<<

The Impact of Exotic Purple Loosestrife (Lythrum salicaria) on Wetland Abundances Author(s): Brian G. Tavernia and J. Michael Reed Source: The American Midland Naturalist, 168(2):352-363. 2012. Published By: University of Notre Dame URL: http://www.bioone.org/doi/full/10.1674/0003-0031-168.2.352

BioOne (www.bioone.org) is a nonprofit, online aggregation of core research in the biological, ecological, and environmental sciences. BioOne provides a sustainable online platform for over 170 journals and books published by nonprofit societies, associations, museums, institutions, and presses. Your use of this PDF, the BioOne Web site, and all posted and associated content indicates your acceptance of BioOne’s Terms of Use, available at www.bioone.org/page/terms_of_use. Usage of BioOne content is strictly limited to personal, educational, and non-commercial use. Commercial inquiries or rights and permissions requests should be directed to the individual publisher as copyright holder.

BioOne sees sustainable scholarly publishing as an inherently collaborative enterprise connecting authors, nonprofit publishers, academic institutions, research libraries, and research funders in the common goal of maximizing access to critical research. Am. Midl. Nat. (2012) 168:352–363

The Impact of Exotic Purple Loosestrife (Lythrum salicaria) on Wetland Bird Abundances

1 BRIAN G. TAVERNIA AND J. MICHAEL REED Biology Department, Tufts University, Medford, Massachusetts 02155

ABSTRACT.—The exotic invasive wetland plant purple loosestrife (Lythrum salicaria) is often considered to have negative impacts on native plant and , but this is debated. Clarifying its influence would provide insight into appropriate management actions following invasion. We investigated the influence of L. salicaria cover and density on abundances of wetland bird species that are associated with a variety of vegetation structures. We found evidence of relationships between L. salicaria measures and abundance for most species we examined, but these relationships did not always agree with our predictions based on species’ habitat associations. Some bird species positively responded whereas others negatively responded to increasing L. salicaria cover or density. Response curves varied in complexity and included linear and quadratic relationships as well as interactions. Our results suggested that L. salicaria did not categorically decrease habitat quality for all wetland bird species, and it may have had a positive influence on quality for some species. This ambiguity is not unique to L. salicaria invasion but applies to many changing habitat features. Therefore, there is likely no single appropriate strategy for managing L. salicaria when the goal is to maintain a diverse avian community in which species have divergent habitat preferences.

INTRODUCTION Exotic invasive plants can alter the composition and structure of vegetation communities and potentially change the availability of resources (e.g., nest sites, foraging habitat) needed by breeding (e.g., Schmidt and Whelan, 1999; Heckscher, 2004). One such species is purple loosestrife (Lythrum salicaria), an exotic invasive perennial wetland plant introduced to the northeastern U.S.A. and Canada from Eurasia in the 1800s; it now occurs in 48 U.S. states and nine Canadian provinces (Thompson et al., 1987; Wilson et al., 2004). The spread of exotic invasive wetland plants can be facilitated by landscape disturbances such as those associated with urbanization (Silliman and Bertness, 2004; Zedler and Kercher, 2004). Consequently, L. salicaria invasion might be one mechanism by which observed landscape changes affect wetland bird distributions (e.g., DeLuca et al., 2004; Tavernia and Reed, 2010). Although the impact of L. salicaria on wetland plant communities is debated (e.g., Anderson, 1995; Blossey et al., 2001), evidence suggests that it reduces aboveground biomass of co-occurring native plants (Farnsworth and Ellis, 2001) and shifts the structure of emergent vegetation to a more shrub-like form (Hill, 2000). Because wetland bird species differ in their preferences for different vegetation structures (Weller, 1999), changes in structure as a results of L. salicaria invasion could alter the distribution and abundance of wetland bird species. We found few published investigations that have examined effects of Lythrum salicaria invasion on wetland birds given the extent of its invasion in North America. Consequently, we tested predictions regarding the influence of L. salicaria cover and density on the breeding abundance of a suite of wetland bird species. Our predictions were based on the structure of vegetation stands typically considered habitat for these species and on

1 Corresponding author’s present address: Department of Forestry, University of Missouri, Columbia 65211; e-mail: [email protected]

352 2012 TAVERNIA &REED:INVASIVE PURPLE LOOSESTRIFE AND WETLAND BIRDS 353

TABLE 1.—Models relating bird species abundances to Lythrum salicaria cover and density and a suite of additional habitat variables at 64 survey points within 30 wetlands. Habitat variables are: water depth (WD), fine-leaved emergent cover (FC), cattail cover (CC), shrub density (SHD), shrub cover (SHC), vegetation height (VH), tree and snag number (TSN), water cover (WC), L. salicaria cover (LC), and L. salicaria density (LD). The best model (DAICc 5 0) and others receiving support (DAICc # 2) are reported for each bird species. Akaike weights (wi) represent the likelihood that each candidate model is the best model among those considered. Pseudo-r2 values are reported as measures of the amount of variability explained

2 Species Model DAICc wi Pseudo-r Virginia rail 21.15 + 1.24WD 2 0.25FC + 0.34CC + 2.42WC + 1.27LD 0 0.33 0.22 20.67 + 1.12WD 2 0.65FC 2 0.07CC + 2.07WC 0.40 0.27 0.19 20.93 + 1.35WD 2 0.44FC + 0.05CC + 2.09WC + 1.09LC 1.00 0.20 0.21 21.32 + 1.07WD 2 0.07FC + 0.58CC + 2.64WC 1.56 0.15 0.24 + 1.04LD + 3.94LD*WC marsh 23.87 + 1.91WD + 1.94CC + 0.94VH 2 3.48WC 2 2.96LD 0 0.52 0.25 23.28 + 2.23WD + 2.13CC + 0.62VH 2 2.97WC 1.03 0.31 0.20 song sparrow 2.30 + 1.38SHD 2 0.51SHC + 0.02TSN 2 0.77VH 2 0 0.50 0.27 0.99LC + 1.01LD 2 2.58LC*LD 2.09 + 2.45SHD 2 0.63SHC + 0.004TSN 2 0.70VH 1.72 0.21 0.17 2.10 + 2.38SHD 2 0.60SHC 2 0.01TSN 2 0.71VH 2 0.44LC 1.88 0.20 0.19 swamp sparrow 3.08 + 1.90LC 0 0.52 0.07 red-winged 7.52 2 1.38SHD + 1.98SHC + 1.32CC + 4.40WC 0 0.50 0.27 blackbird 7.27 + 0.76SHD + 1.59SHC + 1.06CC + 4.27WC 1.78 0.21 0.34 + 1.79LC 2 2.34LD + 9.07LC*LD previously published results. Specifically, dense stands of vegetation may impede the movement of Virginia rails (Rallus limicola) (Conway, 1995), a relatively ambulatory species, so we expected a negative relationship between the density of L. salicaria and this species’ abundance. However, this effect may be reduced if L. salicaria stands are interspersed with a relatively open cover type, such as water pools. Consequently, with respect to Virginia rail abundance, we tested for the presence of an interaction between L. salicaria density and the cover of water pools. For marsh ( palustris), we predicted a negative linear response to increasing L. salicaria cover and density because this species typically breeds in areas with simple vertically structured vegetation (e.g., cattail) (Kroodsma and Verner, 1997). We expected the abundances of yellow warblers (Setophaga petechia), common yellowthroats (Geothlypis trichas), and song sparrows (Melospiza melodia) to respond in a positive linear fashion to L. salicaria cover and density, as these species are typically associated with shrubby areas (Guzy and Ritchison, 1999; Lowther et al., 1999; Arcese et al., 2002). Hill (2000) reported a possible positive, unimodal relationship between swamp sparrow (Melospiza georgiana) densities and L. salicaria cover and density and suggested that L. salicaria may provide song posts and nest building opportunities. Consequently, we predicted positive, unimodal relationships between swamp sparrow abundance and Lythrum salicaria cover and density. Previous studies have reported no relationship between L. salicaria and red-winged blackbird (Agelaius phoeniceus) abundance or nest success (Whitt et al., 1999; Hill, 2000; Maddox and Wiedenmann, 2005), so we expected no relationship between red-winged blackbird abundance and L. salicaria cover or density. A clear understanding of how these bird species respond to L. salicaria may help to predict likely impacts of wetland invasion by L. salicaria and can inform evaluation of alternative management activities such as L. salicaria eradication vs. control. 354 THE AMERICAN MIDLAND NATURALIST 168(2)

METHODS Study wetlands.—We conducted our study in the Greater Boston area of eastern Massachusetts (42u449 to 42u29N and from 71u289 to 70u579W). Historically, agricultural conversion and road and building construction have contributed to the loss of 58–64% of the original wetland habitat of Massachusetts. Currently, wetlands cover 6–7% of the state (United States Fish and Wildlife Service, 1995). Herbarium specimens suggest that Lythrum salicaria was established in Massachusetts at least as early as 1831 (Stuckey, 1980), and it now occurs throughout (Sorrie and Somers, 1999). Using aerial photographs taken in Jul. of 2007 (E Google, date accessed: 3/15/2009), we selected a subset of 94 candidate wetlands in the Greater Boston area meeting two selection criteria: (1) the wetland appeared to contain areas of emergent vegetation where Lythrum salicaria was prevalent and (2) the wetland was large enough to accommodate at least two 50-m radius bird survey points separated by a distance of 200 m. We selected sites with enough space for at least two survey points to increase the number of points that could be surveyed in a morning. Sixty-nine candidate wetlands were eliminated from consideration for one of two reasons: (1) field visits indicated a lack of L. salicaria in the wetland (i.e., aerial photograph interpretation was not always reliable) or (2) we were unable to get the landowners’ permissions to access wetlands. We increased the total number of study wetlands to 30 by including five wetlands observed during the course of field visits that met our two selection criteria and for which we obtained permission to survey. Wetland bird surveys.—We established pairs of survey points within each wetland site. For each wetland site, survey points were $50 m from an upland edge, $200 m from other survey points, and were within vegetation patches that qualitatively differed in Lythrum salicaria cover and density as determined by a non-intrusive, visual inspection (see Results for subsequent quantitative description). In two of our larger wetlands, we were able to establish an additional pair of survey points. At each point, we surveyed birds in a 50-m fixed-radius circular plot, counting all unique individuals heard or seen. Surveys were conducted from 0500 to 1000 h when there was no sustained rain and wind was ,20 km/h. We conducted surveys twice at each point, separated by at least 2 wk, as our target species differed in their breeding phenologies. Surveys were conducted from 15 May to 30 Jun. 2009, and consisted of a 10-min passive visual and listening period followed by a 6-min broadcast period during which we played calls of the behaviorally cryptic Virginia rail to increase its detection (Conway and Gibbs, 2005). Counts were preceded by a 2-min period to allow birds to adjust to our presence. Species targeted during the passive observation period included: marsh wren, yellow warbler, common yellowthroat, song sparrow, swamp sparrow, and red-winged blackbird. Surveys followed a removal protocol that allowed us to investigate and remove bias that increasing L. salicaria cover and density may have had on species detection probabilities (Farnsworth et al., 2002) (see Lythrum salicaria and Wetland Bird Abundances). Specifically, we divided passive and broadcast periods into 2-min intervals and recorded the interval in which each bird was first detected. The removal protocol assumes that bird populations remain closed during the survey period, and given the small territory sizes of the species that we surveyed, this was a reasonable assumption. Each minute of broadcast consisted of a 30-sec broadcast of calls followed by 30 sec of silence. The call broadcast system consisted of an iPod Nano (Apple Inc.) and portable speakers (SDI Technologies, Inc., model: iH2OW). Habitat surveys.—We measured vegetation and other habitat features along three 50-m transects radiating away from each bird survey point at 0u, 120u, and 240u. Habitat sampling points were established at 10-m intervals along each transect, resulting in 5 points per 2012 TAVERNIA &REED:INVASIVE PURPLE LOOSESTRIFE AND WETLAND BIRDS 355 transect (15 habitat sampling points per bird survey point). At each habitat sampling point, we placed a 1-m2 frame and recorded occupancy (presence/absence) of plants associated with vegetation types as well as the presence of open water ($0.25 m2). Vegetation types consisted of individual species or groups of species with similar structure and included: Lythrum salicaria, cattail ( spp.), fine-leaved emergent (e.g., Carex, Scirpus, Spartina) and shrub (e.g., Myrica). To assess the density of each vegetation type, we held a 1-m long pole horizontally at a height of 1-m above the ground or water surface and counted the number of touches made by vegetation types within each decimeter. For some vegetation types (e.g., L. salicaria, cattail), dead plants from previous years still contributed considerably to the structure present at a sampling point; consequently, dead individuals were included in density determinations. We measured density at a 1-m height because (1) we were interested in dominant vegetation types that all exceed 1-m in height and (2) high densities of shorter plant species (e.g., ferns) precluded accurate density determinations for dominant vegetation types at lower heights. We acknowledge that, in some instances, density estimates may have been biased against fine-leaved emergent species with shorter stems (e.g., Carex). At each habitat sampling point, we also recorded maximum vegetation height (overall, not specific to a vegetation type), maximum water depth, and the number of trees and snags within 2.5 m. For each wetland bird survey point, cover values for each vegetation type and open water were determined as the proportion of occupied habitat sampling points. Across the 15 habitat sampling points for each bird survey point, we averaged density values (number of touches per decimeter along a 1-m horizontal pole) of each vegetation type, maximum vegetation height, water depth and the number of trees and snags. We measured both cover and density of vegetation types because bird species may respond to interactions between these two metrics. For example, a vegetation type may occupy a high proportion of an area (i.e., high cover value), but birds may only respond to the vegetation type when it has also achieved a high density. Habitat surveys were conducted from 14 Jun. to 10 Jul. 2009. Habitat surveys occurred after the emergence of wetland plants, and while plant height increased during the survey period, the stem density would not have changed appreciably. Lythrum salicaria and wetland bird abundances.—If increasing Lythrum salicaria cover or density influences species detection probabilities, then an observed decrease of a species related to L. salicaria cover or density could be an artifact of counts unadjusted for detection probability. We used the program SURVIV (White, 1983) in conjunction with species- specific detection histories obtained via removal methods (Farnsworth et al., 2002) to estimate species detection probabilities in areas with low vs. high L. salicaria cover and density. To define low vs. high L. salicaria conditions, we looked for natural breaks in frequency distributions of bird survey points with respect to L. salicaria cover and density. Based on these distributions, we defined our cut-offs with respect to cover and density as: #60% cover vs. .60% cover, #0.35 touches/dm vs. .0.35 touches/dm. For five of our seven target species, detection probabilities could be estimated using SURVIV and there was no indication of a difference in detection probability across plots. Small sample sizes prevented SURVIV from producing detection probability estimates for marsh wrens and Virginia rails. For these species, a simple comparison of the proportion of individuals detected in each survey interval for low and high L. salicaria points suggested that detection probabilities did not differ. On the basis of these results, we used raw count data for our statistical models relating bird species abundances to habitat features. We used variance inflation factors to identify and sequentially remove habitat variables showing evidence of collinearity with other predictors (Zuur et al., 2009a). For these 356 THE AMERICAN MIDLAND NATURALIST 168(2) analyses, we considered a variance inflation factor $3 to be an indication of strong collinearity with other predictor variables. This procedure was carried out using the vif function (CAR library; Fox, 2009) in R v. 2.10.1 (R Development Core Team, 2009) and led to the elimination of cattail density and fine-leaved emergent density. Based on our predictions, we defined sets of species-specific, a priori models to evaluate the influence of Lythrum salicaria cover and density on abundance. For all bird species, we evaluated the following models: Abundance~intercept Abundance~LC Abundance~LD Abundance~LCzLDzLC Ã LD where LC is L. salicaria cover and LD is L. salicaria density. For four species, our models included additional, non-L. salicaria habitat variables based on the following citations for each species. (1) Virginia rail models included cattail cover, fine-leaved emergent cover, water cover, and water depth (Conway, 1995). (2) Marsh wren models included cattail cover, water cover, water depth, and vegetation height (Kroodsma and Verner, 1997). (3) Shrub cover, shrub density, tree and snag density, and vegetation height were included in song sparrow models (Arcese et al., 2002). (4) Red-winged blackbird models included cattail cover, shrub cover, shrub density, and water cover (Yasukawa and Searcy, 1995). Non-L. salicaria variables did not improve the fit of an intercept-only model for the yellow warbler, common yellowthroat, and swamp sparrow, so these variables were not included in models for these species. When non-L. salicaria habitat variables were included in models for the other bird species, our objective was to help ensure that L. salicaria effects we detected were not due to correlations with other habitat measures that might better explain the variance in our dependent variable. We considered additional models for the swamp sparrow and Virginia rail. Because non- linear relationships have been suggested for the swamp sparrow (Hill, 2000), we evaluated two additional models: Abundance~LCzLC2 Abundance~LDzLD2 For the Virginia rail, we considered an additional model to determine if the influence of L. salicaria density depended on water cover: Abundance~LDzWCzLD Ã WC where WC is water cover; this model for the Virginia rail also included the non-L. salicaria habitat variables listed above. Model performances were compared using AICc scores and Akaike weights. The model with the lowest AICc score was considered the best model, whereas those within 2 AICc units were considered to be competitive (Burnham and Anderson, 2002). If the intercept-only model was best or was within 2 AICc units of the best model, we considered this evidence that Lythrum salicaria had a negligible influence on a species’ abundance. We used Akaike weights, or model probabilities, to determine the relative support for models in the competitive model set. Pseudo- r2 values (Kramer, 2005; Zuur et al., 2009b) were also reported as measures of model fit. When models including interactions received support, we explored the nature of interactions using simple slopes (Quinn and Keough, 2002). This approach examines slopes of correlations between abundance and a habitat variable at 2012 TAVERNIA &REED:INVASIVE PURPLE LOOSESTRIFE AND WETLAND BIRDS 357 three values of a second habitat variable (mean and mean 6 3 standard errors). For example, we investigated how the slope relating Virginia rail abundance to L. salicaria density changed at three levels of water cover. Variables involved in interactions were centered using their mean values to avoid collinearity with the interaction term (Quinn and Keough, 2002). Models relating species’ abundances to habitat features at bird survey points were fitted using regression techniques in R v 2.10.1 and were species specific. Histograms of species- specific counts averaged across our two visits to each survey point indicated that a variety of statistical distributions was present, with some species showing normal distributions (yellow warbler, common yellowthroat, swamp sparrow, red-winged blackbird), others with Poisson distributions (Virginia rail, song sparrow) and one with a zero-inflated distribution (marsh wren). Given the nested nature of our data (i.e., pairs of survey points nested within wetlands), we initially used the gls and lme functions (NLME library; Pinheiro et al., 2009) to assess the fit of fixed effects vs. mixed effects models including wetland identity as a random variable; error distributions were assumed to be either Gaussian or Poisson as appropriate. The maximum observed count at a point across our two visits was used as the response variable for Poisson regression models. Unless AICc scores indicated stronger support for mixed-effect models (i.e., .2 units lower), a fixed effects model was implemented using the glm function (STATS library; R Development Core Team, 2009). Such a comparison was not possible for marsh wrens using the zero-inflated negative bionomal (ZINB) regression function, zeroinfl (PSCL library; Jackman, 2009). However, the inclusion of site identity in the ZINB regression model did not improve model fit, suggesting that a fixed-effects approach was appropriate. For marsh wrens, we used a ZINB regression because AICc scores indicated that it provided a better fit than did a zero-inflated Poisson regression model. For all but one species, plots of model residuals vs. predicted values showed homogeneous variance, as was the case when residuals were plotted against predictor variables. For yellow warblers, it was necessary to cube abundance estimates to achieve homogeneous scatter of residuals. There was no evidence of overdispersion in our Poisson models (W , 1.5, Zuur et al., 2009b). Moran’s I test (ArcMap 9.2) suggested that model residuals for the swamp sparrow and red-winged blackbird showed negative spatial autocorrelation whereas those for the marsh wren and song sparrow showed positive spatial autocorrelation. Models incorporating spatial effects into error terms or covariates (Zuur et al., 2009b; Beale et al., 2010) failed to improve model fit and did not result in independent residuals. This was likely because our residuals had anisotropic patterns of autocorrelation; methods for coping with anisotropic patterns of autocorrelation are still being developed (Carl et al., 2008). For these reasons, we reported models for these species that did not account for these spatial effects.

RESULTS

Lythrum salicaria cover around bird survey points averaged 0.62 6 0.03 SE (min and max: 0–1). Lythrum salicaria density around bird survey points averaged 0.22 touches/dm 6 0.03 SE (min and max: 0–1.06). For pairs of bird survey points within a wetland, the average difference in L. salicaria cover was 0.22 6 0.04 SE whereas the average difference in density was 0.23 touches/dm 6 0.04 SE. Comparisons of models relating avian species-specific abundances to Lythrum salicaria measures and other habitat variables indicated idiosyncratic responses by the species examined (Table 1). Nevertheless, the best models for four bird species included L. salicaria metrics. Two species responded positively, one negatively, and one with a condition 358 THE AMERICAN MIDLAND NATURALIST 168(2)

TABLE 2.—Simple slopes used to explore relationships between species abundances and interactions involving Lythrum salicaria density (Density) and cover (Cover) as well as water cover (Water); simple slopes are bivariate correlations between species abundances and habitat features. The interactions are labeled as ‘variable 1 | variable 2’ and should be interpreted as ‘the response (slope) of the species to variable 1 at each of 3 levels (Low, Mean, High) of variable 2’. Low and high levels of habitat metrics correspond to the mean 2 3 standard errors and the mean + 3 standard errors, respectively. Means and standard errors were 0.22 touches/dm 6 0.03 SE for L. salicaria density, 0.62 6 0.03 SE for L. salicaria cover, and 0.16 6 0.03 SE for water cover

Levels Species Interaction Low Mean High Virginia rail Water | Density 2.27 2.64 3.01 Density | Water 0.74 1.04 1.34 red-winged blackbird Density | Cover 21.44 22.34 23.23 Cover | Density 2.65 1.79 0.93 song sparrow Density | Cover 1.27 1.01 0.76 Cover | Density 20.74 20.99 21.23 dependent response (i.e., there was an interaction). Three species were insensitive to L. salicaria. Three out of four competitive models for Virginia rail included Lythrum salicaria effects, and the best model included a positive response to L. salicaria density. Based on Akaike weights, support for the best model was approximately equal to the support for the model with only non-L. salicaria variables. For marsh wren, a model that included a negative influence of L. salicaria density was 1.6 times more likely than a model with only non-L. salicaria habitat variables. Two of three models for song sparrow included L. salicaria effects. The best model which included an interaction between L. salicaria cover and density was approximately 2.4 times more likely than a model with only non-L. salicaria variables. The best model for swamp sparrow included a positive influence of L. salicaria cover, and there were no competitors for this model. For the red-winged blackbird, the best model included only non-L. salicaria variables, but a competitor included an interaction between L. salicaria cover and density. The model lacking L. salicaria effects was approximately 2.4 times more likely than the model including the interaction. The intercept-only model was best for the common yellowthroat and yellow warbler, so we did not report model results for these species. Interactions suggested a positive response by Virginia rails to water cover increased in strength as Lythrum salicaria density increased (Table 2). Interactions also suggested that, as water cover increased, a positive response of Virginia rails to L. salicaria density also increased (Table 2). For the red-winged blackbird, interactions suggested a negative response to L. salicaria density was stronger as L. salicaria cover increased. A positive response of red-winged blackbirds to L. salicaria cover became reduced as L. salicaria density increased. Interactions in song sparrow models indicated that a positive response to L. salicaria density decreased as L. salicaria cover increased. Interactions also suggested an increase in L. salicaria density led to an increasingly negative response to L. salicaria cover by song sparrows (Table 2).

DISCUSSION Alien, invasive species can negatively affect native plant and animal communities (Pysˇek and Richardson, 2010), although some native species may also benefit from the habitats 2012 TAVERNIA &REED:INVASIVE PURPLE LOOSESTRIFE AND WETLAND BIRDS 359 created by invasives (e.g., Heckscher, 2004). Based on species-specific habitat associations and previously published results, we predicted some negative and some positive relationships between breeding abundances for a suite of wetland bird species and invasive alien Lythrum salicaria cover and density. Our results provided evidence of L. salicaria effects for most species, but observed relationships did not always support predictions. Despite a lack of support for some species-specific predictions, our general expectation that both negative and positive responses would be observed was supported. Lavoie (2010) recently stated there was insufficient evidence to conclude that L. salicaria has largely negative effects on wetlands after reviewing studies addressing impacts on vascular plants, invertebrates, amphibians, birds, and ecosystem processes. Our results provided evidence that L. salicaria likely does not have a uniformly negative effect on wetland bird abundances. Based on potential changes in the structure and density of vegetation stands following Lythrum salicaria invasion, we predicted negative effects of L. salicaria cover and density on the abundances of marsh wrens and Virginia rails. Marsh wrens typically breed in cattail or bulrush marshes (Kroodsma and Verner, 1997), and these vegetation types have a simple vertical structure unlike the multi-stemmed and shrubby structure of L. salicaria. Previous researchers found that marsh wrens avoided placing their nests in L. salicaria (Rawinski and Malecki, 1984; Maddox and Wiedenmann, 2005) and that marsh wren abundance was lower in areas that were dominated by L. salicaria (Whitt et al., 1999; Hill, 2000). Consistent with these studies, our best supported model for marsh wrens included a negative effect of L. salicaria density. For the Virginia rail, a relatively ambulatory species, we predicted a negative relationship between abundance and L. salicaria density, but we also expected this effect to be ameliorated if dense L. salicaria stands were interspersed with open water pools. In contrast to our predictions, the best supported model for Virginia rail abundance suggested a positive influence of L. salicaria density. Consequently, our results suggested that, if there is a negative effect of L. salicaria density on Virginia rails, our L. salicaria stands were insufficiently dense. Our prediction of an interaction between L. salicaria density and water cover was partly supported. We found that a positive influence of water cover was greater as L. salicaria density increased, supporting the notion that open areas, such as water pools, were increasingly important to Virginia rails as L. salicaria density increased. For the marsh wren and Virginia rail, we also found some support for models without Lythrum salicaria measures, suggesting that L. salicaria does not affect abundances of these species in our study area. Indeed, this was true for each bird species that had L. salicaria measures incorporated into their best model. However, for these species, support for models lacking L. salicaria measures was low (Akaike weight #0.31), and the combined support for models including L. salicaria measures was high (Akaike weight $0.66). The relatively minor support for models lacking L. salicaria measures reinforces the importance of supported models including L. salicaria measures. Based on habitat affiliations, we anticipated positive responses by some birds to Lythrum salicaria cover and density; yellow warblers and common yellowthroats breed in shrubby areas (Guzy and Ritchison, 1999; Lowther et al., 1999), and song sparrows are typically associated with moist shrubby habitats (Arcese et al., 2002). In contrast to our predictions, we found no evidence that yellow warblers or common yellowthroats responded to L. salicaria cover or density. Our results supported those of Whitt et al. (1999) and Hill (2000) who also reported that these two species were insensitive to L. salicaria. Hill (2000) also found that song sparrows were insensitive to L. salicaria, but we observed a complex relationship between this species’ abundance and L. salicaria cover and density. Supported models included an interaction between L. salicaria cover and density and a negative 360 THE AMERICAN MIDLAND NATURALIST 168(2) relationship between abundance and L. salicaria cover. Taken together, these models suggested that song sparrows may benefit from dense stands of L. salicaria if the stands do not cover a large proportion of an area (i.e., high cover). Despite an early report that red-winged blackbirds clustered nests in stands of Lythrum salicaria even when stands of cattail were available for nesting (Rawinski and Malecki, 1984), subsequent studies have reported no relationship between red-winged blackbird abundance or density and L. salicaria cover or density (Whitt et al., 1999; Hill, 2000). Maddox and Wiedenmann (2005) found no difference in red-winged blackbird nesting success between L. salicaria and cattail stands, but they also noted that the phenologies of cattail and L. salicaria meant that nests were initiated later in L. salicaria stands. Our prediction that the abundance of red-winged blackbirds would be unrelated to L. salicaria cover and density was supported to the extent that the top model for red-winged blackbirds lacked L. salicaria measures. However, we also found evidence that red-winged blackbird abundance may be influenced by the interaction between L. salicaria cover and density. When L. salicaria density was low, there was a relatively strong, positive response of red-winged blackbirds to L. salicaria cover, and it is possible that these may have been the conditions under which Rawinski and Malecki (1984) observed clustering of red-winged blackbird nests in L. salicaria stands. However, the interaction term further indicated that red-winged blackbirds negatively respond to L. salicaria density regardless of cover level. We expected that swamp sparrows could be modeled using a non-linear (positive unimodal) relationship with Lythrum salicaria. Whitt et al. (1999) reported higher abundances of swamp sparrows at sites with L. salicaria as a dominant plant species. In contrast, Hill (2000) reported a trend that suggested a positive unimodal relationship between swamp sparrow densities and L. salicaria cover and density. Hill (2000) observed swamp sparrow use of L. salicaria as song posts and in nest building activities and suggested that L. salicaria may enhance territory quality when present at low levels. However, our results were consistent with Whitt et al. (1999) because we found support only for a positive, linear response of swamp sparrow abundance to L. salicaria cover. It is possible that our results differed from those of Hill (2000) because different L. salicaria metrics were used (i.e., cover and density were defined differently) or because our two studies examined a different range of L. salicaria conditions. Regardless, habitat relationships developed for swamp sparrows explained little variability, suggesting that unmeasured habitat features are important to this species on our study area. There has been considerable controversy surrounding the impacts of exotic invasive species (e.g., Brown and Sax, 2005), including the debate about the ecological effects of Lythrum salicaria (Anderson, 1995; Blossey et al., 2001; Lavoie, 2010). Evidence from our study and cited studies indicates that L. salicaria does not have a uniformly negative impact on wetland birds and that it can have a positive influence on some species. In a wetland management context, it has long been recognized that various bird species have different affinities for a variety of habitat features (e.g., Fredrickson and Taylor, 1982). From this perspective, the issue of L. salicaria’s presence and prevalence in wetlands does not appear to be a new or unique concern. Consequently, the management actions that should be taken to curb L. salicaria spread and reduce its prevalence are debatable (Anderson, 1995; Blossey et al., 2001; Lavoie, 2010), and probably will revolve around specific wetland management goals. Wetland managers would benefit from manipulative studies that evaluated the effect of L. salicaria control treatments (e.g., biocontrol) on wetland bird abundances; these studies could also be used to test predictions based on L. salicaria effects observed to-date. 2012 TAVERNIA &REED:INVASIVE PURPLE LOOSESTRIFE AND WETLAND BIRDS 361

Ultimately, the importance of the impacts of invasive species can only be fully understood by assessing effects on species persistence. Hence, future investigations of Lythrum salicaria’s influence on direct fitness measures, such as nesting success (e.g., Van Horne, 1983), and on the potential role of wetland networks in maintaining populations across a landscape would provide valuable insights into the conservation and management implications of L. salicaria invasion. The presence of complex and contrasting responses to L. salicaria invasion suggests that no single strategy is appropriate for managing L. salicaria within the context of bird species diversity.

Acknowledgments.—We thank D. DesRochers, S. Keyel, and M. Anteau for comments on an earlier version of this manuscript. S. Melvin provided field training. We thank N. Skaff for assistance in the field. C. Conway and C. Nadeau supplied sound recordings used in our playback surveys. C. Lattin assisted with audio software. F. Chew and C. Orians provided feedback on habitat sampling surveys. B. Paramenter and P. Florance provided assistance with GIS data gathering and D. Marshall provided advice on statistical analyses. T. Schneider Bayard provided additional statistical guidance. We also thank the numerous public and private landowners that provided access to their properties. Funding was provided by the A.V. Stout Fund of the Norcross Wildlife Foundation, the Charles Blake Fund of the Nuttall Ornithological Club, the Frances M. Peacock Scholarship of the Garden Club of America, a Tufts University’s Graduate Student Research Award and a fellowship from Tufts Institute of the Environment. Ideas for this paper were developed while doing related research supported by the Massachusetts Environmental Trust, Massachusetts Water Resources Research Center and Sigma Xi.

LITERATURE CITED

ANDERSON, M. G. 1995. Interactions between Lythrum salicaria and native organisms: a critical review. Environ. Manage., 19:225–231. ARCESE, P., M. K. SOGGE,A.B.MARR, AND M. A. PATTEN. 2002. Song sparrow (Melospiza melodia). In: A. Poole (ed.). The birds of North America online. Cornell Lab of Ornithology, Ithaca. Available via the birds of North America online: http://bna.birds.cornell.edu/bna/species/704. Accessed 20 Apr. 2010. BEALE, C. M., J. J. LENNON,J.M.YEARSLEY,M.J.BREWER, AND D. A. ELSTON. 2010. Regression analysis of spatial data. Ecol. Lett., 13:246–264. BLOSSEY, B., L. C. SKINNER, AND J. TAYLOR. 2001. Impact and management of purple loosestrife (Lythrum salicaria) in North America. Biodivers. Conserv., 10:1787–1807. BROWN,J.H.AND D. F. SAX. 2005. Biological invasions and scientific objectivity: reply to Cassey et al. (2005). Austral. Ecol., 30:481–483. BURNHAM,K.P.AND D. R. ANDERSON. 2002. Model selection and multimodel inference. Springer-Verlag, New York. 514 p. CARL, G., C. F. DORMANN, AND I. KU¨ HN. 2008. A wavelet-based method to remove spatial autocorrelation in the analysis of species distributional data. Web Ecol., 8:22–29. CONWAY, C. J. 1995. Virginia rail (Rallus limicola). In: A. Poole (ed.). The birds of North America online. Cornell Lab of Ornithology, Ithaca. Available via the birds of North America online: http:// bna.birds.cornell.edu/bna/species/173. Accessed 20 Apr. 2010. ——— AND J. P. GIBBS. 2005. Effectiveness of call-broadcast surveys for monitoring marsh birds. Auk, 122:26–35. DELUCA, W. V., C. E. STUDDS,L.L.ROCKWOOD, AND P. P. MARRA. 2004. Influence of land use on the integrity of marsh bird communities of Chesapeake Bay, U.S.A. Wetlands, 24:837–847. FARNSWORTH,E.J.AND D. R. ELLIS. 2001. Is purple loosestrife (Lythrum salicaria) an invasive threat to freshwater wetlands? Conflicting evidence from several ecological metrics. Wetlands, 21:199–209. FARNSWORTH, G. L., K. H. POLLOCK,J.D.NICHOLS,T.R.SIMONS,J.E.HINES, AND J. R. SAUER. 2002. A removal model for estimating detection probabilities from point-count surveys. Auk, 119:414–425. FOX, J. 2009. Car: companion to applied regression. R package version 1.2-16. 362 THE AMERICAN MIDLAND NATURALIST 168(2)

FREDRICKSON,L.H.AND T. S. TAYLOR. 1982. Management of seasonally flooded impoundments for wildlife. U.S. Department of the Interior and Fish and Wildlife Service, Washington, D.C. Resource Publication 148. 36 p. GUZY,M.J.AND G. RITCHISON. 1999. Common yellowthroat (Geothylpis trichas). In: A. Poole (ed.). The birds of North America online. Cornell Lab of Ornithology, Ithaca. Available via the birds of North America online: http://bna.birds.cornell.edu/bna/species/448. Accessed 20 Apr. 2010. HECKSCHER, C. M. 2004. Veery nest sites in a mid-Atlantic Piedmont forest: vegetative physiognomy and use of alien shrubs. Am. Midl. Nat., 151:326–337. HILL, J. D. 2000. Avian use of purple loosestrife (Lythrum salicaria) in southern Michigan wetland complexes. Master of Science Thesis, Michigan State University, East Lansing. 63 p. JACKMAN, S. 2009. Pscl: Classes and Methods for R Developed in the Political Science Computational Laboratory, Stanford University. Department of Political Science, Stanford University, Stanford, California. R package version 1.03.3. http://pscl.stanford.edu/ 2 KRAMER, M. 2005. R statistics for mixed models. P. Conf. Appl. Stat. Agr., 17:148–160. KROODSMA,D.E.AND J. VERNER. 1997. Marsh wren (Cistothorus palustris). In: A. Poole (ed.). The birds of North America online. Cornell Lab of Ornithology, Ithaca. Available via the birds of North America online: http://bna.birds.cornell.edu/bna/species/308. Accessed 20 Apr. 2010. LAVOIE, C. 2010. Should we care about purple loosestrife? The history of an invasive plant in North America. Biol. Invas., 12:1967–1999. LOWTHER, P. E., C. CELADA,N.K.KLEIN,C.C.RIMMER, AND D. A. SPECTOR. 1999. Yellow warbler (Dendroica petechia). In: A. Poole (ed.). The birds of North America online. Cornell Lab of Ornithology, Ithaca. Available via the birds of North America online: http://bna.birds.cornell.edu/bna/ species/454. Accessed 20 Apr. 2010. MADDOX,J.D.AND R. N. WIEDENMANN. 2005. Nesting of birds in wetlands containing purple loosestrife (Lythrum salicaria) and cattail (Typha spp.). Nat. Area. J., 25:369–373. PINHEIRO, J., D. BATES,S.DEBROY,D.SARKAR, THE RCORE TEAM. 2009. Nlme: linear and nonlinear mixed effects models. R package version 3.1-93. PYSˇEK,P.AND D. M. RICHARDSON. 2010. Invasive species, environmental change and management, and health. Annu. Rev. Env. Resour., 35:25–55. QUINN,G.P.AND M. J. KEOUGH. 2002. Experimental design and data analysis for biologists. Cambridge University Press, Cambridge. 556 p. RDEVELOPMENT CORE TEAM. 2009. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project.org. RAWINSKI,T.J.AND R. A. MALECKI. 1984. Ecological relationships among purple loosestrife, cattail and wildlife at the Montezuma National Wildlife Refuge. New York Fish Game J., 31:81–87. SCHMIDT,K.A.AND C. J. WHELAN. 1999. Effects of exotic Lonicera and Rhamnus on nest predation. Conserv. Biol., 13:1502–1506. SILLIMAN,B.R.AND M. D. BERTNESS. 2004. Shoreline development drives invasion of Phragmites australis and the loss of plant diversity on New England salt marshes. Conserv. Biol., 18:1424–1434. SORRIE,B.A.AND P. SOMERS. 1999. The vascular plants of Massachusetts: a county checklist. Massachusetts Natural Heritage and Endangered Species Program, Westborough. 186 p. STUCKEY, R. L. 1980. Distributional history of Lythrum salicaria (purple loosestrife) in North America. Bartonia, 47:3–20. TAVERNIA,B.G.AND J. M. REED. 2010. Spatial, temporal, and life history assumptions influence consistency of landscape effects on species distributions. Landscape Ecol., 25:1085–1097. THOMPSON, D. Q., R. L. STUCKEY, AND E. B. THOMPSON. 1987. Spread, impact and control of purple loosestrife in North American wetlands. U.S. Fish and Wildlife Service. 55 p. UNITED STATES FISH AND WILDLIFE SERVICE. 1995. The Silvio O. Conte National Fish and Wildlife Refuge Final Action Plan and Environmental Impact Statement. Northeast Region, National Wildlife Refuge System, Hadley, Massachusetts. VAN HORNE, B. 1983. Density as a misleading indicator of habitat quality. J. Wildl. Manage., 47:893–901. WELLER, M. W. 1999. Wetland birds: habitat resources and conservation implications. Cambridge University Press, Cambridge. 316 p. 2012 TAVERNIA &REED:INVASIVE PURPLE LOOSESTRIFE AND WETLAND BIRDS 363

WHITE, G. C. 1983. Numerical estimation of survival rates from band-recovery and biotelemetry data. J. Wildl. Manage., 47:716–728. WHITT, M. B., H. H. PRINCE, AND R. R. COX Jr. 1999. Avian use of purple loosestrife dominated habitat relative to other vegetation types in a Lake Huron wetland complex. Wilson Bull., 111:105–114. WILSON, L. M., M. SCHWARZLAENDER,B.BLOSSEY, AND C. BELL RANDALL. 2004. Biology and biological control of purple loosestrife. Forest Health Technology Enterprise Team, Morgantown, West Virginia. 78 p. YASUKAWA,K.AND W. A. SEARCY. 1995. Red-winged blackbird (Agelaius phoeniceus). In: A. Poole (ed.). The birds of North America online. Cornell Lab of Ornithology, Ithaca. Available via the birds of North America online: http://bna.birds.cornell.edu/bna/species/184. Accessed 20 Apr. 2010. ZEDLER,J.B.AND S. KERCHER. 2004. Causes and consequences of invasive plants in wetlands: opportunities, opportunists, and outcomes. Crit. Rev. Plant Sci., 23:431–452. ZUUR, A. F., E. N. IENO, AND C. S. ELPHICK. 2009a. A protocol for data exploration to avoid common statistical problems. Methods Ecol. Evol., 1:3–14. ———, ———, N. J. WALKER,A.A.SAVELIEV, AND G. M. SMITH. 2009b. Mixed effects models and extensions in ecology with R. Springer Science + Business Media, New York. 596 p.

SUBMITTED 21 JULY 2011 ACCEPTED 10 FEBRUARY 2012