INTRODUCTION

Reed canarygrass (Phalaris arundinacea L.) is a prolific invader that can easily spread throughout wetlands and create monospecific stands (Lavergne and Molofsky 2004; Adams and Galatowitsch 2005). These stands of pure reed canarygrass are undesirable because they create areas of low plant diversity (Kercher et al. 2004) and undesirable wildlife habitat (Kirsch et al. 2007; Evans-Peters et al. 2012). Reed canarygrass has been used as a forage source for livestock and has been used in environmental plantings to treat wastewater and control erosion. These uses have led to the spread of reed canarygrass across landscapes (Galatowitsch et al. 1999; Green and Galatowitsch 2002; Kercher and Zedler 2004; Adams and Galatowitsch 2005; Kim et al. 2006; Kidd and Yeakley 2015). Reed canarygrass can exist and thrive in a variety of growing conditions: (1) increased nitrate (N) in the soil, (2) high amounts of soil organic matter, (3) flooding conditions, (4) shade, and (5) heavy soil disturbance (Green and Galatowitsch 2002; Kellogg et al. 2003; Kercher and Zedler 2004). Not only is this grass tolerant of flooding but it can withstand fluctuating water levels (Galatowitsch et al. 2000). These characteristics are why reed canarygrass has been classified as an invasive species and a management problem for wetlands.

Several methods have been used to varying degrees of success to control the invasion of reed canarygrass. Herbicide application has been used with total control being elusive (Bahm et al. 2014). Replacement of reed canarygrass using cultivation accompanied with herbicide and planting of desirable vegetation has shown some success in reducing the monospecific stands but again total control was not achieved (Adams and Galatowitsch 2005; Kim et al. 2006). Fertilizing with nitrate was not effective at reducing reed canarygrass canopy cover (Green and Galatowitsch 2002) nor was burning (Adams and Galatowitsch 2006; Kim et al. 2006). The use of grazing as a control method had only been mentioned and prior to 2010 there was no information on the effects grazing has on controlling reed canarygrass.

The goal of this study was to investigate if grazing cattle could be an effective practice for reducing the canopy cover of reed canarygrass in two restored wetland complexes in western Minnesota. Our wetland complexes included northern wet prairie, prairie mixed-cattail marsh, and prairie wet meadow/sedges, while the uplands included northern dry prairie and northern mesic prairie (MNDNR 2019). Because reed canarygrass has been planted as a forage for livestock it is predicted that grazing will be effective at reducing reed canarygrass canopy cover. Reed canarygrass is a palatable grass to cattle, with high crude protein and low content of alkaloids, creating a desirable feed (Vetsch et al. 1999). Grazing by cattle should defoliate the plant, suppress flower development, and reduce or eliminate seed production and spread. When reed canarygrass is grazed the plant does not create monoculture stands (Paine and Ribic 2002; Hillhouse et al. 2010; Kidd and Yeakley 2015).

Two grazing treatments were used at two separate sites. The first treatment used rotational grazing with a high stock density of cattle using a short grazing duration (7 d or less). Cattle were rotated through four paddocks three times (spring, early summer, and fall). This treatment design is similar to previous studies by Oates et al. (2011), Rinella and Bellows (2016), and James et al. (2017). These previous research studies showed a decrease in invasive plant abundance and increase in native species richness from grazing.

The second treatment used patch-burn grazing. This treatment used one large paddock, grazed season-long with cattle, with portions burned periodically. Biondini et al. (1999) showed bison selectively grazed new burned sites over unburned and previously burned. Spiess et al. (2020) showed cattle selectively grazed burn patches when patch-burn grazing in western North Dakota. Others (Cummings et al. 2007; Diamond et al. 2012; Scasta et al. 2016) predicted combining grazing and burning would reduce the abundance of invasive species from the dominant species to a component of the plant community. While this study is not designed to reduce reed canarygrass with burning alone, which Lavergne and Molofsky (2006) showed to be ineffective, it is believed burning paired with grazing will boost reed canarygrass suppression compared to burning alone. We predict that grazing in both approaches will reduce canopy cover of reed canarygrass.

Our second prediction is if reed canarygrass is reduced, the plant communities will change to a more diverse community. While much research has been conducted on reed canarygrass and grazing practices (patch-burn, targeted, rotational), only a few studies address controlling reed canarygrass on restored wet meadows (Green and Galatowitsch 2002; Adams and Galatowitsch 2005, 2006). No studies have looked at both rotational and patch-burn grazing as control methods for reed canarygrass on restored wet meadows.

The two sites selected for the research have as their restoration objectives:

The research objectives were:

METHODS

This research was conducted at two study sites, the Brantner and Williams sites, each located near Glyndon, Minnesota, in Clay County. The sites were located in the northeastern region of the Tallgrass Prairie (USDI US Fish and Wildlife Service 2019).

Williams Site

The Williams site (46°48′41.46″N, 96°25′48.87″W) is managed by The Nature Conservancy (TNC) and is part of the Bluestem Prairie Scientific and Natural Area. The site is 534 ha in size (Figure 2). TNC acquired the site in 2005 and began restorations in 2008, completing the first phase of the restoration in 2010. A perimeter fence was established on the site in 2011 and 2012, and a patch-burn grazing program implemented.

Upland soils were highly permeable sands, loamy sands, poor to well drained, moderately permeable to permeable, to fine and medium textured loams. The wetland soils formed in periodically inundated shallow basins and classified as mollisols (USDA Natural Resource Conservation Service 2006; MNDNR 2019). Native plant communities are tallgrasses and sedges with rushes in flooded areas (see plant list for Brantner site) (MNDNR 2019).

TNC used a local restoration contractor, Prairie Restoration Inc., who provided locally sourced seed. The differing parts of the site were sprayed with a glyphosate and clopyralid herbicides mix in July 2008, June 2009, and June 2010. Differing parts of the site were seeded in 2008, July and August 2009, and July 2010. The seed used was harvested directly from several local native prairie sites and broadcasted over the site as a bulk seed mix. The 2008 bulk seed mix was made up of 39% big bluestem and 4.1% Indiangrass (Sorghastrum nutans L.(Nash)). Other species seeded included two invasive grasses (Poa pratensis L. and Bromus spp.) at trace amounts, nine native grasses at 5.8%, and 3% various native forb species. The remaining 48.1% was inert matter. Snow seeding of the bulk seed collected during the fall of 2010 was conducted in March 2011, with big bluestem at 20%, Indiangrass 19%, leadplant (Amorpha canescens Pursh) 8.2%, and prairie dropseed (Sporobolus heterolepis) 5.4% of the mix; 5% of the seed mixture contained 12 native and one nonnative grasses, and 1% various native forb species. The remaining 42.4% was inert matter.

Figure 1.

The Brantner site showing the boundary, paddock fences, and labeled sample points. The UG label represents non-grazed exclosures. The first number of the sample point labels denotes the paddock while the number after the decimal is the point ID.

Areas seeded in 2009 and 2010 were clipped for weed control. Bird's-foot trefoil (Lotus corniculatus L.), an invasive legume, was sprayed on the previously seeded areas with aminopyralid in 2010. In September 2011, 21 ha of the site was mowed and the bird's-foot trefoil found on 40 ha was sprayed using aminopyralid. The restoration project restored 476 ha upland prairie and 40 ha of wetlands.

Study treatments

This study used two different grazing practices: (1) rotational grazing (RG) using a high stock density of cattle for short periods (7 d) and (2) patch-burn grazing (PBG). The Brantner site used the RG within a four-paddocks system. The four-paddock rotation was designed to graze each paddock twice starting late May through July when reed canarygrass was vigorously growing, and then a single rotation in the fall starting late August. The cattle were kept in other paddocks on the site when not used in the rotations. Each of the four paddocks was considered a replicate and approximately 6 ha each. A herd of 25–30 cross-bred cattle with calves was used to graze the RG system. The paddocks were considered to be moderately grazed according to regional conditions (Jeff Duchene, regional range specialist, NRCS, pers. comm.).

The RG system was designed to start in the summer of 2012 and end after the summer of 2018. Due to unforeseen watering problems, the cows were not always properly rotated in 2012, 2013, and 2014. Because there was only one watering source the gates to the pastures were left open and the cows were allowed to move freely through all four paddocks. The grazing period from 2012 to 2014 was restricted to May through July and a period from later August into September. This resulted in the cows being allowed to graze for a time similar to what would have happened if the rotational system was going. When the water system was not functioning from 2012 to 2014 grazing by the cows resulted in an uncontrolled application of the grazing treatment since cows could select to spend more time grazing in a certain paddock over the others. In 2015 until the end of the experiment in 2018 the watering system functioned as planned so that the cows were rotated through the four paddocks in a controlled fashion so the grazing treatment was identical for each paddock with stock density, grazing time, and frequency being the same.

Four locations were selected outside the four paddocks and used as non-grazed controls. These sites were fenced creating an exclosure measuring 10 × 10 m. In the areas around the grazing exclosure cattle were allowed to graze at different times. The non-grazed locations were dispersed over the site to decrease the chance they were subject to conditions that would invalidate them as controls.

Figure 2.

The Williams site showing the sample points, boundary fence, and the burn units that were burned during the study. The UG label is the non-grazed exclosures.

The Williams site had the PBG system installed with season-long grazing every year from mid-May to September, and a burn conducted every year with a burn return rate of 6 y (Figure 2). The Williams site was one 485 ha paddock. Not all sampling points had a burn occur at their location by the end of the study in 2017. An average of 578 AUMs was grazed on the site for 4–5 mo per year. This stock density resulted in the site being lightly to moderately grazed according to regional conditions over the time of the study (Jeff Duchene, regional range specialist, NRCS, pers. comm.). Because watering locations were not uniformly distributed, certain areas had less cattle use compared to other areas. The intent of PBG is to attract cattle to sites irrespective of watering locations so they get preferred grazing use for 1–2 y after a burn. Four non-grazed locations were selected on similar soils and plant communities and used as controls. These locations were areas built in 2011 and not included in the grazed paddock. The non-grazed locations were dispersed over the site to decrease the chance they were subject to conditions that would invalidate them as controls. The burning schedule resulted in 66% of points being burned once. Approximately 65% of the study site was burned once.

The Brantner site (RG) had three randomly selected sampling points within each of the four paddocks. This resulted in a total of 12 points assessed for a grazing effect along with four non-grazed control plots. The Williams site (PBG) had 12 points randomly located in the fenced paddock. If a random point did not fall within a reed canarygrass patch, the point was moved to the closest patch. A global navigation satellite system receiver was used to navigate to the sampling points each sampling period. At each sampling point, three 1 m² quadrats were arranged in a triangle 1 m from the sample point. In each quadrat the canopy cover of all species was ocularly estimated to the nearest percent (Daubenmire 1959). Bare ground and litter cover were also recorded. Canopy cover estimation was determined by the same observer to maintain consistency over time. Plants not present in the quadrats were recorded if found within a 4 m² area around the three quadrats. The additional species within the 4 m² area were given a 0.5% canopy cover in the analysis. All sampling occurred in September to capture the effects of grazing over the growing season.

RESULTS

A total of 102 species were identified at the Brantner site and 98 at the Williams site over the course of the study. Grazing reduced the canopy coverage of reed canarygrass by 50% from pretreatment condition for both the Brantner and Williams sites 7 and 5 y after grazing, respectively (Figure 3). In contrast, the non-grazed locations at both the Brantner and Williams sites did not show any reductions in reed canarygrass canopy coverage over the same time period (Figure 3). Pictures of the fence line contrasts (Figure 4) show the effect of the canopy cover reduction from grazing with a decrease in the height and stems of reed canarygrass. The PERMANOVA analysis found grazed reed canarygrass patch plant communities did differ (P < 0.05) compared to the pretreatment communities for both the Brantner and Williams sites 7 and 5 y after grazing, respectively. The non-grazed plant communities did not differ (P > 0.05) compared to the pretreatment communities for both Brantner and Williams sites 5 and 7 y after grazing, respectively.

Nonmetric multidimensional scaling ordinations of the reed canarygrass patch plant community data for the Brantner site found three explanatory axes where the stress was 10.8 (Figure 5). The Brantner NMS ordination had 80% of data represented by the first two axes with the third only representing 11%. Because only 11% of the variability was explained by the third axis, this axis was not included in the results. Knowing that the grazed plant communities differed from pretreatment communities the Brantner site directional vectors can be used to show where the plant communities were trending. The trend was that all the grazed samples moved from the positive end of axis 1 where a high cover of reed canarygrass was correlated, to the negative end of axis 1 where a high cover of wooly sedge (Carex pellita Muhl. ex Willd.) occurred (see Table 1 for canopy coverage values for species interpreted within the NMS analysis). The major effect of the RG treatment over time was to decrease the cover of reed canarygrass with a subsequent increase in wooly sedge cover. The negative end of axis 1 where the grazed sites were trending over time has two exotic grass species, Kentucky bluegrass and redtop (Agrostis gigantean Roth), with correlations of higher cover (Figure 5). Axis 2 has four native forb species correlated with the positive end of the axis. Axis 2 appears to be related to sample differences as opposed to change over time due to RG grazing treatment which axis 1 shows.

The NMS ordination of the Williams site found two explanatory axes with a stress of 18.4 accounting for 81% of the variability in the data (Figure 6). The NMS ordination of the Williams sites shows the same trends as the Brantner site, although the grazing response and axes scores are reversed. The major effect of PBG over time was to decrease the cover of reed canarygrass with a subsequent increase in wooly sedge cover. The three native species correlated with axis 2 are different but the patterns are the same. There were no consistent directional changes noted for both the Brantner and Williams non-grazed samples, which matches the analysis that showed the communities did not differ over the study.

Figure 3.

Percent canopy coverage of reed canarygrass for grazed and non-grazed patches before grazing (pretreatment) and so many years after grazing for the Brantner and the Williams sites. Treatments with the same letters are not significantly different (P > 0.05). The error bars show standard deviation.

Native species richness was compared from pretreatment levels to 7 and 5 y after grazing for the Brantner and Williams sites, respectively. Native species richness did not differ (P = 0.976) between pretreatment levels and 7 y after grazing in the reed canarygrass patches at the Brantner site (Table 2). Native species richness in the reed canarygrass grazed patches increased (P = 0.039) from the pretreatment level to 4 y after grazing at the Williams site (Table 2). The increase in richness was also found for total and exotic species richness at the Williams site. There were no changes in native, total, or exotic species richness in the non-grazed locations over the course of the study (Table 2).

DISCUSSION

The RG treatment at the Brantner site was not consistently applied in the first 3 y due to watering issues, but was consistently applied as planned in the final 4 y. This inconsistent application meant that replications were not treated the same until later in the study. We maintain that the early inconsistent grazing application did not unduly affect the RG treatment as it was applied later in the study. The early inconsistent grazing application would have introduced variability in how the sites would react to grazing, but the NMS results of the Brantner sites showed a uniform trend across all the grazed sample sites. Therefore, we contend any interpretation of the RG effects from the study should apply to rotational grazing. Concerning the lack of replication in the PBG treatment, we want to make the case that even though the PBG is an un-replicated experiment that given the experiment covered a large area and the results agree with the RG results, we can assert that the interpretation of the results can be applied to the grazing of reed canarygrass.

We found that grazing can reduce reed canarygrass cover resulting in an increase of both native and grazing-resistant species cover. One of the species that increased was a native sedge (Carex pellita), which is the type of species desired in restored wet meadow communities. Grazing did not decrease native species richness over time in the reed canarygrass patches using RG while increasing native species richness using PBG. Others (Hillhouse et al. 2010; Guretzky et al. 2018) also found that grazing of reed canarygrass did not result in lower native species richness. Therefore, the reduced cover of reed canarygrass from grazing is moving the plant community toward one with a higher cover of native wet meadow species and does not lower native species richness. While grazing is an effective way of reducing reed canarygrass canopy cover, grazing did not guarantee eradication but rather is a way to limit reed canarygrass effects on restored wet meadows.

Figure 4.

Pictures showing fence line contrast between grazed and non-grazed areas. The picture on the right is from the Brantner site with the grazed portion on the left. The picture on the left is from the Williams site with the grazed portion on the right. Both pictures show a reduction in the height and old stems of reed canarygrass.

Figure 5.

Ordination NMS graph representing the shift in the plant communities over time for the Brantner site for both grazed reed canarygrass patches and non-grazed patches. The directional successional arrows show the shift in the plant community in the reed canarygrass patches from the start of the experiment to the end of the experiment. The dotted line highlights the non-grazed patches with dotted directional successional arrows showing the shift from the start of the experiment to 7 y after grazing. Species that were highly correlated (R values >0.4 or <–0.4) with the axes are represented on the graph. Values after the axis labels show the amount of variability explained by the axis.

Table 1.

Average cover of plant species that had a Pearson's correlation coefficient of R ≥ 0.4 or R ≤ –0.4 with the significant NMS axes for either the Brantner or Williams site. Values for pretreatment and so many years after grazing started are shown.

The reduction in height and old stems that is reflected in the reduced canopy coverage of reed canarygrass from grazing is probably conducive to the increase of native sedges and other native species. By reducing the height, old stems, and canopy coverage more of the area is open to sunlight, reducing competition so other species were able to establish and grow such as native sedges (Hillhouse et al. 2010; Marchetto et al. 2021). The result of more sedge species may also be a result of saturated soils and spring flooding, with sedges more tolerant of certain wet conditions.

One of the consequences of grazing in the reed canarygrass patches is that other introduced, invasive grazing-resistant grasses, such as Kentucky bluegrass and redtop, increase in cover within the changed plant community. Both grasses were present on the site prior to restoration and were not eradicated during the restoration process. Kentucky bluegrass can be a more difficult plant to suppress and reduce, creating a new concern (DeKeyser et al. 2013). Kentucky bluegrass can be problematic for restorations and its long-term management is a problem for restorations (DeKeyser et. al. 2015; Larson et al. 2017). Because the grazed wet meadows have other species increasing like the native sedge and the wet meadow habitat not ideal for Kentucky bluegrass, it is unknown if this introduced, invasive grass will be a long-term problem as seen in other restoration studies (Larson et al. 2017). Although redtop is adapted to wet meadows, it is unclear if it is an issue for wet meadow restorations (Cao et al. 2019). This species warrants more research on its effects in wet meadow restoration, since it can increase once reed canarygrass is reduced.

Grazing has mixed results with some studies showing a reduction in exotic plants and others no change (Paine and Ribic 2002; Launchbaugh and Walker 2006; Hillhouse et al. 2010; James et al. 2017; Marchetto et al. 2021). Our study demonstrated a reduction in reed canarygrass from RG without any reduction in native species richness. James et al. (2017) along with Marchetto et al. (2021) reported grazing does not decrease and can increase native species abundance and provide other benefits to conservation. This agrees with our study and shows how RG can control certain invasive species in restorations while providing benefits to conservation.

Patch burn grazing has been advocated as a way to reduce invasive species dominance (Scasta et al. 2016) and the results of this study confirm that PBG can reduce the dominance of reed canarygrass. We speculate some of the conditions that made PBG successful were the burns attracting cattle to reed canarygrass patches making reed canarygrass accessible through removing litter and old stems (Sheaffer et al. 1990; Biondini et al. 1999). Once attracted to the patches, reed canarygrass cover was reduced by grazing and native species like sedges increased. In addition, native species richness increased. This increase in native species richness may be a function of a matured restoration with native richness naturally increasing even without grazing. Nonetheless, it appears grazing is not a hindrance to increasing native species richness even though it might not be a factor in promoting native species richness.

Figure 6.

Ordination NMS graph representing the shift in the plant communities over time for the Williams site for both grazed reed canarygrass patches and non-grazed patches. The directional successional arrows show the shift in the plant community in the reed canarygrass patches from the start of the experiment to the end of the experiment. The dotted line highlights the non-grazed patches with dotted directional successional arrows showing the shift from the start of the experiment to 7 y after grazing. Species that were highly correlated (R values >0.4 or <–0.4) with the axes are represented on the graph. Values after the axis labels show the amount of variability explained by the axis.

One suggestion for managing the reed canarygrass patches is once a reduction in reed canarygrass has been achieved, removing grazing and applying rest to the sites would be beneficial to the plant community and meet restoration goals. Based on the findings of Kidd and Yeakley (2015), we recommend that rest from grazing should be short term if at all. Rest could potentially aid in the invasions of other exotic species such as Kentucky bluegrass (DeKeyser et al. 2015). The tradeoff between reducing reed canarygrass and increasing exotic species is to be expected, as some plants will fill the niche left after reed canarygrass is reduced.

Table 2.

Average species richness for the Brantner or Williams sample points comparing pretreatment values with so many years after grazing. Different letters denote a significant difference (P < 0.05) among dates within a richness category row. Categories and rows with no letters were not significantly different (P > 0.05).

Because both grazing systems, RG and PBG, resulted in reduced cover of reed canarygrass and movement of the plant community toward restoration goals, we speculate that grazing rather than the grazing system maybe the primary factor in plant community changes. This may mean that any grazing system that managers consider would be effective, and grazing from light to moderate levels will produce the desired results. Further studies on grazing systems and levels of grazing use will need to be conducted to establish if grazing systems are important or, as we speculate, the act of grazing at effective levels is the major cause for change.

CONCLUSION