Biological Conservation 159 (2013) 292–302

Contents lists available at SciVerse ScienceDirect

Biological Conservation

journal homepage: www.elsevier .com/locate /b iocon

Ecological tradeoffs in non-native plant management

0006-3207/$ - see front matter � 2012 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.biocon.2012.10.017

⇑ Corresponding author. Present address: Department of Natural Resources andEnvironmental Science, Mail Stop 370, Fleishmann Agriculture, Room 121, Univer-sity of Nevada, Reno, NV 89557-0013, USA. Tel.: +1 4065997576; fax: +14069943933.

E-mail addresses: tskurski@gmail.com (T.C. Skurski), bmax@montana.edu(B.D. Maxwell), lrew@montana.edu (L.J. Rew).

Tanya C. Skurski ⇑, Bruce D. Maxwell, Lisa J. RewDepartment of Land Resources and Environmental Sciences, Montana State University, PO Box 173120, Bozeman, MT 59717-3120, USA

a r t i c l e i n f o a b s t r a c t

Article history:Received 10 July 2012Received in revised form 13 October 2012Accepted 22 October 2012Available online 23 January 2013

Keywords:Invasive plantsPlant community impactsRemoval experimentManagement effectsNative plant conservationNatural areas

Controlling non-native plants in natural areas should, ideally, not only reduce target species’ abundance,but also benefit broader management objectives such as conserving native species, improving wildlifehabitat, and maintaining ecosystem function. In this context, the effectiveness and non-target impactsof control strategies, such as broadleaf herbicides, must be weighed against the impacts of non-nativeplants themselves. We undertook this relative assessment for Centaurea stoebe, one of the most wide-spread and heavily managed non-native plant species in the Intermountain West, USA. While effective-ness and plant community impacts of herbicide treatment for C. stoebe have been assessed, field-basedexperiments quantifying community-level impacts of C. stoebe are rare. In a three-year experiment insagebrush–grassland communities of southwest Montana, USA, we found that the broadleaf herbicide,picloram, was highly effective at reducing C. stoebe, but also caused a significant loss of native forb coverand a significant increase in non-native grass cover, primarily Bromus tectorum. There was a significantincrease in native forb cover in response to manual removal of C. stoebe, which would seem to indicateC. stoebe had been suppressing native forbs. However, there was an equivalent increase in native forbcover with no treatment. In some communities, C. stoebe appears to have a negligible effect on native forband grass cover and richness. Depending on management objectives, the loss of native forb cover andpotential secondary invasion may outweigh the benefits of reduced target non-native plant abundance;thus, highlighting an ecological tradeoff of non-native plant management in natural areas.

� 2012 Elsevier Ltd. All rights reserved.

1. Introduction

Non-native plants pose a significant threat to biodiversity andecosystem structure and function (Vitousek et al., 1997; Macket al., 2000; Simberloff, 2011; Vilà et al., 2011), motivating aggres-sive and sustained control programs throughout the world. Theinvasion of non-native plants into natural areas and protectedlands is particularly troublesome given the importance of theseareas for conserving biological diversity and providing ecosystemservices. Consequently, controlling non-native plants is eitherstrongly advised or mandated for most natural areas and publiclands. However, controlling non-native plants is only one of manyland management objectives in natural areas, and whether com-mon non-native plant control strategies are congruent with theseother objectives is unclear.

The Intermountain West of North America (‘the west’ hereaf-ter)—a region of both high conservation and human use value—exemplifies many of the challenges of managing natural areas for

multiple natural resource objectives. Numerous non-native plantsspecies have established and spread throughout the region, andone of the species of greatest concern is Centaurea stoebe L. (spot-ted knapweed; formerly known as Centaurea maculosa Lam.). C.stoebe is a short-lived perennial native to Europe and northernand central Asia that was accidentally introduced to North Americain contaminated alfalfa (Medicago sativa L.) seed in the 1890s(Roché and Talbott, 1986). C. stoebe is now one of the most wide-spread non-native plants and is a designated ‘‘noxious weed’’ inthe 11 contiguous western states (NRCS, 2012), meaning stateand federal land management agencies are mandated to controlC. stoebe and minimize the economic, ecological, and human healthimpacts that it causes (Executive Order 13112 of 1999).

The major impacts associated with C. stoebe in natural areas ofthe west (i.e. grasslands, shrublands, sagebrush steppe) includedisplaced native plant species, reduced forage quality and quantity,and negatively altered wildlife habitat (Hirsch and Leitch, 1996;DiTomaso, 2000; Duncan, 2005). A number of traits have beenindentified that make C. stoebe highly competitive against nativeNorth American plant species, including extensive mycorrhizal col-onization (Marler et al., 1999; Callaway et al., 2004a; Harner et al.,2010), production of allelopathic compounds (Ridenour and Call-away, 2001; Bais et al., 2003; Thorpe et al., 2009), greater nutri-ent-use efficiency relative to native species (Blicker et al., 2002;

T.C. Skurski et al. / Biological Conservation 159 (2013) 292–302 293

Zabinski et al., 2002; Thorpe et al., 2006), greater compensatorygrowth after herbivory relative to native species (Walling andZabinski, 2006), and cultivation of soil biota in the invaded rangethat increase C. steobe’s growth (Callaway et al., 2004b). Whenseeded into experimental monocultures, C. stoebe significantly re-duced the biomass of several native western USA forb and grassspecies (Maron and Marler, 2008b). It is generally assumed thatC. stoebe displaces native plant species in natural plant communi-ties because the presence and abundance of C. stoebe is often neg-atively correlated with native plant species cover, richness anddiversity (Tyser and Key, 1988; Kedzie-Webb et al., 2001; Ortegaand Pearson, 2005; May and Baldwin, 2011). The only experimen-tal (manipulative) study of C. stoebe impacts under natural condi-tions that we are aware of is Lesica and Shelly (1996). Theyfound that removing C. stoebe in two western Montana bunchgrasscommunities led to significantly higher seedling recruitment of arare native Brassicaceae, Mt. Sapphire rockcress (Arabis fecundaRollins); thus, demonstrating a species-specific competitive effectof C. stoebe. Collectively, these studies and others have greatly ad-vanced our understanding of the biology, ecology, and potentialimpacts of C. stoebe. Nevertheless, quantitative field-based evi-dence linking C. stoebe to plant community-level impacts remainsscarce.

C. stoebe is the most heavily managed non-native plant on pub-lic lands in the west (based on quantity of herbicide used, numberof hectares treated, and hours of labor (USFS et al., 2010)). Manage-ment objectives in natural areas typically include maintaining na-tive plant diversity and wildlife habitat, increasing forageproduction, sustaining ecosystem services, and conserving speciesof concern. It is generally assumed that reducing non-native plantsin natural areas will benefit these objectives (Pearson and Ortega,2009). However, the legal mandate to control noxious weedsmay compel action before there is sufficient knowledge of howother land management objectives will be affected. Assessing theimpacts of non-native plant species, as well as the impacts of thestrategies used to control them, are crucial steps in managing nat-ural areas for multiple conservation objectives (Buckley, 2008;Pearson and Ortega, 2009; Downey et al., 2010).

While there have been numerous studies of the effectiveness(i.e. reduction of C. stoebe abundance) and plant community im-pacts of herbicide treatments for C. stoebe (e.g. Rice and Toney,1998; Sheley et al., 2000; Crone et al., 2009; Ortega and Pearson,2010), there have not been similar experimental studies of thecommunity-level impacts of C. stoebe. Thus, it is unclear how thebenefits of C. stoebe management compare with the impacts of C.stoebe itself.

We conducted replicated removal experiments to quantify therelative impacts of C. stoebe and herbicide treatment of C. stoebeon sagebrush–grassland plant communities in southwest Montana.Removal experiments, along with experimental additions, providethe strongest evidence for assessing non-native plant impacts andavoid the confounding effects of environmental variation that arisein multi-site comparison studies (Adair and Groves, 1998). Theobjective of removal experiments is to approximate how a plantcommunity would be structured in the absence of the non-nativeplant species. Non-native plants are removed with as little distur-bance as possible and, after a recovery period, impacts are quanti-fied as the change in plant community metrics from beforeremoval.

The objectives of this study were to: (1) quantify impacts of C.stoebe on sagebrush–grassland plant communities in southwesternMontana; (2) compare the impacts of C. stoebe with those of herbi-cide control of C. stoebe, small-scale disturbance, and no treatment;and (3) determine whether invaded plant communities differedfrom nearby uninvaded plant communities and, if so, whetherplant communities treated for C. stoebe became more similar to

nearby uninvaded plant communities 3 years after treatment.The response variables measured for all three objectives were thepercent cover and species richness and composition of native andnon-native forbs and grasses, as well as the aboveground biomassof two native forage grasses. We focused on growth form (forbsand grasses) and origin groups (native and non-native) becausethese are common land management metrics and may be betterindicators of a range of ecosystem processes, such as nutrient cy-cling, decomposition, and water retention, than individual speciesidentity alone (Diaz and Cabido, 2001; Cadotte et al., 2011).

2. Methods

2.1. Study area

The study was conducted in the sagebrush–grassland habitattype of southwest Montana. These communities occupy the transi-tion zone between valley bottoms and Douglas fir-dominated for-est (Pseudotsuga menziesii (Mirb.) Franco). Perennial grasses featureprominently, ranging between 40% and 60% canopy cover (Mueg-gler and Stewart, 1980). The dominant species are bluebunchwheatgrass (Pseudoroegneria spicata (Pursh) Á. Löve), westernwheatgrass (Pascopyrum smithii (Rydb.) Á. Löve), and Idaho fescue(Festuca idahoensis Elmer), and frequent species include green nee-dlegrass (Nassella viridula (Trin.) Barkworth), slender wheatgrass(Elymus trachycaulus (Link) Gould ex Shinners), prairie junegrass(Koeleria macrantha (Ledeb.) Schult.), squirreltail (Elymus elymoides(Raf.) Swezey), and Sandberg bluegrass (Poa secunda J. Presl). Forbcanopy cover ranges from 20% to 30% and typically includes:arrowleaf balsamroot (Balsamorhiza sagittata (Pursh) Nutt.), buck-wheat (Eriogonum species), rosy pussytoes (Antennaria roseaGreene), hairy false goldenaster (Heterotheca villosa (Pursh) Shin-ners var. villosa), prairie sagewort (Artemisia frigida Willd.), silverylupine (Lupinus argenteus Pursh), Indian paintbrush (Castilleja spe-cies), fleabane (Erigeron species), phlox (Phlox species), scarletglobemallow (Sphaeralcea coccinea (Nutt.) Rydb.), and milkvetch(Astragalus L. species). Shrub canopy cover averages 5–10%, com-prised mainly of big sagebrush (Artemisia tridentata Nutt.) andmountain big sagebrush (Artemisia tridentata Nutt. ssp. vaseyana(Rydb.) Beetle). Average litter cover in these communities is 40–60% and bare ground 5–10%. Average temperatures in the studyarea are 19 �C in the summer and �4 �C in the winter. The averageannual precipitation is 480 mm, with most falling in April, May,and June.

High forage production and diverse plant communities makesagebrush–grasslands in this region important habitat for numer-ous large mammals (e.g. elk (Cervus canadensis), mule deer (Odocoi-leus hemionus), pronghorn (Antilocapra americana) and black bear(Ursus americanus)); small mammals (e.g. deer mouse (Peromyscusmaniculatus), Uinta ground squirrel (Urocitellus armatus), bushy-tailed woodrat (Neotoma cinerea)); songbirds (e.g. Brewer’ssparrow (Spizella breweri), Vesper sparrow (Pooecetes gramineus),western meadowlark (Sturnella neglecta)); upland game birds(e.g. greater sage-grouse (Centrocercus urophasianus), sharp-tailedgrouse (Tympanuchus phasianellus)); and raptors (e.g. golden eagle(Aquila chrysaetos), ferruginous hawk (Buteo regalis)). Many taxathat are at-risk due to declining population trends and habitat loss(‘‘species of concern’’) inhabit sagebrush–grasslands in this regionfor at least part of the year (MNHP, 2011).

There is a long history of human activity in the region, includingmining, logging, farming, and ranching. In addition, due to moder-ate topography, open landscapes and proximity to wildlands, thelast quarter century has seen rapid growth in recreational useand exurban development in lowland habitats such as sage-brush–grasslands (Gude et al., 2006). All of these activities and

294 T.C. Skurski et al. / Biological Conservation 159 (2013) 292–302

their associated transportation networks have likely contributed tothe introduction and spread of non-native plant species (Olliffet al., 2001; Parks et al., 2005; Mortensen et al., 2009).

2.2. Experimental design

Nine C. stoebe populations were randomly selected across thestudy area and were located between 45�310N, 111�150W;45�170N, 110�490W; and 45�100N, 110�510W (Fig. 1 and KML fileonline) at elevations between 1515 m and 1790 m. The only selec-tion criteria for the study populations were: that they were in nat-ural areas (i.e. non-crop) and that they were located in areas thatretained some native vegetation. This latter criterion was easilymet since most C. stoebe populations in natural areas of this regionexist within a matrix of native plant communities (although othernon-native plant species are common as well). The study popula-tions were on south–southeast aspects, 2–25% (1–14�) slope andwere 60–600 m away from roads. Populations ranged from 150to 650 m2 in spatial extent and were between 0.1 and 50 km apart.Initial C. stoebe cover was 5–25%. All of these characteristics aretypical of C. stoebe populations found in sagebrush–grasslands inthis region.

Within each of the nine C. stoebe population, two 0.25 m2 plotswere established for each of four treatments: (1) no treatment, (2)manual removal of C. stoebe, (3) ground disturbance (with smallhand tools) similar to that created in the manual removal treat-ment but with no plant material removed, and (4) herbicide appli-cation (four treatments � nine populations � two replicate plotsper population = 72 plots). A concern regarding removal experi-

Fig. 1. Study area located in the sagebrush–grassland habitat of southwest Montana, Udistributed across the study area at elevations between 1515 m and 1790 m. KML file p

ments is that observed changes may be in response to the soil dis-turbance created in the removal process rather than to the absenceof the non-native plant. We addressed this concern by including aground disturbance treatment that mimicked the disturbance cre-ated by C. stoebe removal, but with no plant material removed.Thus, if C. stoebe was having an impact, the effect of manual re-moval of C. stoebe on response variables would be greater thanthe effects of ground disturbance with no removal and no treat-ment. The area disturbed varied depending on the size and amountof C. stoebe removed in the adjacent removal plot; thus the areadisturbed ranged from 10% to 50% of the plot surface area. We con-sider ‘no treatment’ a treatment in the sense that it is one potentialnon-native plant management option. Within populations, plotswere established to maximize the similarity among treatments interms of canopy cover, percent bare ground, ratio of forbs tograsses, percent cover of C. stoebe, and species richness. Plots wereseparated by at least 1 m within populations. To examine differ-ences between invaded and uninvaded areas, and whether thischanged over time, we established two additional plots approxi-mately 2 m outside of each C. stoebe population. Comparisons weremade with uninvaded areas close to C. stoebe populations to min-imize differences in habitat conditions and disturbancecharacteristics.

Treatments were applied in late June 2007 after pre-treatmentdata had been collected, which coincides with the middle of thespray season for most local land management agencies. Herbicideplots were broadcast sprayed with the broadleaf herbicide, piclo-ram (Tordon™ 22 K), which is the most common herbicideused for C. stoebe, at the recommended rate (1.17 l ha�1 =

SA. Experiments were conducted at nine Centaurea stoebe L. populations randomlyrovided online.

Table 1Results from one-way analysis of variance for variation in the proportional change inresponse variables from pre-treatment to 3 years post-treatment levels amongsttreatments. Treatments: (1) no treatment, (2) manual C. stoebe removal, (3)disturbance, and (4) herbicide treatment. Only significant response variables shown(p < 0.05); complete results in Table S3 (supplementary material).

Response variable Mean square F3,60 p

C. stoebe cover 22.41 7.98 <0.001Total non-native cover 21.80 4.04 0.011Native forb cover 6.38 3.14 0.032Native grass cover 0.63 3.01 0.043B. tectorum cover 29.74 4.52 0.006

T.C. Skurski et al. / Biological Conservation 159 (2013) 292–302 295

0.28 kg ae ha�1, Dow AgroSciences, 2001). Herbicide was appliedusing a CO2 pressure regulated backpack sprayer, which is a typicalapplication method in natural areas of this region. In manual re-moval plots, C. stoebe plants were carefully hand-pulled when soilswere wet to minimize disturbance and facilitate removal of entireplants. In disturbance plots, hand tools were used to disturb thesame area of soil as in the paired manual removal plots, but withno plant material removed.

2.3. Vegetation sampling

We quantified the percent canopy cover (estimated to the near-est 1%) for each plant species, as well as the percent cover of bareground, lichens and bryophytes (all species together as one covertype), and litter in each plot. Because plant canopies often overlap,the total percent cover per plot could exceed 100%. Measurementswere taken annually in June and July 2007 through 2010. Planttaxonomy was based on Cronquist et al. (1977), Dorn (1984), andLavin and Seibert (2009). We also counted the number of C. stoebestems per plot.

2.4. Biomass estimation

We examined potential effects of C. stoebe and herbicide treat-ment of C. stoebe on forage production in natural areas by quanti-fying the change in aboveground biomass of P. spicata (bluebunchwheatgrass) and N. viridula (green needlegrass). These nativeperennial bunchgrasses were chosen because they are importantforage species for domestic livestock and wildlife and because theywere present in most of the C. stoebe study populations. Wenon-destructively estimated changes in aboveground biomass ofP. spicata and N. viridula within plots by developing allometric rela-tionships between basal diameter, height and aboveground bio-mass for each species (Corbin and D’Antonio, 2004). During peakproductivity (mid-July) in 2008 and 2009, we selected 227 P. spi-cata and 71 N. viridula individuals in off-plot areas representing arange of sizes. Basal diameter and height were recorded for eachindividual, after which plants were clipped at the soil surface,bagged, and dried at 60 �C for 6 days. Linear regression was usedto calculate the relationship between basal diameter, height, andindividual plant dry weight (to the nearest 0.001 g). The r2 forthe regression models were 0.842 and 0.903 for P. spicata and N.viridula, respectively (Table S1 – supplementary material).

2.5. Data analysis

2.5.1. Vegetation metricsThe total percent cover and plant species richness were calcu-

lated per plot, as well as cover and richness values within growthform and origin groups: forbs and grasses, native and non-native. C.stoebe percent cover and density were analyzed as separate re-sponse variables and C. stoebe was excluded from all other re-sponse variable calculations. Cover values were square roottransformed to meet parametric assumptions of normality andhomogeneity of variance. There was a total of 22 response vari-ables (Table 1).

2.5.2. Treatment effectsWe tested for treatment effects on the absolute value of re-

sponse variables over time using a mixed effects model. ‘Plot’was the independent sampling unit. ‘Treatment’ was a fixed effectand ‘plot’ was a repeated measure nested within the continuousrandom effect, ‘year’. This model structure accounts for the non-independence of repeated measures on individual plots throughtime. ‘Population’ was included as a random blocking effect to ac-count for potential unmeasured within-population similarities.

2.5.3. Proportional changeIn addition to examining treatment effects on the absolute va-

lue of response variables over time, we also assessed the propor-tional change for each response variable by treatment.Proportional change takes into account the initial value of responsevariables and was calculated as: (3 years post-treatment value –pre-treatment value)/pre-treatment value for each response vari-able. One-way analysis of variance (ANOVA) was used to examinedifferences in the proportional change in response variables amongtreatments. ‘Population’ was again included as a random blockingeffect. The two approaches, mixed effects over time and propor-tional change, allowed a comparison of transitory versus longer-term treatment effects.

2.5.4. Effect size estimationThe ANOVA results were a useful indication of which response

variables seemed most affected by treatments. However, ulti-mately we were interested in effect sizes, which provide more bio-logically meaningful information than the binary significance/non-significance outcome of ANOVA (Rinella and James, 2010). There-fore, we estimated the mean effect size of treatments for responsevariables identified as significant in the initial mixed model andANOVA analysis using parametric bootstrapping. The 2007–2010proportional changes for each response variable were randomlyresampled within treatment group, creating a new data set fromwhich the mean was calculated. This procedure was repeated1000 times to generate an estimated distribution of the mean ef-fect size of each treatment per response variable.

2.5.5. Comparison with uninvaded areasWe tested for differences between treatment plots inside C. sto-

ebe invaded plant communities and uninvaded plant communitiesboth before, and 3 years after, treatment for all response variablesusing one-way ANOVA. We also quantified the proportion of thetotal per-plot cover represented by each of the growth form andorigin groups (i.e. the proportion of total per-plot vegetation coverthat was native forbs, native grasses, etc.) for inside and outsideplots. We examined whether these proportional representationsdiffered between inside and outside plots, and if so, whether thesedifferences changed over time. Post hoc contrasts were conductedwith Tukey’s multiple comparison test (p < 0.05).

2.5.6. Species compositionWe tested for differences in species composition among treat-

ments and between invaded and uninvaded areas both before,and 3 years after, treatment, using permutation-based multivariateANOVA based on Bray–Curtis dissimilarity matrices. This is a alter-native to ordination methods for examining whether, and to whatdegree, variation in a multivariate response (i.e. species composi-tion) can be explained by a continuous or, as in our experiment,categorical predictor variable (i.e. treatments) (Oksanen, 2011).Test significance was based on 1000 permutations of the abun-dance-based dissimilarity matrices per-treatment (or uninvadedstatus) per-year.

296 T.C. Skurski et al. / Biological Conservation 159 (2013) 292–302

Data were analyzed and plotted using R, version 2.14.1 (RDevelopment Core Team, 2011) including packages ggplot2 (Wick-ham, 2009), nlme (Pinheiro et al., 2012) and vegan (Oksanen et al.,2011).

3. Results

3.1. Pre-treatment conditions

Before treatment (2007), native and non-native species richnessand cover (forbs and grasses), C. stoebe cover and density, P. spicataand N. viridula biomass, and percent bare ground and litter were

Fig. 2. Mean (±1 SE) per plot values for plant response variables by treatment over time (panels. C. stoebe was excluded from all metrics except C. stoebe cover and density.

not significantly different among the four treatment groups (notreatment, disturbance, manual removal, and herbicide), as indi-cated by non-significant one-way ANOVAs (a = 0.05) (Table S2).Pre-treatment species composition was also not significantly dif-ferent among treatment plots, as indicated by non-significant mul-tivariate ANOVA based on 1000 permutation of Bray-Curtisdissimilarity matrices (Table S2).

3.2. Transient versus long-term impacts

Percent cover of native and non-native forbs and grasses, litter,and bare ground; forb richness; and P. spicata biomass all changed

n = 18 plots per treatment). 2007 values are pre-treatment. Note: y-axes vary among

T.C. Skurski et al. / Biological Conservation 159 (2013) 292–302 297

significantly over time in response to treatments (Fig. 2 andTable S3). However, many of these changes were transient andonly significant in the mixed model analysis of treatment effectby year. When assessed as a proportional change from pre-treat-ment levels to 3 years post-treatment, effects were significantlydifferent among treatments (p < 0.05) only for C. stoebe cover, totalnon-native cover (excluding C. stoebe), native forb cover, nativegrass cover, and B. tectorum cover (Table 1 and detailed below).

3.3. C. stoebe cover

Herbicide treatment was highly effective at reducing C. stoebecover (Fig. 2). Before treatment C. stoebe cover across all plotswas 8.7% ± 0.6% (mean ± 1 SE). Three years after treatment, meanC. stoebe cover in herbicide plots was 0.8% ± 0.5%, whereas un-treated and disturbed plots averaged 12.5% ± 2.2% and13.5% ± 3.2% C. stoebe cover, respectively. The proportional changein C. stoebe cover from pre-treatment to 3 years post treatment wassignificantly different among treatments (F3,60 = 7.98, p < 0.001).Overall, C. stoebe cover increased by an estimated 118% ± 2.0%and 145% ± 1.7% with disturbance and no treatment, respectively,and decreased by an estimated 79% ± 4.7% with herbicide treat-ment. The proportional change in C. stoebe cover was not signifi-

Fig. 2. (con

cantly different between manual removal and herbicidetreatments (Fig. 3).

3.4. Native forb, native grass and total non-native cover

The change in total plant cover (not including C. stoebe) was notsignificantly different among treatments; however, the propor-tional representation of forbs and grasses and native and non-na-tive species shifted, with the largest effects resulting fromherbicide treatments (Fig. 3).

The proportional change in native forb cover was significantlydifferent among treatments (F3,60 = 3.143, p < 0.0317). Three yearsafter treatment, native forb cover had increased by 107% ± 1.6%and 98% ± 1.3% in response to manual removal and no treatment,respectively. A significantly smaller increase in native forb coverresulted from disturbance (36% ± 0.6%), whereas native forb coverdecreased by an estimated 22% ± 0.4% in response to herbicidetreatment (Fig. 3 and Table S4).

Native grass cover increased over time with no treatment, man-ual removal, and herbicide treatment; however, the increase wassignificantly greater in response to herbicide (F3,60 = 3.01,p < 0.0427) (Fig. 2). Three years after treatment, the estimated ef-fects on native grass cover were increases of 41% ± 0.3% in response

tinued)

C. stoebe cover (p < 0.001) Native forb cover (p = 0.0317)

Total non-native cover (p = 0.011) B. tectorum cover (p = 0.0064)

Native grass cover (p < 0.001)

aa

b

b

aa

a

b

a a a

b

a

aa

b a

a

a

a

b

Fig. 3. Distributions of treatment effect sizes estimated from 1000 bootstrap samples of the proportional change in the response variable from pre-treatment level to 3 yearspost-treatment = (2010 value – 2007 value)/2007 value. Values on the y-axis correspond to the proportional change in the response variable; however, effect sizes arepresented as a percent change in the text. Therefore, a mean effect size plotted as 1.0 indicates that the response variable increased, on average, by 100% from beforetreatment to 3 years after treatment. Negative values indicate an overall reduction in the response variable over time. Note y-axes vary among panels. p-values are from one-way analysis of variance examining differences in proportional change among treatments (Table 1 and Table S3); only significant response values are shown here (p < 0.05).Mean indicated by (⁄). Different letters, a–c, denote significant differences among means (p < 0.05; Tukey’s HSD test). Note C. stoebe was excluded from the calculation of‘Total non-native cover’. Numerical values of mean effect sizes ± SE provided in Table S4. ‘Control’ = no treatment, ‘Pull’ = manual removal, ‘Disturb’ = disturbance,‘Spray’ = herbicide.

Table 2One-way analysis of variance results for differences between treatment plots and uninvaded plots in vegetation response variables. Only significant response variables are shown(p < 0.05).

Pre-treatment (2007) 3 Years post-treatment (2010)

F4,77 p F4,77 p

Forb proportion 5.15 0.002 3.45 0.016Native forb proportion 8.36 <0.001 5.44 0.001Grass proportion 5.15 0.002 3.53 0.015Native grass proportion 3.62 0.013 1.17 0.339

298 T.C. Skurski et al. / Biological Conservation 159 (2013) 292–302

to herbicide, 18.5% ± 0.4% with no treatment, and 14% ± 0.5% in re-sponse to manual removal. Disturbance tended to have a slightnegative effect on native grass cover (estimated mean decreaseof 4.4% ± 0.3%) (Fig. 3).

Total non-native plant cover (other than C. stoebe) increasedacross all treatments; however, the increase was significantlygreater in response to herbicide relative to the other treatments(F3,60 = 4.041, p < 0.011) (Table 1). Three years after treatment, totalnon-native cover had increased by an estimated 248% ± 2.9% withherbicide, 62% ± 1.4% with no treatment, 19.4% ± 0.6% with distur-bance, and 16.7% ± 0.4% with manual removal (Fig. 3). Thus, therewas a substantial natural increase in non-native cover over time,but this increase was much greater with herbicide treatment.There were marked differences among the treatments in the typeof non-native plants that increased. With herbicide and no treat-ment nearly all of the increase came from non-native grasses andthe majority of this was B. tectorum (Fig. 3 and Table S4). Herbicidetreatment led to an estimated mean increase in B. tectorum cover of200% ± 2.9% beyond the 60% ± 2.0% increase that occurred in un-treated plots. Surprisingly, B. tectorum decreased with disturbanceand manual removal of C. stoebe, 15% ± 0.5% and 1.9% ± 1.4%,

respectively. These treatments elicited a greater response fromnon-native forbs, primarily the annual/biennial Brassicaceae spe-cies, Alyssum desertorum Stapf and Berteroa incana (L.) DC.

It should be noted that there were no response variables forwhich the effect of manual removal of C. stoebe was greater thanthe effect of no treatment, other than C. stoebe cover.

3.5. Biomass, species richness, and species composition

Neither the change in biomass of P. spicata and N. viridula overtime in response to treatments, nor the proportional change in bio-mass from pre-treatment to 3 years post-treatment, was signifi-cantly different among treatments (P. spicata: F3,60 = 0.518,p = 0.671; N. viridula: F3,60 = 0.632, p = 0.597). Native species rich-ness initially declined in response to herbicide treatment, butwas not significantly different among treatments by the end ofthe experiment (Fig. 2). Species composition was also not signifi-cantly different among treatments before, or 3 years after, treat-ment (2007: F3,60 = 0.436, p = 0.986; 2010: F3,60 = 0.851, p = 0.642,based on 1000 permutations of dissimilarity matrices for eachyear). Four native forb species (one annual and three perennials)

Table 3Percent of total live vegetation cover (mean ± 1 SE) represented by response variable groups pre-treatment (2007) and 3 years post-treatment (2010) for treated and uninvadedplots. Different letters (a or b) denote statistical differences among treatments within response variable and year (p < 0.05; Tukey’s HSD test; uninvaded considered a treatmenthere). Results from ANOVA testing for differences between years within treatment shown as ‘p-value for 2007–2010 change’; � indicates statistical difference (p < 0.05). CON = notreatment; DIST = disturbed; REM = manual removal; HERB = herbicide; OUT = uninvaded.

The mean (± 1 SE) total vegetation cover per plot was 24.5% ± 2.7% in 2007 and 32.5% ± 4.07% in 2010.

Forb p-Value for 2007–2010 change (df = 1, 16) Grass p-Value for 2007–2010 change (df = 1, 16)2007 2010 2007 2010

CON 18.7 ± 5.7a 22.8 ± 6.4a 0.395 81.3 ± 5.7a 78.3 ± 6.7a 0.533DIST 20.8 ± 5.1a 31.2 ± 6.2a 0.64 79.2 ± 5.1a 71.0 ± 6.8a 0.175REM 19.3 ± 4.5a 26.4 ± 6.5a 0.002� 80.7 ± 4.5a 75.1 ± 7.0a 0.375HERB 24.0 ± 6.6a 14.8 ± 5.4b 0.186 76.0 ± 6.6a 88.1 ± 5.6b 0.010�

OUT 50.1 ± 7.1b 46.2 ± 6.8a 0.078 49.9 ± 7.1b 54.4 ± 6.7a 0.109

Native forb p-Value for 2007–2010 change Native grass p-Value for 2007–2010 change2007 2010 2007 2010

CON 12.9 ± 3.8a 13.1 ± 5.1a 0.960 57.9 ± 4.7a 54.6 ± 4.8a 0.549DIST 13.8 ± 3.5a 17.3 ± 4.5a 0.078 58.7 ± 5.2a 55.7 ± 8.1a 0.587REM 11.5 ± 3.0a 16.1 ± 5.1a 0.231 55.5 ± 3.1a 54.7 ± 6.5a 0.876HERB 18.1 ± 5.7a 7.6 ± 3.7a 0.030� 59.6 ± 6.4a 62.0 ± 5.2a 0.735OUT 43.5 ± 6.2b 38.2 ± 6.2b 0.182 34.7 ± 7.3b 43.4 ± 6.0a 0.015�

Non-nativeforb

p-Value for 2007–2010change

Non-nativegrass

p-Value for 2007–2010 change

2007 2010 2007 2010CONT 5.8 ± 2.3a 8.6 ± 2.9a 0.242 23.5 ± 5.3a 23.7 ± 5.4a 0.976DIST 7.0 ± 1.9a 11.7 ± 3.4a 0.143 20.5 ± 3.9a 15.3 ± 4.6a 0.216REM 7.9 ± 3.5a 8.8 ± 3.7a 0.821 25.1 ± 5.2a 20.4 ± 4.9a 0.278HERB 6.0 ± 2.5a 4.3 ± 2.1a 0.388 16.4 ± 5.1a 26.2 ± 6.1a 0.117OUT 6.7 ± 3.5a 7.4 ± 2.8a 0.827 15.2 ± 6.7a 11.1 ± 5.8a 0.0801

T.C. Skurski et al. / Biological Conservation 159 (2013) 292–302 299

occurred in at least one herbicide plot before herbicide treatmentbut in zero herbicide plots 3 years after treatment (Table S5). Theseare infrequent species (found in only one or two of the 18 totaltreatment plots); thus, their presence/absence did not have a sub-stantive effect on the mean species richness or composition.

3.6. Recovery toward uninvaded conditions

The total amount of live vegetation cover did not differ insideversus outside C. stoebe populations before or 3 years after treat-ment. However, there were significant differences in the propor-tional representation of forbs, grasses and native/non-nativespecies (Table 2). In 2007, forbs and grasses comprised equal pro-portions of the total plant cover in uninvaded plant communities,whereas forb cover was significantly lower and grass cover signif-icantly higher within C. stoebe invaded communities (Table 3). Dur-ing the 3 years following treatment, forb cover increased and grasscover decreased inside C. stoebe invaded communities except in theareas treated with herbicide. In these plots, forb cover remainedsignificantly lower and grass cover significantly higher, than inuninvaded communities (Table 3). The proportion of the total veg-etation cover comprised of native forbs increased over time in alltreatments except herbicide plots. Before treatment, native forbscomprised approximately 18.1% ± 5.7% of the total vegetation cov-er in herbicide plots, whereas 3 years after treatment they com-prised only 7.6% ± 3.7%, which was a significant reduction(F1,16 = 6.95, p = 0.030). Nevertheless, the proportion of total vege-tation cover represented by native forbs was significantly higher inuninvaded communities both at the beginning and the end of theexperiment (Table 3). Uninvaded communities had significantlyless native grass cover in 2007 relative to C. stoebe invaded com-munities, but this difference had dissipated by 2010 due to a signif-icant increase in native grass cover in uninvaded communities. Theproportion of non-native grass was also lower in uninvaded com-munities, although this difference was not significantly differentfrom inside C. stoebe invaded communities (Table 3).

The percent bare ground and litter cover did not differ betweentreated and uninvaded plant communities before, or 3 years after,treatment (bare ground 2007: F4,77 = 1.648, p = 0.186; 2010:

F4,77 = 1.452, p = 0.24; litter 2007: F4,77 = 1.968, p = 0.123; 2010:F4,77 = 2.089, p = 0.105). While there were differences in speciesfrequency between treatment plots and uninvaded communities(Table S5), species composition (abundance-based) was not signif-icantly different (2007: F4,77 = 0.568, p = 0.967; 2010: F4,77 = 0.760,p = 0.818).

Of the 22 response variables, native grass cover, C. stoebe coverand C. stoebe density were the only parameters that changed signif-icantly (p < 0.05) in uninvaded plant communities from 2007 to2010 (Table 3). Native grass cover increased from 11% ± 2.2% in2007 to 17% ± 3.0% in 2010. The cover and density of C. stoebe ininitially uninvaded communities remained low 3 years after thestart of the experiment: C. stoebe cover in 2010 averaged2% ± 1.2% and C. stoebe density, six individuals ± 3 per 0.25 m2. Inour study area, C. stoebe generally exhibits a patchy distribution,which was reflected at the local-scale by the variability in coverand density in initially uninvaded communities. At the end of theexperiment, one-third of the initially uninvaded communities stillcontained no C. stoebe. The majority of plots that became invadedhad 1–2% C. stoebe cover (3–4 individuals per 0.25 m2) by 2010;although one plot had 5% C. stoebe cover (12 individuals) and an-other 12% C. stoebe cover (27 individuals) by the end of theexperiment.

4. Discussion

Our results highlight several ecological tradeoffs to considerwhen managing C. stoebe in sagebrush–grassland natural areas.Picloram, the common herbicide used for C. stoebe, was highlyeffective at reducing the cover of C. stoebe and had a positive effecton native grass cover 3 years after treatment, which has also beenshown in other systems (Rice et al., 1997; Rice and Toney, 1998;Sheley et al., 2000; Ortega and Pearson, 2010, 2011). However,we also found that herbicide treatment caused a steady and signif-icant increase in non-native grass cover in the 3 years followingtreatment, primarily B. tectorum—another non-native species ofgreat concern for conservation and management of natural areasin the Intermountain West. The tradeoff of one non-native plant

300 T.C. Skurski et al. / Biological Conservation 159 (2013) 292–302

species for another is an increasingly observed side effect of piclo-ram (Ortega and Pearson, 2010, 2011). Other potentially undesir-able consequences of herbicide treatment are reduced native forbcover and the loss of infrequent forb species. Unlike the findingsof Rice et al. (1997) and Rice and Toney (1998), native forb coverin our herbicide treatment plots had not returned to pre-treatmentlevels by 3 years post-treatment. Furthermore, our herbicide plotsdid not become more similar to uninvaded plots over time in termsof native forb cover. Our findings are in this regard are consistentwith several recent studies documenting detrimental effects ofpicloram on native grassland forbs, particularly arrowleaf balsam-root (Sheley and Denny, 2006; Crone et al., 2009; Rinella et al.,2009; Ortega and Pearson, 2010, 2011).

What was not previously known was the impact of C. stoebe it-self on sagebrush–grassland plant communities and how these im-pacts compare with those of herbicide treatment. Field-basedremoval experiments enabled us to quantify causal relationships.If C. stoebe was having a significant impact, we would expect tosee a proportional change in the response variable in the manualremoval plots greater than that seen with the disturbance treat-ment or no treatment. We found a significant increase in nativeforb cover in response to manual removal, which would seem toindicate that C. stoebe had been suppressing native forbs. However,there was an equivalent increase in native forbs with no treatmentand, to a lesser extent, with disturbance. There were no responsevariables, other than C. stoebe cover, for which the effect of manualremoval was greater than the response to either disturbance or notreatment. These results indicate that C. stoebe is not exerting sig-nificant negative impacts on the cover, richness, or biomass of na-tive forbs or grasses in our study system. C. stoebe shows noindication of displacing native species in these sagebrush–grass-lands because its presence has not lead to significant shifts in theexisting plant community or changes in the proportion of bareground. Furthermore, the coincident increase in cover of C. stoebe,native grasses, and native forbs in plots that received no treatmentsuggests that, in some communities, coexistence of native florawith C. stoebe is possible and that an increase in C. stoebe abun-dance does not necessarily lead to competitive exclusion.

As noted in the introduction, there are very few field-basedexperimental studies with which to compare our C. stoebe impactresults. Lesica and Shelly (1996) clearly showed detrimental im-pacts of C. stoebe in a field-based removal experiment; however,the target of impact was a rare native plant species, making com-munity-level comparisons difficult. In a microcosm setting, Maronand Marler (2008a, 2008b) demonstrated that seeding C. stoebeinto native monocultures and mixtures led to a significant reduc-tion in the biomass of several native forb and grass species. Thediscrepancy between these results and our findings in naturalgrassland communities may be related to differences in climate,soil resources, and/or the abundance and propagule supply of C.stoebe and native species. We have not observed, in our region,high density C. stoebe infestations (i.e. 60–100% cover) that havebeen reported in other regions (Tyser and Worley, 1992; Ridenourand Callaway, 2001).

The uninvaded areas that we examined had significantly higherforb cover, particularly native forb cover, and lower grass cover rel-ative to inside C. stoebe populations at the beginning of the exper-iment. All of the treatments (including no treatment) became moresimilar to uninvaded areas in terms of total forb cover, native forbcover, and total grass cover, except for the herbicide treatment.These latter (herbicide) plots became less similar to uninvadedareas over time, increasing in grass cover, particularly non-nativegrass cover, and decreasing in forb cover and richness, particularlynative forb cover and richness.

The amount of bare ground was not significantly different in-side versus outside C. stoebe populations and the amount of bare

ground did not change significantly over time or in response totreatments. While C. stoebe density and cover increased in some‘no treatment’ and previously uninvaded plots, C. stoebe densityand cover remained constant or decreased in other ‘no treatment’and previously uninvaded plots. Together, these findings suggestthat C. stoebe may not always rapidly invade sagebrush–grasslandcommunities, in spite of intermediate natural and anthropogenicdisturbance and persistent bare ground.

It is generally assumed that non-native plants exert negativeimpacts that justify their control and that after control actionsare taken, desirable properties of the community will improve, typ-ically unaided. However, it is increasingly recognized, particularlyin natural areas and wildland ecosystems, that non-native plantcontrol strategies often do not result in desirable long-term out-comes (D’Antonio et al., 2004; Reid et al., 2009). A lack of nativespecies recovery and secondary invasion are common hindrances(e.g. Mason and French, 2007; Turner et al., 2008; Rinella et al.,2009; Larson and Larson, 2010; Kettenring and Adams, 2011).Our results support the idea that ‘‘one size does not fit all’’ in termsof non-native plant management and that the existing conditionsand potential future conditions need to be considered in the deci-sion-making process (Seastedt et al., 2008). We do not advocateabandoning efforts to control C. stoebe. Rather, we suggest greaterconsideration of the actual impacts of C. stoebe and a weighing ofthis information against the costs and benefits of management(Rew et al., 2007). Because of ecological tradeoffs, such as those de-scribed in our study, there may be situations in which the benefitsof herbicide treatment do not justify the costs. If the primary man-agement objective is maximizing grass cover for forage, herbicidetreatment of C. stoebe may be beneficial and cost-effective. How-ever, this approach should be used cautiously in areas where B. tec-torum or other undesirable grass species occur; these may increaseat least as much, if not more, than the desirable forage species.Monitoring unmanaged non-native plant populations in naturalareas is a necessary component in this adaptive management pro-cess (Rew et al., 2007).

Loss of native forb cover and infrequent native forb speciesresulting from herbicide treatment are consequences that mustbe weighed against management objectives (Rinella et al., 2009).Forbs are an important component of sagebrush–grassland habi-tats (Mueggler and Stewart, 1980), providing food and habitat fororganisms across all trophic levels including soil biota, insects,and small and large mammals (Viketoft et al., 2005; Torstensonet al., 2006; Vogel et al., 2007; Cook et al., 2011). These organismsplay important roles in ecosystem processes, such as decomposi-tion, nutrient cycling, and pollination. Based on our results, manynative forbs in this region successfully co-exist with C. stoebe. Thus,typical sagebrush–grassland communities in natural areas (i.e.ones that retain a viable native plant component) that are beingmanaged for biodiversity and conservation may benefit more fromactivities that promote the existing desirable vegetation than fromC. stoebe removal. On the other hand, in communities harboringspecies of concern, such as the rare Mt. Sapphire rockress studiedby Lesica and Shelly (1996), direct control of C. stoebe may be nec-essary for the conservation of rare plant populations.

In conclusion, we have highlighted several ecological tradeoffsto consider when managing non-native plants in natural areas. C.stoebe had negligible impacts on the cover and richness of nativeforbs and grasses in sagebrush–grassland communities examinedin southwest Montana. Our experimental results indicate that itis possible for communities to maintain ecological value (i.e. nativespecies, forage production, wildlife habitat) despite the presence ofthis ubiquitous non-native plant. We cannot say whether this lackof significant plant community impacts by C. stoebe would holdtrue in severely disturbed areas or significantly different climates.In our study system, the common herbicide used to control C. sto-

T.C. Skurski et al. / Biological Conservation 159 (2013) 292–302 301

ebe affected the plant community much more than C. stoebe itselfby decreasing the cover and richness of native forbs and increasingthe cover of undesirable non-native grass. In natural areas andwildland ecosystems where, in addition to controlling non-nativeplants, management objectives include providing high-qualitywildlife habitat, conserving biological diversity, and maintainingecosystem services, careful consideration of the ecological and eco-nomic costs and benefits of non-native plant control is prudent.The loss of forbs and potential for secondary invasion may out-weigh the benefits gained by reducing non-native plant abun-dance. Sound management decisions will consider these tradeoffsin relation to the quantified impacts of non-native plantsthemselves.

Acknowledgements

For their help in the field and lab, we thank K. Neff, D. Campbell,M. Herstand, R. Marks, J. Patel, and P. Lawrence. We thank twoanonymous reviewers for thoughtful comments that improvedthe manuscript. Funding was provided through a USDA-ERSPREISM Grant, Cooperative Agreement No. 433-AEM-4-80111, aNRI-CSREES Grant, Award No. 2009-55320-05033, and a GraduateStudent Fellowship from the Montana Institute on Ecosystems.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.biocon.2012.10.017. These data include Google maps of the most important areasdescribed in this article.

References

Adair, R.J., Groves, R.H., 1998. Impact of Environmental Weeds on Biodiversity: AReview and Development of a Methodology. Biodiversity Group, EnvironmentAustralia, Canberra, Australia.

Bais, H.P., Vepachedu, R., Gilroy, S., Callaway, R.M., Vivanco, J.M., 2003. Allelopathyand exotic plant invasion: from molecules and genes to species interactions.Science 301, 1377–1380.

Blicker, P.S., Olson, B.E., Engel, R., 2002. Traits of the invasive Centaurea maculosaand two native grasses: effect of N supply. Plant Soil 247, 261–269.

Buckley, Y.M., 2008. The role of research for integrated management of invasivespecies, invaded landscapes and communities. J. Appl. Ecol. 45, 397–402.

Cadotte, M.W., Carscadden, K., Mirotchnick, N., 2011. Beyond species: functionaldiversity and the maintenance of ecological processes and services. J. Appl. Ecol.48, 1079–1087.

Callaway, R.M., Thelen, G.C., Barth, S., Ramsey, P.W., Gannon, J.E., 2004a. Soil fungialter interactions between the invader Centaurea maculosa and North Americannatives. Ecology 85, 1062–1071.

Callaway, R.M., Thelen, G.C., Rodriguez, A., Holben, W.E., 2004b. Soil biota and exoticplant invasion. Nature 427, 731–733.

Cook, S.P., Birch, S.M., Merickel, F.W., Lowe, C.C., Page-Dumroese, D., 2011. Bumblebee (Hymenoptera: Apidae) community structure on two sagebrush steppe sitesin southern Idaho. Pan-Pac. Entomol. 87, 161–171.

Corbin, J.D., D’Antonio, C.M., 2004. Competition between native perennial andexotic annual grasses: implications for an historical invasion. Ecology 85, 1273–1283.

Crone, E.E., Marler, M., Pearson, D.E., 2009. Non-target effects of broadleaf herbicideon a native perennial forb: a demographic framework for assessing andminimizing impacts. J. Appl. Ecol. 46, 673–682.

Cronquist, A., Holmgren, A.H., Holmgren, N.H., Reveal, J.L., Holmgren, P.K., 1977.Intermountain Flora – Volume 6: The Monocotyledons. Columbia UniversityPress, New York.

D’Antonio, C.M., Jackson, N.E., Horvitz, C.C., Hedberg, R., 2004. Invasive plants inwildland ecosystems: merging the study of invasion processes withmanagement needs. Front. Ecol. Environ. 2, 513–521.

Diaz, S., Cabido, M., 2001. Vive la difference. plant functional diversity matters toecosystem processes. Trends Ecol. Evol. 16, 646–655.

DiTomaso, J.M., 2000. Invasive weeds in rangelands: species, impacts, andmanagement. Weed Sci. 48, 255–265.

Dorn, R., 1984. Vascular Plants of Montana. Mountain West Publishing, Cheyenne,Wyoming.

Dow AgroSciences, 2001. Tordon 22K. Specimen label, C & P Press, New York.Downey, P.O., Williams, M.C., Whiffen, L.K., Auld, B.A., Hamilton, M.A., Burley, A.L.,

Turner, P.J., 2010. Managing alien plants for biodiversity outcomes – the needfor triage. Invas. Plant Sci. Man. 3, 1–11.

Duncan, C.L., 2005. Spotted knapweed Centaurea stoebe L. In: Duncan, C.L., Clark, J.K.(Eds.), Invasive Plants of Range and Wildlands and Their Environmental,Economic, and Societal Impacts. Weed Science Society of America, Lawrence,KS, pp. 51–68.

Gude, P.H., Hansen, A.J., Rasker, R., Maxwell, B., 2006. Rates and drivers of ruralresidential development in the Greater Yellowstone. Landscape Urban Plann.77, 131–151.

Harner, M.J., Mummey, D.L., Stanford, J.A., Rillig, M.C., 2010. Arbuscular mycorrhizalfungi enhance spotted knapweed growth across a riparian chronosequence.Biol. Invasions 12, 1481–1490.

Hirsch, S.A., Leitch, J.A., 1996. The Impact of Knapweed on Montana’s Economy.Department of Agricultural Economics, Agricultural Experiment Station, NorthDakota State University.

Kedzie-Webb, S.A., Sheley, R.L., Borkowski, J.J., Jacobs, J.J., 2001. Relationshipsbetween Centaurea maculosa and indigenous plant assemblages. West. N. Am.Nat. 61, 43–49.

Kettenring, K.M., Adams, C.R., 2011. Lessons learned from invasive plant controlexperiments: a systematic review and meta-analysis. J. Appl. Ecol. 48, 970–979.

Larson, D.L., Larson, J.L., 2010. Control of one invasive plant species allows exoticgrasses to become dominant in northern Great Plains grasslands. Biol. Conserv.143, 1901–1910.

Lavin, M., Seibert, C., 2009. Grasses of Montana. Montana State UniversityHerbarium, Bozeman. <http://www.montana.edu/mlavin/herb/mtgrass.pdf>.

Lesica, P., Shelly, J.S., 1996. Competitive effects of Centaurea maculosa on thepopulation dynamics of Arabis fecunda. Bull. Torrey Bot. Club 123, 111–121.

Mack, R.N., Simberloff, D., Lonsdale, W.M., Evans, H., Clout, M., Bazzaz, F.A., 2000.Biotic invasions: causes, epidemiology, global consequences, and control. Ecol.Appl. 10, 689–710.

Marler, M.J., Zabinski, C.A., Callaway, R.M., 1999. Mycorrhizae indirectly enhancecompetitive effects of an invasive forb on a native bunchgrass. Ecology 80,1180–1186.

Maron, J.L., Marler, M., 2008a. Effects of native species diversity and resourceadditions on invader impact. Am. Nat. 172, S18–S33.

Maron, J.L., Marler, M., 2008b. Field-based competitive impacts between invadersand natives at varying resource supply. J. Ecol. 96, 1187–1197.

Mason, T.J., French, K., 2007. Management regimes for a plant invader differentiallyimpact resident communities. Biol. Conserv. 136, 246–259.

May, L., Baldwin, L., 2011. Linking field based studies with greenhouse experiments:the impact of Centaurea stoebe (=C. maculosa) in British Columbia grasslands.Biol. Invasions 13, 919–931.

Montana Natural Heritage Program (MNHP), 2011. Montana field guide. Retrievedon March 14, 2012. <http://FieldGuide.mt.gov>.

Mortensen, D.A., Rauschert, E.S.J., Nord, A.N., Jones, B.P., 2009. Forest roads facilitatethe spread of invasive plants. Invas. Plant Sci. Man. 2, 191–199.

Mueggler, W.F., Stewart, W.L., 1980. Grassland and Shrubland Habitat Types ofWestern Montana. Intermountain Forest and Range Experiment Station, USForest Service, Ogden, Utah.

NRCS, 2012. Natural Resources Conservation Service, USDA. Plants database,<http://www.plants.usda.gov/java/noxComposite>.

Oksanen, J., 2011. Multivariate analysis of ecological communities in R: vegantutorial. <http://cc.oulu.fi/~jarioksa/opetus/metodi/vegantutor.pdf>.

Oksanen, J., Guillaume, B.F., Kindt, R., Legendre, P., Minchin, P.R., O’Hara, R.B.,Simpson, G.L., Solymos, P., Stevens, M.H.H., Wagner, H. 2011. vegan: communityecology package. R package version 2.0-2. <http://CRAN.R-project.org/package=vegan>.

Olliff, T., Renkin, R., McClure, C., Miller, P., Price, D., Reinhart, D., Whipple, J., 2001.Managing a complex exotic vegetation program in Yellowstone National Park.West. N. Am. Nat. 61, 347–358.

Ortega, Y.K., Pearson, D.E., 2005. Weak vs. strong invaders of natural plantcommunities: assessing invasibility and impact. Ecol. Appl. 15, 651–661.

Ortega, Y.K., Pearson, D.E., 2010. Effects of Picloram application on communitydominants vary with initial levels of spotted knapweed (Centaurea stoebe)invasion. Invas. Plant Sci. Man. 3, 70–80.

Ortega, Y.K., Pearson, D.E., 2011. Long-term effects of weed control with picloramalong a gradient of spotted knapweed invasion. Rangeland Ecol. Manage. 64,67–77.

Parks, C.G., Radosevich, S.R., Endress, B.A., Naylor, B.J., Anzinger, D., Rew, L.J.,Maxwell, B.D., Dwire, K.A., 2005. Natural and land-use history of the Northwestmountain ecoregions (USA) in relation to patterns of plant invasions. Perspect.Plant Ecol. Evol. Syst. 7, 137–158.

Pearson, D.E., Ortega, Y.K., 2009. Managing invasive plants in natural areas: movingbeyond control. In: Kingely, R.V. (Ed.), Weeds: Management, Economic Impactsand Biology. Nova Publishers, Hauppauge, NY, pp. 1–21.

Pinheiro, J., Bates, D., DebRoy, S., Sarkar, D., R Development Core Team, 2012. nlme:linear and nonlinear mixed effects models. R package version 3.1-103.

R Development Core Team, 2011. R: a language and environment for statisticalcomputing. R Foundation for Statistical Computing, Vienna, Austria. ISBN: 3-900051-07-0. <http://www.R-project.org/>.

Reid, A.M., Morin, L., Downey, P.O., French, K., Virtue, J.G., 2009. Does invasive plantmanagement aid the restoration of natural ecosystems? Biol. Conserv. 142,2342–2349.

Rew, L.J., Lehnhoff, E.A., Maxwell, B.D., 2007. Non-indigenous species managementusing a population prioritization framework. Can. J. Plant Sci./Rev. Can.Phytotech. 87, 1029–1036.

Rice, P.M., Toney, J.C., 1998. Exotic weed control treatments for conservation offescue grassland in Montana. Biol. Conserv. 85, 83–95.

302 T.C. Skurski et al. / Biological Conservation 159 (2013) 292–302

Rice, P.M., Toney, J.C., Bedunah, D.J., Carlson, C.E., 1997. Plant community diversityand growth form responses to herbicide applications for control of Centaureamaculosa. J. Appl. Ecol. 34, 1397–1412.

Ridenour, W.M., Callaway, R.M., 2001. The relative importance of allelopathy ininterference. The effects of an invasive weed on a native bunchgrass. Oecologia126, 444–450.

Rinella, M.J., James, J.J., 2010. Invasive plant researchers should calculate effectsizes, not p-values. Invas. Plant Sci. Man. 3, 106–112.

Rinella, M.J., Maxwell, B.D., Fay, P.K., Weaver, T., Sheley, R.L., 2009. Control effortexacerbates invasive-species problem. Ecol. Appl. 19, 155–162.

Roché, B.F., Talbott, C.J., 1986. The collection history of Centaurea found inWashington State. Agriculture Research Center Research BulletinXBO978. Washington State University Cooperative Extension Service,Pullman, WA.

Seastedt, T.R., Hobbs, R.J., Suding, K.N., 2008. Management of novel ecosystems: arenovel approaches required? Front. Ecol. Environ. 6, 547–553.

Sheley, R.L., Denny, M.K., 2006. Community response of nontarget species toherbicide application and removal of the nonindigenous invader Potentilla rectaL. West. N. Am. Nat. 66, 55–63.

Sheley, R.L., Duncan, C.A., Halstvedt, M.B., Jacobs, J.S., 2000. Spotted knapweed andgrass response to herbicide treatments. J. Range Manage. 53, 176–182.

Simberloff, D., 2011. How common are invasion-induced ecosystem impacts? Biol.Invasions 13, 1255–1268.

Thorpe, A.S., Archer, V., DeLuca, T.H., 2006. The invasive forb, Centaurea maculosa,increases phosphorus availability in Montana grasslands. Appl. Soil Ecol. 32,118–122.

Thorpe, A.S., Thelen, G.C., Diaconu, A., Callaway, R.M., 2009. Root exudate isallelopathic in invaded community but not in native community: field evidencefor the novel weapons hypothesis. J. Ecol. 97, 641–645.

Torstenson, W.L.F., Mosley, J.C., Brewer, T.K., Tess, M.W., Knight, J.E., 2006. Elk, muledeer, and cattle foraging relationships on foothill and mountain rangeland.Rangeland Ecol. Manage. 59, 80–87.

Turner, P.J., Scott, J.K., Spafford, H., 2008. The ecological barriers to the recovery ofbridal creeper (Asparagus asparagoides (L.) Druce) infested sites: Impacts onvegetation and the potential increase in other exotic species. Austral. Ecol. 33,713–722.

Tyser, R.W., Key, C.H., 1988. Spotted knapweed in natural area fescue grasslands –an ecological assessment. Northwest Sci. 62, 151–160.

Tyser, R.W., Worley, C.A., 1992. Alien flora in grasslands adjacent to road and trailcorridors in Glacier National Park, Montana (USA). Conserv. Biol. 6, 253–262.

USFS, BLM, USFWS, NPS, 2010. Reports and databases obtained directly from USForest Service, Bureau of Land Management, US Fish and Wildlife Service, andthe National Park Service on non-native plant management: area infested, areatreated, control strategies and chemicals used.

Viketoft, M., Palmborg, C., Sohlenius, B., Huss-Danell, K., Bengtsson, J., 2005. Plantspecies effects on soil nematode communities in experimental grasslands. Appl.Soil Ecol. 30, 90–103.

Vilà, M., Espinar, J.L., Hejda, M., Hulme, P.E., Jarošík, V., Maron, J.L., Pergl, J.,Schaffner, U., Sun, Y., Pyšek, P., 2011. Ecological impacts of invasive alien plants:a meta-analysis of their effects on species, communities and ecosystems. Ecol.Lett. 14, 702–708.

Vitousek, P.M., D’Antonio, C.M., Loope, L.L., Rejmánek, M., Westbrooks, R., 1997.Introduced species: a significant component of human-caused global change.NZ. J. Ecol. 21, 1–16.

Vogel, J.A., Debinski, D.M., Koford, R.R., Miller, J.R., 2007. Butterfly responses toprairie restoration through fire and grazing. Biol. Conserv. 140, 78–90.

Walling, S.Z., Zabinski, C.A., 2006. Defoliation effects on arbuscular mycorrhizae andplant growth of two native bunchgrasses and an invasive forb. Appl. Soil Ecol.32, 111–117.

Wickham, H., 2009. Ggplot2: Elegant Graphics for Data Analysis. Springer, NewYork.

Zabinski, C.A., Quinn, L., Callaway, R.M., 2002. Phosphorus uptake, not carbontransfer, explains arbuscular mycorrhizal enhancement of Centaurea maculosain the presence of native grassland species. Funct. Ecol. 16, 758–765.