characteristic plants was greater in sown plots (3–19%) than unsown plots (<1–5%). The same was true in 30 of 35 comparisons for the individual species (sown: <1–44%; unsown: <0–21%). For comparison, the frequency of each species in nearby natural pools was 5–48%. Three of four analyzed species also strongly benefitted from “priority effects”: they were more frequent in pools where they were sown in the first year of the study than in pools where they were sown in the second year, after other species (see original paper for data). Methods: Between 1999 and 2001, seeds of five focal herb species (native species characteristic of Californian ephemeral pools) were sown onto 192 plots (each 0.25 m2 and in a separate excavated pool). Of these, 128 were sown with a mix of five species (600 seeds/plot, over 1–2 years) and 64 were sown with a single species (100–300 seeds/plot, over 1–3 years). The mixed-species plots received one of two planting orders (species A+B+C then species A+D+E, or species A+D+E then species A+B+C). Sixty-four additional plots (pools) were not sown. Each spring between 2002 and 2008, the frequency of the five focal species was recorded in each plot, using a grid of one hundred 2.5-cm2 cells. Some natural pools (number not specified) on the base were also surveyed in 1998 and 1999. This study was based on the same pools as (10).Study and other actions tested
A replicated study in 2005–2006 of 22 lakeshore restoration sites in Minnesota, USA (Vanderbosch & Galatowitsch 2010) reported that 17–40% of sown/planted species reliably established across multiple sites, and that no planted/sown species established in some individual sites. In the seasonally flooded zone, only 22 of 128 sown/planted species reliably established (survived in >75% of sites where planted, or ≥25% cover in ≥1 site). Fifty-six species failed to establish at any site. However, some sown/planted species established at 100% of sites. In the permanently flooded zone, 10 of 25 sown/planted species reliably established. Six species failed to establish at any site. Sown/planted species completely failed to establish at 27% of sites. Methods: In summer 2005 and spring 2006, plant species and their cover were surveyed in 22 urban lakeshore restoration projects. Native plants had been introduced between 1999 and 2004. Species lists were obtained from project reports or interviews with staff. Almost all introduced plants were emergent herbs, and most (but not all) were wetland species. Some plants were sown and some were directly planted (as plugs or on pre-vegetated coconut-fibre mats). The study does not distinguish between the effects of sowing and planting. Most sites were protected with fences and/or wave breaks, at least for the first growing season after sowing/planting.Study and other actions tested
A replicated, controlled study in 1999–2008 in 256 excavated ephemeral pools on one air force base in California, USA (Collinge et al. 2011) found that plots sown with seeds of pool-characteristic herbs typically contained a greater abundance of native pool-characteristic plants than unsown plots (if a dense seed mix was used), but that sowing did not significantly affect the abundance of non-native plants. All data were reported as frequencies, added together for all species in each group. Over seven years of monitoring, plots densely sown with a mix of herb species typically supported a greater abundance of native, pool-characteristic plants than unsown plots (9 of 14 comparisons; other comparisons no significant difference). However, densely sown plots typically supported a similar abundance of non-native plants to unsown plots (9 of 14 comparisons; other comparisons lower abundance in sown plots). The study does not report data for native, generalist plants. In contrast, plots sparsely sown with single species typically supported a similar abundance – to unsown plots – of both native pool-characteristic plants (13 of 14 comparisons) and non-native plants (14 of 14 comparisons). Methods: Between 1999 and 2001, seeds of native, pool-characteristic herbs were sown onto 192 plots (each 0.25 m2 and in a separate excavated pool). Of these, 128 were densely sown (600 seeds/plot; mix of five species) and 64 were sparsely sown (100–300 seeds/plot; one species). Sixty-four additional plots (pools) were not sown. Each spring between 2002 and 2008, the frequency of every plant species was recorded in each plot, using a grid of one hundred 2.5-cm2 cells. This study was based on the same pools as (8).Study and other actions tested
A replicated study in 2003–2004 in six wet meadows in Iowa, USA (Kettenring & Galatowitsch 2011) reported that <1–21% of sown sedge Carex spp. seeds germinated within two growing seasons. A higher proportion of seeds germinated in recently rewetted meadows (7–21%) than natural meadows (<1–4%). For seeds sown in natural meadows in the spring, a higher proportion germinated when chilled over the previous winter than when kept at room temperature (see Action: Chill seeds before sowing). Within each wetland type and seed treatment, germination rate did not significantly differ between species (see original paper). Methods: In autumn 2002 and spring 2003, seeds of 4–5 sedge species were sown into the wet meadow zone of six prairie pothole wetlands (900–8,100 wild-collected seeds/species/pothole, split across 9–27 plots/species). Three meadows were natural and three had been rewetted one year previously (so were still developing vegetation, and had drier soil with no sedge seeds). Autumn-sown seeds were held in place with plastic mesh and barriers. Amongst spring-sown seeds, half had been chilled (1–5°C) over the previous winter whilst half had been kept at room temperature. Seedlings that emerged from the soil were counted for two growing seasons.Study and other actions tested
A replicated, before-and-after study in 2010–2013 aiming to restore an ephemeral freshwater marsh on cropland in South Dakota, USA (Zilverberg et al. 2014) reported that sown prairie cordgrass Spartina pectinata occurred in 0–67% of sampled quadrats after two growing seasons and 31–78% of sampled quadrats after four, depending on elevation. After two growing seasons, 0–10% of quadrats at low elevations (≤10 cm from wetland bottom) and 57–67% of quadrats at higher elevations (>10 cm from wetland bottom) contained at least one cordgrass stem. After four growing seasons, cordgrass plants had spread (possibly from adjacent plots with transplanted cordgrass). There was at least one stem in 15–31% of quadrats at low elevations and 66–78% of quadrats at higher elevations. The height and above-ground biomass of cordgrass were greatest at mid-low elevations (see original paper). Methods: Four plots were established in a historically cultivated ephemeral wetland. Each plot ran perpendicular to the slope of the wetland, so included a range of elevations. Spring floodwaters were typically 50 cm deep. In spring 2010, each plot was sown with cordgrass seed (10 kg/ha). All plots were mown once in 2011 to control weeds. Each autumn from 2011 to 2013, cordgrass presence and height were surveyed in 1-m2 quadrats along the length of each plot. Biomass was sampled in 2013 only.Study and other actions tested
A replicated study in 2013–2014 in a degraded floodplain swamp in Florida, USA (Smith et al. 2016) reported that seeds of six wetland herb species did not germinate within a year of sowing, when sown into a clearing. Methods: In November 2013, herb seeds were sown (600 viable seeds/m2) into fourteen 1.5 x 1.5 m plots, in a clearing in a floodplain swamp. Seven plots received a mix of four native species (common rush Juncus effusus, pine barren goldenrod Solidago fistulosa, purple bluestem Andropogon glomeratus, and red-top panic grass Panicum longifolium). Seven plots received a mix of two native species (common rush and goldenrod). Three months before sowing, all plots were sprayed with herbicide to control, but not eradicate, invasive Mexican petunia Ruellia simplex. Plots were monitored monthly for one year after sowing. Surface water was present in 6 of 12 months and was up to 21 cm deep.Study and other actions tested