Use herbicide to control problematic plants: freshwater marshes
Overall effectiveness category Trade-off between benefit and harms
Number of studies: 17
Background information and definitions
Herbicides can be applied to an entire area of vegetation, or targeted at individual problematic species (e.g. by painting onto individual plants, or shielding non-target vegetation). Herbicides could be applied once, or repeatedly to kill established vegetation or recurrent growth from the seed bank. To maximize contact with target species and minimize non-target effects, herbicides might be applied during/at the start of the dry season, as the tide is going out, or on calm rather than windy days (e.g. Tobias et al. 2016). Often, herbicide application will follow or be followed by other interventions, such as mowing, burning or physical removal of problematic plants.
Caution: In many herbicides, the active chemicals are not specific to the problematic species so can cause collateral damage to desirable species. Relying on herbicides as the only tool to manage problematic plants can lead to the development of herbicide resistance in future generations (Powles et al. 1997). Herbicides can have severe negative side effects on biodiversity, the environment and human health (Pimentel et al. 1992). Accordingly, herbicide use – particularly in or near wetlands or water bodies – is limited in many countries.
Bear in mind that the effects of herbicide might be highly dependent on the chemical used, how it is applied (e.g. season and number of applications), and local site conditions (e.g. nutrient availability, water levels, proximity of untreated invaded vegetation) (Tobias et al. 2016). Also, similarity between treated and untreated, degraded areas might not be an undesirable outcome for this action: similarity in vegetation cover after months or years could suggest, for example, that native vegetation abundance has recovered after being initially depressed by herbicide.
For this action, “vegetation” refers to overall or non-target vegetation. Studies that only report responses of target problematic plants have not been summarized.
Pimentel D., Acquay H., Biltonen M., Rice P., Silva M., Nelson J., Lipner V., Giordano S., Horowitz A. & D’Amore M. (1992) Environmental and economic costs of pesticide use. BioScience, 42, 750–760.
Powles S.B., Preston C., Bryan I.B. & Jutsum A.R. (1997) Herbicide resistance: impact and management. Advances in Agronomy, 58, 57–93.
Tobias V.D., Block G. & Laca E.A. (2016) Controlling perennial pepperweed (Lepidium latifolium) in a brackish tidal marsh. Wetlands Ecology and Management, 24, 411–418.
Supporting evidence from individual studies
A replicated, randomized, paired, controlled study in 1991–1994 in a floating freshwater marsh invaded by purple loosestrife Lythrum salicaria in Ontario, Canada (Gabor et al. 1996) found that the effect of spraying herbicide on the density of sedges, grasses and purple loosestrife depended on the dose. After 2–3 years, the density of sedges Carex spp. did not significantly differ between sprayed plots (68–322 stems/m2) and unsprayed plots (150–199 stems/m2). Grass density did not significantly differ between plots sprayed with low-medium herbicide doses (sprayed: 80–177 stems/m2) and unsprayed plots (47–65 stems/m2). However, it was significantly greater (vs no spraying) in plots sprayed with high herbicide doses (124–161 stems/m2). Purple loosestrife density did not significantly differ between plots sprayed with low herbicide doses (27–41 stems/m2) and unsprayed plots (36–56 stems/m2). However, it was significantly lower (vs no spraying) in plots sprayed with medium-high herbicide doses (1–15 stems/m2). Methods: In summer 1991, twelve 30-m2 plots were established, in three blocks of four, in a loosestrife-invaded marsh. Three plots (one random plot/block) were sprayed with each dose of triclopyr amine: low (4 kg/m2), medium (8 kg/m2) or high (12 kg/m2). The other three plots were not sprayed. In July and August 1993 and 1994, plant stems were counted in three 1-m2 quadrats/plot.Study and other actions tested
A replicated, randomized, controlled, before-and-after study in 1990–1993 of 17 freshwater marshes dominated by cattails Typha spp. in North Dakota, USA (Linz et al. 1996) found that marshes sprayed with herbicide had lower overall and live vegetation coverage than unsprayed marshes, but greater coverage of dead vegetation. After 1–2 years, coverage of emergent vegetation was significantly lower in sprayed marshes (70% of marsh area) than in unsprayed marshes (88% of marsh area). Sprayed marshes also had lower coverage of live vegetation (sprayed: 29%; unsprayed: 70%) but greater coverage of dead vegetation (sprayed: 40%; unsprayed: 17%). Before intervention, marshes destined for each treatment had statistically similar coverage of emergent vegetation (sprayed: 84–90%; unsprayed: 89%; live and dead not separated). Methods: In July 1990 and 1991, glyphosate herbicide (Rodeo®) was sprayed on to a total of 12 cattail-dominated marshes (5.8 L/ha across 50–90% of each marsh). Five similar marshes were left unsprayed. Emergent vegetation coverage was estimated from aerial photographs of each marsh, taken before (June 1990) and 1–2 years after (August 1991–1993) intervention. This study used a subset of the marshes in (3).Study and other actions tested
A replicated, randomized, controlled, before-and-after study in 1990–1993 of 23 freshwater marshes dominated by cattails Typha spp. in North Dakota, USA (Linz et al. 1996) found that marshes sprayed with herbicide had similar overall vegetation coverage to unsprayed marshes, but less live vegetation and more dead vegetation. After 1–2 years, coverage of emergent vegetation did not significantly differ between sprayed marshes (61–81% of marsh area) and unsprayed marshes (76–85% of marsh area). However, sprayed marshes had lower coverage of live vegetation (sprayed: 14–39%; unsprayed: 61–69%) including cattails (sprayed: 31%; unsprayed: 65%), and greater coverage of dead vegetation (sprayed: 25–58%; unsprayed: 15–16%). Before intervention, marshes destined for each treatment had statistically similar coverage of emergent vegetation (sprayed: 70–91%; unsprayed: 87%; data not reported for live, dead and cattail coverage). Methods: In July 1990 and 1991, glyphosate herbicide (Rodeo®) was sprayed on to a total of 16 cattail-dominated marshes (5.8 L/ha across 50–90% of each marsh). Seven similar marshes were left unsprayed. Emergent vegetation coverage was estimated from aerial photographs of each marsh, taken before (June 1990) and 1–2 years after (August 1991–1993) intervention. Some of the marshes in this study were also used in (2).Study and other actions tested
A controlled, before-and-after study in 1995–1998 in a freshwater marsh dominated by common reed Phragmites australis in Connecticut, USA (Farnsworth & Meyerson 1999) found that applying herbicide to the vegetation (along with cutting/mowing) increased the evenness of the plant community and the abundance and richness of non-reed species. After three years, treated plots contained a more even plant community, less dominated by one or two species, than an untreated plot (data reported as a coefficient of variation; see original paper for data on individual species abundance). Treated plots also had greater plant species richness (excluding common reed; treated: 5–7 species/m2; untreated: 3 species/m2) and contained a greater density of non-reed stems (treated: 78–97 stems/m2; untreated: 15 stems/m2). Common reed was less abundant in treated plots, in terms of stem density (treated: 6–19 stems/m2; untreated: 36 stems/m2) and frequency (treated: 45–64%; untreated: 98% of surveyed quadrats contained common reed). Before intervention, all plots had relatively similar plant species richness (excluding common reed; 2–3 species/m2), non-reed density (7–23 stems/m2) and reed density (33–40 stems/m2). Methods: In 1995, two 0.4-ha plots were treated in a reed-dominated, tidal, freshwater marsh. In August, each plot was sprayed with herbicide (Rodeo® 1%). In autumn, one plot was cut by hand and one was mown mechanically; cuttings were left in place. A third adjacent plot was neither sprayed with herbicide nor cut/mown. The study does not distinguish between the effects of applying herbicide and cutting/mowing. In late summer before (1995) and after (1996–1998) intervention, plant stems were identified and counted in fifty 1-m2 quadrats/plot.Study and other actions tested
A replicated, randomized, paired, controlled, before-and-after study in 1988–1991 in two wet meadows that had been cleared of vegetation in New York State, USA (Morrison 2002) found that controlling regrowth of invasive purple loosestrife Lythrum salicaria (by applying herbicide to large shoots and pulling up seedlings) had no significant effect on plant species richness, diversity or vegetation cover. After three years, plots with and without control of loosestrife regrowth had statistically similar plant species richness (control: 7; no control: 8 species/m2), plant diversity (data reported as a diversity index), total vegetation cover (control: 67–82%; no control: 79%), grass-like plant cover (control: 60–75%; no control: 70–73%) and forb cover (control: 5–20%; no control: 8–10%). Purple loosestrife cover was 0% in plots where regrowth had been controlled, but still only 2% in plots where regrowth had not been controlled. For data on the cover of other individual plant species, see original paper. Before intervention and within each meadow, plots destined for each treatment had statistically similar plant species richness (8–9 species/m2), plant diversity, total vegetation cover (103–143%), grass-like plant cover (16–58%), forb cover (25–56%) and purple loosestrife cover (23–63%). Methods: In 1988, six pairs of 1-m2 plots were established across two loosestrife-invaded wet meadows. In September, all vegetation was dug up and removed from the plots (see Action: Physically remove problematic plants). In six of the plots (one random plot/pair), loosestrife regrowth was controlled twice/year thereafter (painting large shoots with glyphosate and pulling up seedlings; the study does not distinguish between the effects of these interventions). In the other plots, loosestrife regrowth was not controlled. Plant species and their cover were surveyed before initial removal (August 1988) and three years after (September 1991).Study and other actions tested
A study in 1998–2003 in a degraded floodplain marsh in the Northern Territory, Australia (Paynter 2004) reported that following herbicide application, physical damage and prescribed burning to control invasive mimosa Mimosa pigra, some herbaceous plants recolonized the site along with mimosa. After one year, cover of all vegetation other than mimosa was approximately 31–80%. This included 12–45% total cover of grasses/sedges. Mimosa cover was approximately 0–17%, depending on the area within the marsh. The number of new mimosa seedlings each year declined over time, from 1 seedling/m2 in the first year after intervention was complete, to <0.5 seedlings/m2 in the second and third years, then 0 seedlings/m2 in the fourth year. Methods: Three interventions were applied to a 100-ha patch of mimosa-dominated floodplain. In April 1998, the site was sprayed with herbicide (metsulfuron methyl). In October 1999, the dead vegetation was crushed using a chain tied between two bulldozers, then the site was burned (fire lasting several days). The study does not distinguish between the effects of these interventions. Vegetation was surveyed in the dry season (July–October), in up to three areas of the marsh (where no vegetation had been introduced) and for up to four years after intervention was complete. This study was in the same area as (7), but used a different experimental set-up.Study and other actions tested
A replicated, randomized, paired, controlled, before-and-after study in 1997–1999 in a floodplain wetland invaded by mimosa Mimosa pigra in the Northern Territory, Australia (Paynter & Flanagan 2004) found that spraying the vegetation with herbicide did not reduce the cover of non-mimosa vegetation 1–2 years later. In three of six comparisons, non-mimosa cover was higher in sprayed plots (55–74%) than unsprayed plots (15–38%). In the other three comparisons, there was no significant difference between treatments (sprayed: 32–47%; unsprayed: 15–38%). Sprayed plots consistently had lower mimosa coverage (six of six comparisons) and density (six of six comparisons). Results for mimosa biomass were mixed, but never significantly higher in sprayed than unsprayed plots (see original paper for data). Before intervention, the abundance of both mimosa and other vegetation were statistically similar in plots destined for each treatment (data not reported). Methods: In April 1998, thirty-two 100 x 200 m plots were established, in four sets of eight, on a mimosa-invaded floodplain. Twenty-four plots (six random plots/set) were sprayed with herbicide in April 1998 and/or January 1999. In half of the plots, vegetation was also crushed with bulldozers in late 1998. Vegetation was surveyed one year before spraying (late 1997/early 1998) and approximately 1–2 years after the latest spray (late 1999), in four 1–5 m2 quadrats/plot and by aerial photography (mimosa coverage). This study was in the same area as (6), but used a different experimental set-up.Study and other actions tested
A replicated, randomized, paired, controlled study in a freshwater marsh invaded by antelope grass Echinochloa pyramidalis in eastern Mexico (López Rosas et al. 2006) found that spraying the vegetation with herbicide had no significant effect on overall plant density, richness or diversity, the relative abundance of common plant species, or the absolute abundance of common native plant species. After 4–8 months, sprayed and unsprayed plots contained a statistically similar overall plant density (six of six comparisons; sprayed: 57–81; unsprayed: 54–93 plants/0.49 m2), species richness (six of six comparisons; sprayed: 5–7; unsprayed: 3–5 species/0.49 m2) and diversity (two of two comparisons; data reported as a diversity index). Accordingly, all six monitored plant species had a similar relative abundance in sprayed and unsprayed plots (five native species, plus antelope grass; see original paper for data). The five native plant species had statistically similar cover in sprayed and unsprayed plots in 13 of 14 comparisons (both treatments: 0–19% cover/species). In contrast, antelope grass had lower cover in sprayed plots in four of six comparisons (for which sprayed: 22–78%; unsprayed: 99–100%). Methods: In January (year not reported), twenty-one 0.49-m2 plots were established (in seven sets of three) in a degraded marsh, invaded by antelope grass. Fourteen plots (two random plots/set) were sprayed with glyphosate herbicide (Roundup®). In seven of these, the most abundant native plant species was shielded with plastic tubes. The final seven plots (one random plot/set) were not sprayed. All 21 plots were enclosed, underground, by a plastic barrier. Vegetation was surveyed between May and September later that year (relative biomass in September only).Study and other actions tested
A replicated, randomized, paired, controlled study in 2000–2004 in two wet meadows invaded by reed canarygrass Phalaris arundinacea in Minnesota, USA (Reinhardt Adams & Galatowitsch 2006) found that plots sprayed with herbicide contained less overall plant biomass than unsprayed plots after 2–3 growing seasons, but more non-canarygrass plant biomass. Two to three growing seasons after the last herbicide application, sprayed plots contained less total above-ground plant biomass (320–720 g/m2) than unsprayed plots (520–900 g/m2). Sprayed plots contained less reed canarygrass biomass (10–480 g/m2) than unsprayed plots (420–880 g/m2). However, they contained more biomass of other plants. This was true for total biomass of sown species (sprayed: 0–70 g/m2; unsprayed: 0 g/m2) and species that had not been sown (sprayed: 170–550 g/m2; unsprayed: 30–100 g/m2). Methods: In the early 2000s, one hundred and sixty 25-m2 plots were established, in 40 sets of four, across two canarygrass-invaded wet meadows. One hundred and twenty plots (three random plots/set) were sprayed with herbicide (Roundup® Ultra): in late May, August or September and in one or two years. The remaining 40 plots were not sprayed. Half of the plots under each herbicide treatment were also burned in mid-May. All plots were sown with a mixture of grass and forb seeds in the spring after the final herbicide application. Dry biomass samples were taken in August in the two years after herbicide application.Study and other actions tested
A replicated, randomized, paired, controlled, before-and-after study in 2004–2005 in two freshwater marshes invaded by alligatorweed Alternanthera philoxeroides in Alabama and Georgia, USA (Allen et al. 2007) found that spraying the vegetation with herbicide had no significant effect on native plant biomass after 1–2 growing seasons. Native plant biomass varied a lot depending on herbicide type, dose and application date, but was statistically similar in sprayed and unsprayed plots in 24 of 24 paired comparisons (sprayed: <1–210 g/0.25 m2; unsprayed: 76–129 g/0.25 m2). Alligatorweed biomass did not significantly differ between treatments in 14 of 24 comparisons (for which sprayed: 18–92 g/0.25 m2; unsprayed: 54–78 g/0.25 m2) but was lower in sprayed plots in the other 10 comparisons (for which sprayed: <1–22 g/0.25 m2; unsprayed: 54–78 g/0.25 m2). The study also provided short-term data on alligatorweed cover. This was depressed after 2–4 weeks for all herbicide types, doses and application dates (sprayed: 0–32%; unsprayed: 25–69%; before spraying: 17–62%). Methods: Sixty-four 5 x 5 m plots (in four sets of 16) were established across two alligatorweed-invaded freshwater marshes, managed for waterfowl. Herbicide was applied to 48 of the plots (12 random plots/set), in all possible combinations of herbicide type (triclopyr amine or imazapyr), dose (low, medium or high) and application date (April or July 2004). Alligatorweed cover was surveyed one week before spraying and for 12 weeks after. Vegetation was cut from plots, then dried and weighed, in October 2004 and 2005.Study and other actions tested
A study in 2005–2007 of a dune slack in England, UK (Smith & Kimpton 2008) reported that after cutting and applying herbicide to grey willow Salix cinerea scrub, ground vegetation recolonized. In 2006, approximately one year after removing willows, 80% of the site was covered with vegetation (mostly herbaceous). There were 108 vascular plant taxa, including 98 natives. Approximately 54 taxa were characteristic of dune slacks. In 2007, approximately two years after removing willows, 95% of the site was covered with vegetation (still mostly herbaceous). There were 111 vascular plant taxa, including 107 natives. Approximately 65 taxa were characteristic of dune slacks. Twenty-eight taxa recorded in 2006 were not present in 2007, but 31 new taxa had colonized the site. Methods: In November/December 2005, dense grey willow scrub in a dune slack (low-lying area amongst dunes) was controlled. Grey willows were cut at ground level, then herbicide (Roundup® Biactive Plus) was applied to the largest stumps. The study does not distinguish between the effects of these interventions. Cut material was burned on site. Vascular plant taxa and their overall coverage were surveyed in August/September 2006 and 2007.Study and other actions tested
A replicated, randomized, paired, controlled study in 2006–2008 in a wet meadow being invaded by hybrid cattail Typha x glauca in Wisconsin, USA (Hall & Zedler 2010) found that spraying cattail with herbicide had no significant effect on cover of sedges Carex spp. after two growing seasons. Plots where cattails had been sprayed with herbicide had statistically similar sedge cover (14–29%) to unsprayed plots (11–38%). No sedge seedlings were found in any plot. Methods: Sixteen 4 x 8 m plots were established (in two sets of eight) on the boundary between native wet meadow vegetation and a patch of hybrid cattail. In May 2006, cattail plants in eight plots (four random plots/set) were sprayed with herbicide (Rodeo® 0.75%). The other eight plots were not sprayed. Sedge cover was surveyed in October 2007, in four 1-m2 quadrats/plot.Study and other actions tested
A replicated, randomized, controlled study in 2007–2008 in a freshwater marsh invaded by common reed Phragmites australis in Ohio, USA (Back et al. 2012) reported that plots sprayed with herbicide were more likely to contain free-growing filamentous algae than unsprayed plots, but found that all plots contained a similar abundance, diversity and community of biofilm algae. After approximately one year, free-growing algae occurred in 13 of 30 samples in sprayed plots (vs 1 of 15 samples in unsprayed plots; statistical significance not assessed). Meanwhile, biofilm algae reached a statistically similar abundance in sprayed and unsprayed plots. This was true for both density (sprayed: 1,700–2,800 cells/cm2; unsprayed: 1,100–1,700 cells/100 cm2) and biomass (sprayed: 5–41 μg chlorophyll/cm2; unsprayed: 5–41 μg chlorophyll/cm2). Sprayed and unsprayed plots also supported a similar diversity of biofilm algae (data reported as a diversity index) and a similar community composition of the most abundant group: diatoms (data reported as a graphical analysis; statistical significance of difference not assessed). Common reed was less abundant in sprayed than unsprayed plots, in terms of both density (sprayed: 2–3 live stems/m2; unsprayed: 36 live stems/m2) and cover (sprayed: 1–3%; unsprayed: 49%). Methods: In June 2007, fifteen contiguous 20 x 20 m plots were established in a reed-invaded, lakeshore marsh. Ten random plots were sprayed once with herbicide (five with glyphosate-based AquaNeat®; five with imazapyr-based Habitat®). The other five plots were not sprayed. Vegetation was surveyed in June–August 2008. Free-growing algae were surveyed in 10 x 10 cm quadrats. Biofilms were surveyed on fallen, submerged reed stems. Reeds were surveyed, in 0.5-m2 quadrats, along a central transect in each plot.Study and other actions tested
A replicated, randomized, paired, controlled, before-and-after study in 2005–2008 in five wet meadows in South Dakota, USA (Bahm et al. 2014) found that controlling problematic plants by mowing, applying herbicide and planting native upland plants increased plant species richness and cover of unplanted native species. All plots were initially dominated by reed canarygrass Phalaris arundinacea (>80% cover). After 1–3 growing seasons, plant species richness was higher in treated than untreated plots in 19 of 21 comparisons (for which treated: 2–5 species/0.25 m2; untreated: 2 species/0.25 m2). Treated plots also had greater cover of unplanted native species in 17 of 21 comparisons (for which treated: 8–57%; untreated: 3–21%) and lower cover of reed canarygrass in 21 of 21 comparisons (treated: 1–66%; untreated: 91–93%). After 2–3 growing seasons, no treatment outperformed those involving imazapyr. Plots treated with imazapyr never had lower plant species richness and unplanted native cover than plots treated with other herbicides, and never had higher cover of reed canarygrass (see original paper for data). Methods: Forty 3 x 40 m plots were established across five canarygrass-invaded wet meadows (eight plots/meadow). Between autumn 2005 and spring 2006, thirty-five plots (seven random plots/set) were mown, sprayed with herbicide (seven chemical x timing combinations), and planted with 14 native upland species. Subsequent targeted mowing of “noxious weeds” was also carried out. The study does not distinguish between the effects of these interventions. Vegetation was surveyed at the end of each growing season 2006–2008, in nine 0.25-m2 quadrats/plot.Study and other actions tested
A replicated, controlled, before-and-after study in 2013–2014 in eight artificial marshes invaded by hybrid cattail Typha x glauca in Michigan, USA (Lawrence et al. 2016) found that applying herbicide to the vegetation reduced native plant richness, density and biomass. After one year, there were no living plants in marshes treated with herbicide: richness, density and biomass, of both native plants and hybrid cattail, were zero. All metrics were significantly lower than in untreated marshes (where native richness: 4.5 species/2 m2; native density: 300 stems/m2; native biomass: 440 g/m2; cattail density: 44 stems/m2; cattail biomass: 745 g/m2). Results were generally similar after one month, although native plant richness had not yet declined in treated marshes (see original paper). Before treatment, vegetation was statistically similar in all marshes (native plant richness: 3.5–4.3 species/2 m2; native plant density: 190–230 stems/m2; native plant biomass: 350–430 g/m2; cattail density: 66–69 stems/m2; cattail biomass: 1,080–1,130 g/m2). Methods: In July 2013, glyphosate-based herbicide was spread onto all plant stems in four experimental marshes (1 x 2 m area, 1 m soil depth). The marshes had been created in 2002 and planted with cattail (i.e. deliberately invaded) in 2004. Four additional marshes were left untreated. Plant species, density and height were recorded in all marshes immediately before, one month after and one year after treatment. Dry above-ground biomass was calculated from height measurements.Study and other actions tested
A before-and-after study in 2006–2010 of a floodplain wetland invaded by limpo grass Hemarthria altissima in Florida, USA (Toth 2016) reported that over the four years after applying herbicide, cover of wet-prairie indicator species and total plant species richness were typically higher than before intervention. Statistical significance was not assessed. In the spring before applying herbicide, the wetland had 96% limpo grass cover and <1% cover of native wet-prairie indicator species. There were 3 plant species/100 m2. Between one and four years after applying herbicide, limpo grass cover ranged from 2% to 22%. Indicator species cover ranged from 1% to 13%. There were between 13 and 30 plant species/100 m2. Methods: In May 2006, glyphosate herbicide was applied to 6 ha of a recently rewetted, limpo-grass-invaded floodplain. Plant species and their cover were surveyed in twelve 100-m2 plots, before intervention (spring 2006) and for approximately four years after (spring 2007–summer 2010). This study was in the same area as (17), but used a different plot.Study and other actions tested
A replicated, before-and-after, site comparison study in 2007–2010 of floodplain wet prairies invaded by limpo grass Hemarthria altissima in Florida, USA (Toth 2016) reported that applying herbicide reduced limpo grass cover and maintained cover of other wetland-characteristic herbs in some plots 2–3 years later, but failed to do so in others. Statistical significance was not assessed. In three of seven treated plots, applying herbicide reduced limpo grass cover (before: 47%; 2–3 years after: 4%). Cover of other wetland-characteristic herbs was similar before (21%) and after (18%) applying herbicide. In the other four of seven treated plots, applying herbicide failed to prevent an increase in limpo grass cover 2–3 years later (before: 47%; 2–3 years after: 59%). Meanwhile, cover of other wetland-characteristic herbs declined (before: 21%; after: 10%). Nearby untreated plots were always dominated by wetland-characteristic herbs (before: 54%; after: 43%) with little limpo grass cover (before: 5%; after: 1%). Across all seven treated plots, three key metrics increased over time after herbicide application, reaching similar or higher levels three years after intervention than before: cover of native wet-prairie indicator species (before: 2–5%; after: 9%), total vegetation cover (before: >80%; after: >80%), and total plant species richness (before: 13–17; after: 23 species/100 m2). Methods: Eighteen 100-m2 plots were established in wet prairies on a recently rewetted floodplain, with varying limpo grass cover. In autumn 2007, glyphosate herbicide was applied to seven plots with the greatest limpo grass cover. Plant species and their cover were surveyed in all 18 plots, before intervention (spring 2006–summer 2007) and for approximately three years after (spring 2008–summer 2010). This study was in the same area as (16), but used different plots.Study and other actions tested