- Top of page
Although different invasive plant removal methods such as hand-weeding or herbicide application can affect native communities through soil disturbance or non-target herbicide effects for example (Aarssen & Epp 1990; Campbell et al. 1991; McLellan, Fitter & Law 1995), previous experiments explicitly evaluating native community responses have typically focused on a single removal method (D’Antonio et al. 1998; Alvarez & Cushman 2002; Hulme & Bremner 2006). Removal treatments such as hand-weeding often disturb the soil, which can disrupt root systems and mycorrhizal networks of non-target plants (McLellan, Fitter & Law 1995), but can also facilitate native (Biggerstaff & Beck 2007) and non-native (Mack & Lonsdale 2002; Ogden & Rejmanek 2005; Mau-Crimmins 2007) plant establishment. Chemical treatments may have strong effects on a particular plant functional group if the herbicide is designed to kill only monocotyledonous or dicotyledonous plants, for example (Pavlik, Nickrent & Howald 1993; Cione, Padgett & Allen 2002), and may leave toxic residues that inhibit native plant recruitment. Further, pre-emergent herbicides may inhibit seed germination of non-target species. Other control methods such as grazing, fire, mowing or shading may also have unexpected or unintended consequences. To develop effective invasive plant management techniques that promote native community recovery, studies that test multiple removal methods for specific plant invaders are needed.
In addition to the direct application of restoring invaded communities, invasive plant removal experiments provide information on the effects of the invader on native communities (Gould & Gorchov 2000; Alvarez & Cushman 2002; Diaz et al. 2003). The impacts of plant invasions can be evaluated through both introduction (e.g. Robinson, Quinn & Stanton 1995) and removal (e.g. Alvarez & Cushman 2002) experiments. Experimental introductions may provide the best measure of the effects of plant invasions, but small-scale invasion experiments in containers and common gardens yield limited insights for natural systems. However, adding noxious species to natural areas can present obvious management and ethical problems. In contrast, long-term removal experiments in natural areas can help to quantify the effects of plant invasions on native communities (Diaz et al. 2003). Because different removal methods could have different effects on native communities, studies designed to evaluate the effects of invasions through removal experiments should test multiple removal methods.
We evaluated the response of native plant communities to removal of Microstegium vimineum (Trin.) A. Camus (Japanese stiltgrass), a C-4 non-native annual grass that is rapidly invading eastern USA forests (Winter, Schmitt & Edwards 1982; Horton & Neufeld 1998). Microstegium was introduced to the USA from south-east Asia in the early 1900s (Fairbrothers & Gray 1972) and is currently invasive in more than 20 states (USDA & NRCS 2005). Invasions by Microstegium and other species threaten to reduce the diversity of woodland herbaceous communities and inhibit forest regeneration and succession (Winter, Schmitt & Edwards 1982; Barden 1987; Oswalt, Oswalt & Clatterbuck 2007). Microstegium is a prolific seed producer (Tu 2000) and is dispersed through water, animals, and anthropogenic activities. It most commonly invades forest openings, riparian areas, and along roads, streams and trails (Redman 1995; Tu 2000). Microstegium is highly shade tolerant and can reproduce under deeply shaded conditions (Winter, Schmitt & Edwards 1982; Horton & Neufeld 1998). Multiple methods have been used to control Microstegium, including mowing, fire, selective (grass-specific) and non-selective (glyphosate) herbicides, and hand-pulling (Tu 2000; Czarapata 2005), but studies testing the effectiveness of different treatments for removing invasions and the specific responses of native communities to removal treatments are needed (Judge et al. 2005a,b; Flory 2008).
Previously, we reported on the effectiveness of Microstegium removal methods over multiple growing seasons (Flory 2008). We found that two herbicide-based removal methods each reduced Microstegium biomass by 99%, while a hand-weeding treatment was slightly less effective (87% reduction, Flory 2008). Here, we use the same system to evaluate the specific responses of native plant communities to multiple invasive plant removal methods. After two seasons of treatment, we quantified the species richness, biomass, and cover of specific native plant functional groups, the number of colonizing tree seedlings, and overall native plant community richness, evenness, and diversity. We evaluated specific native plant functional groups separately because Microstegium and other invaders may be having disproportionate effects on different native plant groups. In addition, the method used to remove invasions may differentially affect functional group responses to removal. Specifically, we asked the following questions: (i) Do the responses of native plant communities depend on the method used to remove invasive Microstegium? (ii) Do the responses of native plant functional groups differ among removal methods? (iii) Does removal result in greater native plant diversity and biomass? Answers to these questions have applications for management of this specific invader, but are also broadly relevant to native plant community responses to removal methods of any plant invader.
- Top of page
The responses of the four functional groups (graminoids, forbs, woody species, and ferns) of native species to Microstegium removal differed significantly based on the removal method (Table 2, Fig. 2). For graminoids, there was greater species richness when Microstegium was removed with hand-weeding (HW) but lower richness with POST + PRE compared to REF plots (Fig. 2a). In contrast, forb richness was greater with both the HW and POST treatments compared to REF plots (Fig. 2c). Although there were trends for greater woody species richness with HW and POST (Fig. 2e), there were no statistically significant differences among treatments. There were also non-significant trends for greater fern species richness with the POST treatment (Fig. 2g). There was a significant site × treatment interaction for forb biomass (Table 2) due to greater positive responses of forb species to removal at some sites than others.
Table 2. Results of anova tests for the effects of treatment, site, and their interaction on graminoid, forb, woody species, and fern richness and biomass in fall 2006
| ||Graminoid||Forb||Woody species||Fern|
|Treatment|| 3||14·52||< 0·0001||3·39||0·02||21·27||< 0·0001|| 3·75||0·01||2·81||0·04||1·75||0·16||3·19||0·02||2·30||0·08|
|Site|| 7||27·68||< 0·0001||8·51||< 0·0001||9·27||< 0·0001||13·68||< 0·0001||3·02||0·0045||6·32||< 0·0001||8·30||< 0·0001||4·50||< 0·0001|
|Trt × Site||21|| 1·17||0·28||1·49||0·08||1·6||0·05|| 3·23||< 0·0001||1·36||0·14||1·03||0·43||0·93||0·56||1·39||0·12|
Figure 2. Average (± SE) species richness and biomass of four functional groups per 0·25 m2 quadrat, graminoids (a,b), forbs (c,d), woody species (e,f), and ferns (g,h) in REF, HW, POST, and POST + PRE plots for fall 2006. Different letters indicate significant differences at P < 0·05.
Download figure to PowerPoint
Graminoid biomass was lower when Microstegium was removed with POST compared to REF plots (Fig. 2b) but forb biomass was greater across all removal treatments (Fig. 2d). Woody species biomass was also greater across all removal treatments but the trends were not significant (Fig. 2f). Likewise, there was a non-significant trend for greater fern biomass with POST (Fig. 2h).
Overall native community diversity was greater with the HW and POST, but not POST + PRE, compared to REF plots (Table 3, Fig. 3). Native species richness was greater with HW and POST (Fig. 3a) but native species evenness was unaffected by the treatments (Fig. 3b), indicating that changes in diversity were driven by changes in species richness. There were greater positive responses in overall native community richness at some sites than others resulting in a significant site × treatment interaction (Table 3). There was 24% and 21% greater native community diversity when Microstegium was removed using HW and POST respectively compared to REF plots.
Table 3. Results of anova tests for the effects of treatment, site, and their interaction on native community richness, evenness, and diversity in fall 2006
|Treatment|| 3||25·91||< 0·0001||0·55||0·65||8·94||< 0·0001|
|Site|| 7||20·18||< 0·0001||2·26||0·03||7·50||< 0·0001|
|Treatment × site||21|| 1·72||0·03||1·33||0·16||1·33||0·16|
Figure 3. Average (± SE) native species richness (a), evenness (b), and diversity (c) 0·25 m−2 quadrat for fall 2006 in REF, HW, POST, and POST + PRE plots. Different letters indicate significant differences at P < 0·05.
Download figure to PowerPoint
After 2 years of treatment, spring graminoid percentage cover was greater with the HW treatment compared to REF plots but not with POST or POST + PRE (Table 4, Fig. 4a). However, forb cover was greater across all removal treatments compared to REF plots. The greatest difference in forb cover (43%) was observed in POST-treated plots (Fig. 4b). However, there were greater positive responses in forb cover at some sites than others, resulting in a site × treatment interaction (Table 4). Tree seedling density was 123% greater with POST than in REF plots, but tree seedling colonization was no greater with HW or POST + PRE (Fig. 4c).
Table 4. Results of anova tests for the effects of treatment, site, and their interaction on graminoid and forb cover and the number of tree seedlings per plot in spring 2007
|Source||d.f.||Graminoid cover||Forb cover||Tree seedlings|
|Treatment|| 3|| 4·55||0·0039||21·68||< 0·0001|| 3·5||0·02|
|Site|| 7||27·94||< 0·0001||15·39||< 0·0001||22·27||< 0·0001|
|Treatment × site||21|| 1·5||0·08|| 1·67||0·04|| 1·4||0·12|
Figure 4. Average (± SE) graminoid (a) and forb (b) percentage cover and the number of tree seedlings (c) 0·25 m−2 quadrat for spring 2007 in REF, HW, POST, and POST + PRE plots. Different letters indicate significant differences at P < 0·05.
Download figure to PowerPoint