Advertisement

BioControl

, Volume 63, Issue 3, pp 449–460 | Cite as

Analyses of nine years of citizen-based biological control monitoring of Dalmatian toadflax, Linaria dalmatica (Plantaginaceae) in Idaho, USA

  • Aaron S. Weed
  • Joseph Milan
  • Mark Schwarzlaender
Article

Abstract

Dalmatian toadflax has been a target for biological control in North America since the 1960s. The stem-mining weevil Mecinus janthiniformis was first released in Canada and the western United States in the mid-1990s. Since 2007, a citizen-based monitoring program in Idaho, USA has supplemented data collection to help evaluate the impact of M. janthiniformis on Dalmatian toadflax abundance and assess changes in the surrounding plant community. We monitored and analysed trends in toadflax, weevil, and the plant community abundance following weevil releases at the regional and site level (34 sites) across the state of Idaho, USA. Significant declines in toadflax cover and stem density were recorded across the majority of sites. Weevil populations have established at all release sites. The mechanistic model indicated that the population dynamics of toadflax at our sites are negatively affected by M. janthiniformis abundance. When averaged across the region, 15 years after weevil release, Dalmatian toadflax stem density and cover declined by 93 and 84%, respectively. We observed significant declines in toadflax abundance in over 75% of the sites. Changes to the surrounding plant community following weevil releases were less consistent among sites. At the regional scale we found evidence for an overall increase in average cover of native perennial grasses and other exotic weeds (primarily annual grasses and exotic forbs) but a decline in native forbs.

Keywords

Post-release assessment Weed population dynamics Stem-miner Curculionidae 

Introduction

The invasive weed Dalmatian toadflax (Linaria dalmatica ssp. dalmatica (L.) Mill.; syn.: L. dalmatica (L.) Mill. and L. genistifolia (L.) Mill. ssp. dalmatica (L.) Maire & Petitm.) (Plantaginaceae) was introduced into North America from Europe near the turn of the 20th century (Alex 1962) and is a noxious weed of natural areas and dry rangelands in western North America (USDA-NRCS 2017). Dalmatian toadflax is a short-lived, herbaceous perennial that primarily invades well-drained, coarse-textured soils and is commonly associated with disturbed, sparsely vegetated habitats (Robocker 1974; Vujnovic and Wein 1997; Blumenthal et al. 2012). Individual plants produce multiple prostrate and erect, flowering stems from mature, woody rootstocks and each plant is estimated to produce up to 500,000 seeds that can remain viable for up to ten years in the soil (Vujnovic and Wein 1997). Dalmatian toadflax also spreads through horizontal roots.

Dalmatian toadflax has been a target for biological control in North America since the 1960s (De Clerck-Floate and Harris 2002; Wilson et al. 2005). Three plant-feeding beetle species arrived inadvertently, Brachypterolus pulicarius L. (Kateretidae) and two Rhinusa [Gymnetron] species (Curculionidae). Two other species of herbivores, Lepidoptera, (Calophasia lunula (Hufnagel) (Noctuidae) and Eteobalea species (Cosmopterigidae), and the weevil, Rhinusa linariae (Panzer), were intentionally released to reduce reproduction and population density of various toadflax weeds in North America. All of the deliberately introduced biocontrol agents except for R. linariae and Eteobalea species are well established in North America, but most have little population-level impact on the target weed (Wilson et al. 2005; Sing et al. 2016). Much later, in the early 1990s, a stem-mining weevil Mecinus janthinus Germar was released in North America in another attempt to reduce toadflax abundance (De Clerck-Floate and Harris 2002; Nowierski 2004). Morphological and molecular studies have subsequently determined that Mecinus janthiniformis Toševski & Caldara attacks Dalmatian toadflax, while the near-cryptic species M. janthinus attacks yellow toadflax, Linaria vulgaris Mill (Toševski et al. 2011).

Mecinus janthiniformis is currently established throughout Canada and the western United States (Wilson et al. 2005; Toševski et al. 2011; Sing et al. 2016). Releases of M. janthiniformis were made in Canada in 1991 and in the United States in 1995 (De Clerck-Floate and Harris 2002; Nowierski 2004). Adults emerge in early spring, undergo maturation feeding on leaves, and oviposit into stems. Larvae emerge from eggs after about seven days and create mines within the vascular tissue of the stems (Jeanneret and Schroeder 1992). Pupation occurs during late summer and adults remain in the stems to overwinter. One generation is completed per year.

Declines in stem and patch density in western USA and Canada have been associated with the release of M. janthiniformis (De Clerck-Floate and Miller 2002; Sing et al. 2008; van Hezewijk et al. 2010; Weed and Schwarzländer 2014), whose adult and larval feeding can reduce shoot height and decrease reproduction of Dalmatian toadflax (Schat et al. 2011; Goulet et al. 2013). In 2007, the Idaho Biological Control Task Force in partnership with the University of Idaho, Nez Perce Biocontrol Center, Idaho State Department of Agriculture, U.S. Forest Service and U.S. Bureau of Land Management, initiated a state-wide effort to monitor changes of Dalmatian toadflax and the surrounding plant community following releases of M. janthiniformis. Prior analysis of the monitoring data from 2007 to 2011 demonstrated that variation in stem density of Dalmatian toadflax was linked to negative feedback processes related to abundance in the year prior (e.g., probably intraspecific competition), variation in precipitation, and weevil abundance (Weed and Schwarzländer 2014). These results concluded that reductions in Dalmatian toadflax stem density were associated with increasing weevil abundance. In this study, we expand upon our earlier work by assessing regional and site-level trends in toadflax, weevil, and plant community abundance following weevil releases from a longer monitoring time period (2007–2015). We then fit mechanistic models to estimate the timing and influence of biotic factors on the population dynamics of Dalmatian toadflax and M. janthiniformis. These analyses were undertaken to determine whether our interpretation of Dalmatian toadflax biocontrol impact has changed as we continue to monitor sites for longer periods and because analyses of M. janthiniformis time series have not been conducted.

Methods

Long-term monitoring of Dalmatian toadflax biological control

Only a brief description of the monitoring program is provided here and we refer readers to Weed and Schwarzländer (2014) for a full description of the monitoring program, site selection, and data collection procedures. The monitoring protocol was developed with the goal to encourage participation of land managers, stakeholders and citizen-scientists. Sites for permanent monitoring were selected following a request for study sites to land managers and county weed personnel throughout the state of Idaho. Most study sites were selected based on accessibility and availability of local participants to monitor sites. The decision to select sites in a non-random fashion was weighed against multiple trade-offs including the logistical constraints of sampling many sites across a wide geographic region within a short time at locales often in rugged terrain and to encourage citizen participation. Our inquiries lead to the establishment of permanent sampling sites at 38 Dalmatian toadflax infestations by 2011. Sites were located in infestations of Dalmatian toadflax ranging from relatively small, isolated patches of plants near road sides to large, contiguous patches. Habitat types of the sites ranged from high elevation coniferous forests to areas previously or recently disturbed by grazing and fire, and river canyon grasslands. Some sites were located in actively managed rangelands with cattle and most sites were under active integrated management targeting multiple species of noxious weeds with herbicides and biological control (e.g., yellow star thistle, spotted knapweed, etc.).

In this study, we focused the analysis on 34 of the 38 sites where we have documented dates of weevil releases, where 150–600 weevils (221 ± 17, mean ± SE) were released per site from 2000 to 2009. Variation in the number of weevils released per site was usually due to availability of weevils to land managers. Vegetation monitoring was conducted at each site along a permanent 20 m transect placed randomly within a Dalmatian toadflax infestation. Ten permanent 0.25 × 0.5 m (0.125 m2) sampling plots were marked every 2 m along each transect. In each year, the density of live Dalmatian toadflax stems was counted in each plot. Plant community composition in each plot was characterized by visually estimating the percentage cover of Dalmatian toadflax, native perennial grasses, native forbs, other exotic invasives (forbs and annual grasses), litter, moss, and bare ground. Monitoring personnel were trained in the identification of common plant species of each broad vegetation cover class and workshops were conducted annually to improve data quality. The abundance of M. janthiniformis was estimated at each site in each year using six, 3 min counts of adults (Carney 2003). During each count the observer walked systematically through the infestation surrounding each transect carefully counting every visible weevil within 3 min. In 2010 and 2011, stems were collected and dissected from 11 sites (9–13 stems collected per site each year; 182 total stems) to assess adult oviposition in addition to taking adult counts. These dissections confirmed that 3 min adult counts are highly correlated with attacks per cm of stem (i.e., larval population size) (Pearson correlation coefficient = 0.61) suggesting that adult counts provide a reasonable index for assessing weevil impact (unpublished data). Vegetation and weevil sampling were conducted on the same day at each site. The monitoring was timed to coincide with peak weevil abundance at each site in that year. The most appropriate monitoring times during each year were based upon prior experience with monitoring M. janthiniformis at the site. All data were submitted annually to the US Bureau of Land Management and managed in a GIS database.

Statistical analysis

Prior to all analyses we calculated the average Dalmatian toadflax stem density and cover, average cover of all plant cover types, and average weevil abundance at each site in each year. These site-level averages were used in the analyses described below. We first assessed trends in the plant community and M. janthiniformis abundance at sites with known weevil release dates (34 sites). To evaluate trends in these variables across all sites (regional scale) a mixed effects model was fit to the annual site average of Dalmatian toadflax (stem density and cover), M. janthiniformis counts, or the associated plant community cover. The main effect in the model was years since the release of M. janthiniformis and site set was set as the random effect. To estimate the trend in the same response variables at each site a linear model was fit to the data at each site as a function of the number of years since the release of M. janthiniformis. All model residuals were checked for normality. From these models, we estimated the trend (β 1) at the regional scale and site-level. The β 1 value estimates the direction and magnitude of the trend in each response variable following weevil release.

Mechanistic models were then used to evaluate the timing and influence of biotic factors on Dalmatian toadflax population dynamics. Prior to analysis we removed the linear temporal trend from each site-level time series of stem density to avoid estimating spurious correlations that can arise when trends are present in time series. Interannual changes in stem density (\(R_{t}^{N}\)) were then calculated from the series with the temporal trends removed as \(R_{t}^{N}\) = ln(N ti+1 ) − ln(N ti ), where N ti was the (detrended) average Dalmatian toadflax density (0.125 m−2) at site i in current year t and N ti+1 the average density in the previous year. Similar to Weed and Schwarzländer (2014), a discrete form of a predator–prey model was used as the baseline for evaluating and comparing alternative models to describe the temporal behavior in interannual changes in Dalmatian toadflax stem density
$$R_{t}^{N} = \, f({ \ln }(N_{t}),{ \ln }(N_{t - 1}),{ \ln }(P_{t}),{ \ln }(P_{t - 1}),\varepsilon_{t} )$$
(1)
with N t and N t−1 equal to the previous and lagged year toadflax stem density, P t and P t−1 the average abundance of M. janthiniformis adults in the previous and lagged year, respectively, and \(\varepsilon_{t}\) representing sampling error in density estimates plus density-independent effects on toadflax population dynamics (Berryman 2003).
Using the same monitoring dataset but with observations only from 2007 to 2011, Weed and Schwarzländer (2014) found that stem density in the previous year (N t ) was the strongest source of negative feedback on interannual changes in toadflax stem density at the regional scale. This negative feedback is due to density-dependent processes, such as competition for resources or from natural enemies and affects variation in abundance (Berryman 2003). We conducted a similar analysis with the updated dataset (through 2015) and still found support for direct (N t ) rather than delayed (N t−1) effects on interannual changes in stem density. Based on this, we evaluated model structures including the effects of prior stem density (N t ) and M. janthiniformis (P t−l) abundance and the null model (intercept only) for their ability to describe interannual changes in Dalmatian toadflax stem density (\(R_{t}^{N}\)) at the regional-scale (Table 1). While we recognize and have demonstrated the importance of other factors on Dalmatian toadflax demography (Weed and Schwarzländer 2014) in this study we were most interested in modelling the density-dependence and not the influence of density-independent variation (i.e., weather or plant community context). The current approach differs from the one described in Weed and Schwarzländer (2014) because models were fit in this study to examine different density-dependent processes and with a longer time series of data. Parameters and model fit (second-order Akaike Information Criterion (AICc)) of the five model structures (Table 1) were estimated using a mixed effects model structure including site as the random effect in the nlme package in R (Pinheiro et al. 2011).
Table 1

Maximum likelihood estimates (± SE) of the factors affecting interannual changes in Dalmatian toadflax stem density (\(R_{t}^{N}\)) among different model structures

Model

1

2

3

4

5

∆AICc

0

0.19

31.2

64.0

102.0

Model weight

0.52

0.48

0

0

0

Intercept

0.92 ± 0.14

1.06 ± 0.11

0.89 ± 0.10

0.32 ± 0.09

0.03 ± 0.05

lnN t

− 0.62 ± 0.13

− 0.80 ± 0.07

− 0.88 ± 0.07

  

lnP t

− 0.05 ± 0.04

− 0.10 ± 0.03

 

− 0.11 ± 0.03

 

lnN t × lnP t

− 0.06 ± 0.04

    

N t estimates the influence of average toadflax stem density in the previous year and P t estimates the effect of average abundance of M. janthiniformis adults in the previous year

Lastly, we used a modelling approach similar to the one described above with stem density to assess the factors and timing of their effects on M. janthiniformis population dynamics. We calculated interannual changes in weevil abundance from detrended tome series at each site as \(R_{t}^{P}\) = ln(P ti+1 ) − ln(P ti ). We assumed that interannual changes in weevil abundance are also affected by previous (t) or lagged (t−1) toadflax and weevil abundance denoted by the model
$$R^{P}_{t} = f({ \ln }\left( {N_{t} } \right),{ \ln }\left( {P_{t} } \right),{ \ln }\left( {P_{t - 1} } \right),\varepsilon_{t} )$$
(2)
Preliminary analyses determined that effects of delayed (N t−1) toadflax stem density were never important in describing \(R_{t}^{P}\). Our final model selection approach compared the fit of all the possible candidate models (19 total models) nested within Eq. (2). The fit of each model in the set was then evaluated using AICc with the best fit indicated by the lowest AICc. Models are considered plausible if they have a ΔAICc below 2 (Burnham and Anderson 2002). The multi-model inference analyses were performed using the ‘MuMIn’ package implemented in R version 3.3.1 (Barton 2012).

Results

Trends in Dalmatian toadflax, M. janthiniformis, and the associated plant community

The Dalmatian toadflax biocontrol monitoring program has accumulated data from 34 sites where we have known release dates of M. janthiniformis. Monitoring at six of these sites began the year of the weevil release (since 2007) and four sites were monitored following releases 15 years ago (Fig. 1 and supplementary material Fig. 1A).
Fig. 1

Number of sites monitored (top) and associated trends in mean (± SE) weevil abundance (middle) and Dalmatian toadflx stem density (bottom) among 34 sites following releases of M. janthiniformis in Idaho, USA

Weevils established at all 34 sites and, since release, populations have remained relatively constant when averaged across all of the sites with no indication of a statistical trend since the releases (F 1,207 = 1.34; β 1 = 0.04 ± 0.03; P = 0.249) (Fig. 1). When viewed at the site-level, significant trends (P < 0.05) in weevil abundance since releases were lacking at most sites (24 or 65%). Significant increases (P < 0.05) in weevil abundance were observed at six (22%) sites whereas decreases in weevil abundance were detected at four sites (15%) (Table 2). The sites with a statistically significant trend in weevil abundance were monitored for at least eight to nine consecutive years, except Slate Creek (five years) (Table 2).
Table 2

Model coefficients (± SE) and associated P-values at sites where significant trends in Mecinus janthiniformis abundance (weevils 3-min−1) were observed as a function of years after its release

Site name

Intercept

Years since release

Model R 2

No. years monitored

β 0

P

β 1

P

Big Wood

1.27 ± 0.45

0.027

0.19 ± 0.07

0.029

0.45

9

Caribou

− 0.80 ± 0.59

0.220

0.38 ± 0.08

0.003

0.75

8

Cold Springs

0.56 ± 0.9

0.560

0.46 ± 0.18

0.043

0.44

8

Doty

0.38 ± 0.65

0.581

0.22 ± 0.06

0.007

0.63

9

Idaho Hill Transfer

4.14 ± 0.4

< 0.001

− 0.19 ± 0.1

0.089

0.31

8

Lake Creek

1.93 ± 0.53

0.008

0.18 ± 0.08

0.060

0.33

9

Port Hill

3.65 ± 0.24

< 0.001

0.15 ± 0.04

0.004

0.67

9

Slate Creek

15.81 ± 3.94

0.028

− 1.21 ± 0.39

0.054

0.68

5

Snake River Breaks

3.19 ± 0.86

0.007

− 0.33 ± 0.13

0.041

0.40

9

Ten Ax

4.49 ± 1.4

0.015

− 0.31 ± 0.12

0.041

0.39

9

When averaged across the 34 sites, Dalmatian toadflax stem density has declined by about 93% (F 1,207 = 107.9; β 1 = − 0.19 ± 0.02; P < 0.001) 15 years after weevil release (Fig. 1). Over this same period a 84% reduction in Dalmatian toadflax cover was reported (F 1,207 = 87.6; β 1 = − 0.22 ± 0.02; P < 0.001). Site-level stem density declined at 22 sites since weevil release (Table 3). Lake Creek, which has been monitored for nine years, was the only site where an increase in stem density was observed over time (Table 3). Five of the seven sites lacking a significant trend have only been monitored for five years so it may still be too early to distinguish whether toadflax populations are changing at these sites or not.
Table 3

Model coefficients (± SE) and associated P-values estimating the trend in Dalmatian toadflax stem density as a function of years after release of Mecinus janthiniformis

Site name

Intercept

Years since release

Model R 2

No. years monitored

β 0

P

β 1

P

Baldy Palisades

3.58 ± 0.35

< 0.001

− 0.32 ± 0.05

< 0.001

0.85

9

Battle Creek

1.50 ± 0.29

0.015

0 ± 0.06

0.993

− 0.33

5

Bear Lake

2.62 ± 1.05

0.056

− 0.23 ± 0.16

0.205

0.16

7

Big Wood

3.29 ± 0.23

< 0.001

− 0.26 ± 0.04

< 0.001

0.87

9

Blackhawk Bar

5.31 ± 1.88

0.066

− 0.58 ± 0.26

0.114

0.49

5

Caribou

4.26 ± 0.71

0.001

− 0.35 ± 0.1

0.012

0.62

8

Center Canyon 01

3.88 ± 0.85

0.020

− 0.62 ± 0.2

0.054

0.68

5

Center Canyon 02

2.81 ± 0.39

0.005

− 0.49 ± 0.09

0.013

0.88

5

Center Canyon 03

1.83 ± 0.43

0.023

− 0.35 ± 0.1

0.041

0.73

5

Cold Springs

2.24 ± 0.34

0.001

− 0.26 ± 0.07

0.008

0.66

8

Cove Ranch

2.58 ± 0.16

< 0.001

− 0.15 ± 0.05

0.021

0.62

7

Cuprum

1.97 ± 0.21

< 0.001

− 0.22 ± 0.04

0.004

0.81

7

Doty

2.73 ± 0.57

0.002

− 0.11 ± 0.05

0.060

0.34

9

Hammer Creek

4.08 ± 2.02

0.137

− 0.71 ± 0.48

0.233

0.23

5

Hayden

2.13 ± 0.73

0.022

− 0.23 ± 0.11

0.078

0.29

9

Helispot

1.78 ± 0.42

0.004

− 0.11 ± 0.04

0.020

0.50

9

Huck

3.22 ± 0.59

0.001

− 0.23 ± 0.05

0.003

0.70

9

Idaho Border

1.41 ± 0.17

< 0.001

− 0.21 ± 0.04

0.002

0.79

8

Idaho Hill Transfer

2.04 ± 0.48

0.005

− 0.28 ± 0.11

0.046

0.43

8

Lake Creek

1.23 ± 0.13

< 0.001

0.09 ± 0.02

0.003

0.71

9

Little Salmon River 01

1.19 ± 0.13

0.003

− 0.23 ± 0.05

0.022

0.82

5

Little Salmon River 02

1.42 ± 0.29

0.016

− 0.42 ± 0.12

0.036

0.75

5

Lone Mountain

1.57 ± 0.29

0.001

− 0.19 ± 0.04

0.004

0.69

9

Lyons Bar 01

3.03 ± 0.44

0.006

− 0.26 ± 0.06

0.026

0.80

5

Patterson

4.86 ± 2.53

0.151

− 0.54 ± 0.31

0.182

0.33

5

Pine Bar

2.34 ± 0.22

< 0.001

− 0.18 ± 0.03

< 0.001

0.82

9

Pine Creek

1.83 ± 0.29

0.001

− 0.15 ± 0.03

0.004

0.73

8

Port Hill

1.20 ± 0.37

0.015

0.01 ± 0.06

0.929

− 0.14

9

Rice Creek

2.25 ± 0.74

0.023

− 0.23 ± 0.11

0.077

0.34

8

Slate Creek

4.29 ± 0.92

0.018

− 0.37 ± 0.09

0.026

0.80

5

Snake River Breaks

1.41 ± 0.31

0.003

− 0.18 ± 0.05

0.008

0.61

9

Squaw Creek 01

1.36 ± 0.08

< 0.001

− 0.33 ± 0.03

0.002

0.96

5

Ten Ax

4.25 ± 0.64

< 0.001

− 0.32 ± 0.06

0.001

0.79

9

Whitebird

0.19 ± 0.77

0.816

0.08 ± 0.14

0.609

− 0.20

5

Site-level trends in Dalmatian toadflax cover (Supplementary Material Table A1) generally followed a similar pattern as stem density (decreasing trend over time) due to their high correlation but with a few exceptions (Fig. 2). Significant trends in stem density observed at Cove Ranch, Doty, and Helispot were not apparent for toadflax cover and, although the decline in stem density was only marginally significant at Blackhawk Bar (Table 3), the decrease in toadflax cover was highly significant at this site (P = 0.04; supplementary material Table A1).
Fig. 2

Distribution of the regession coefficients estimating the temporal trend (β 1) of plant community cover types following releases of M. janthiniformis among 34 sites in Idaho, USA

Statistically significant trends in the surrounding plant community also coincided with the period after the release of M. janthiniformis. For instance, across all sites, there was a 12% increase in average native, perennial grass cover (F 1,207 = 3.97; β 1 = 0.03 ± 0.02; P < 0.048) and a slight increase in the cover of other exotic weeds (F 1,207 = 8.60; β 1 = 0.06 ± 0.02; P = 0.004) 0–15 years after weevil release. During this same period mean native forb cover remained unchanged at the regional scale (F 1,207 = 2.01; β 1 = − 0.03 ± 0.02; P = 0.157).

Site-level trends in the surrounding plant community cover were much more variable among sites compared to trends in toadflax cover (supplementary material Tables A2–A4). Native forb cover increased at two sites (0.06%), decreased at six sites (17.6%), while at 28 sites (76.4%) there was no change observed following weevil release (supplementary material Table A2). Significant increases in forb cover were recorded at two sites (Helispot and Doty). Increases in native, perennial grass cover were observed at five (14.7%) sites but declines in native, perennial grass cover were observed at four sites (11.7%) (Fig. 2, supplementary material Table A4). Finally, other exotic weed cover had increased at 12 (80%) sites, and this is the general pattern observed across most sites currently monitored (Fig. 2). Significant declines in other exotic weed cover coincident with weevil releases were found at Lone Mountain, Whitebird, and Pine Bar (supplementary material Table A4).

Factors affecting population dynamics of Dalmatian toadflax

Of the five model structures examined, interannual changes in Dalmatian toadflax at the regional scale were best described by model structures including effects from previous stem density (N t ) and prior weevil (P t ) abundance (Table 1). In other words, changes in toadflax density could not be explained simply by stem density or weevil abundance alone. However, model structures with additive and interactive effects were almost indistinguishable in terms of their information content (i.e. ΔAIC = 0.3; Burnham and Anderson 2002) but there were fewer parameters to estimate for the additive model. Estimated coefficients from the top model (1) indicated that regional population dynamics of Dalmatian toadflax (34 sites) were negatively affected by prior stem density (F 1,168 = 171.7; P < 0.001) and M. janthiniformis adult abundance (F 1,168 = 13.0; P < 0.001) (Table 1).

Factors affecting population dynamics of Mecinus janthiniformis

The multi-model inference approach determined that three model structures provided a similar level of information content (ΔAICc < 2) in explaining interannual changes in weevil abundance (Table 4). Although the ΔAICc of model 3 equalled 2.1, we considered this model structure to be weighted high enough for consideration. The top model (lowest AICc) indicated that stem density in the previous year (N t ) positively affected (F 1,133 = 3.4; P = 0.070) whereas prior weevil abundance negatively affected interannual changes in weevil abundance (F 1,133 = 67.4; P < 0.001) (Table 4; Fig. 3). Or, weevil populations tend to increase following years with higher than average stem density. Conversely, weevil populations tend to decrease following years when weevil populations are relatively high (assuming similar stem density). The negative effect of P t is evidence of direct, negative feedback. The next best model structure (2) suggested a weak negative effect due to delayed weevil abundance (P t−1) on interannual changes in weevil abundance (Table 4). Finally, the third best model suggested an interaction between prior stem density (N t ) and weevil abundance (P t ) (Table 4).
Table 4

Maximum likelihood estimates (± SE) of the factors affecting interannual changes in Mecinus janthiniformis adult abundance (\(R_{t}^{P}\)) among the top (∆AICc ~ 2) and null models

Model

1

2

3

4

∆AICc

0

2.0

2.1

41.6

Model weight

0.43

0.16

0.15

0

Intercept

1.09 ± 0.28

1.17 ± 0.32

1.26 ± 0.40

− 0.12 ± 0.11

lnN t

0.54 ± 0.17

− 0.65 ± 0.09

0.43 ± 0.37

 

lnP t

− 0.65 ± 0.08

− 0.53 ± 0.17

− 0.68 ± 0.12

 

lnP t1

 

− 0.03 ± 0.08

− 0.04 ± 0.08

 

lnN t × lnP t

  

0.04 ± 0.12

 

N t estimates the influence of average toadflax stem density in the previous year and P t−l estimates the effect of average abundance of M. janthiniformis adults in the previous year (t) and lagged year (t−1)

Fig. 3

Conditional plots showing the effect of prior M. janthiniformis abundance, ln(P t), and Dalmatian toadflax stem density, ln(N t), on interannual changes in adult M. janthiniformis abundance (\({\text{R}}_{\text{t}}^{\text{P}}\)) across all monitoring sites. Shaded area denotes 95% confidence interval with respect to the regression models

Discussion

Data from our monitoring suggests that releases successfully established M. janthiniformis at our sites and that most of the sites have stable weevil populations. Our results also indicate that average stem density and cover of Dalmatian toadflax across the entire monitoring program’s sites has declined in concert with weevil releases. This pattern of decline in Dalmatian toadflax stem density and cover coincident with time since weevil release is apparent at almost all the 34 sites examined, with 76% of sites showing evidence of statistically significant reductions in toadflax. Changes to the surrounding plant community cover following weevil releases were less consistent among cover types and sites than toadflax abundance. At the regional scale, we found evidence for an overall increase in average cover of native perennial grasses and other exotic weeds (primarily annual grasses and exotic forbs) but a decline in native forbs following weevil releases. Although biological control has contributed to the notable decline in toadflax abundance across our monitoring sites, our results suggest it has not influenced the recovery of native plant communities at the regional scale. This may suggest that Dalmatian toadflax abundance has a weak effect on native plant species presence in general and community restoration will require additional management tactics.

Results from the mechanistic model evaluating the timing of effects from prior stem density and weevil abundance on interannual changes in toadflax stem density (or more generally toadflax population dynamics) are still consistent with results reported from an analysis of shorter time series (2007–2011) (Weed and Schwarzländer 2014). The current analysis of a longer time series indicated that density-dependent negative feedback from stem density in the prior year is still apparent as is the negative effect associated with M. janthiniformis abundance. Importantly, toadflax population dynamics could not be explained simply by effects from previous stem density, but the models consistently included effects from weevil abundance. Our interpretation of the analysis remains the same as in the 2014 study—effects of prior stem density likely reflect the influence of intraspecific competition (Buckley et al. 2001) and the negative effect associated with weevil abundance likely indicates the impact of this weevil on plant growth that has been documented in laboratory and in other field experiments (Jeanneret and Schroeder 1992; De Clerck-Floate and Miller 2002; Peterson et al. 2005; Sing et al. 2008; van Hezewijk et al. 2010; Jamieson et al. 2011).

The results of the mechanistic model describing toadflax dynamics are important for interpreting the importance of biocontrol on toadflax demography. Since most monitoring sites were selected and releases began when toadflax was relatively abundant, it could be argued that negative feedback processes driven by plant-plant or plant-soil interactions rather than biocontrol herbivory was responsible for the reported reductions in toadflax stem density at the site and regional scale (Crawley 1989). However, results from the population model clearly indicate that changes in stem density are related to the simultaneous influence of both factors, each having a negative effect of toadflax population dynamics. Our current data supports the interpretation that weevil abundance has played a role in the regional declines in stem density. In our previous study we also demonstrated the importance of variation in precipitation (Weed and Schwarzländer 2014). Although we did not test this in this study, it is still expected that inputs of precipitation at these drylands sites will affect variation in toadflax abundance and possibly the impact of herbivory on toadflax demography.

Interannual changes in M. janthiniformis abundance across our monitoring sites were positively affected by prior stem density and negatively affected by prior weevil abundance. While the positive influence of host resources on the abundance of herbivores is widely acknowledged (Price et al. 2011) the finding of direct negative feedback affecting weed biocontrol agent population dynamics is expected from studies of other herbivorous insects but still needs further examination. Direct, negative feedback, such as suggested from prior weevil abundance in this study, is an important stabilizing force (i.e. tendency to return to equilibrium abundance when perturbed) on population fluctuations over time (Berryman 2003). Our finding of direct, negative feedback is strong evidence that it, and perhaps dispersal, plays a role in the observed stability in weevil populations among our sites. The likely cause of negative feedback in M. janthiniformis population dynamics is probably due to intense intraspecific larval competition similar to that documented in the stem miner Lixus cardui Olivier attacking Onopordum (Briese et al. 2004). Briese et al. (2004) demonstrated that population growth of L. cardui is negatively affected by weevil density and plant size due to their influence on larval competition. Numerous lines of evidence suggest similar processes affect M. janthiniformis as well. For example, experiments have documented a positive correlation between oviposition per stem and local weevil abundance in the field (Carney 2003; van Hezewijk et al. 2010) and we have shown decreases in per capita reproduction and weevil mass in the F1 generation with increasing attacking parent weevil density (ASW, unpublished data).

The observation that weevil populations have remained relatively constant across all of our sites (regional sclae) while we observed widespread declines in Dalmatian toadflax is noteworthy since weevil abundance is expected to track weed abundance closely. The best explanation for this regional-scale trend in weevil abundance is most likely due to the variation in weevil density among sites. And, while weevil density tends to be positively correlated with weed density at most of the monitored sites (see supplemental materials Fig. A1), weevil population estimates (based on our 3 min counts) have either remained constant or have increased over time at sites where toadflax populations have declined. This discrepancy may be related to counting weevils nearby versus on our monitoring transects and deserves further inquiry. Although we have observed a strong correlation between our adult counts and attack density near transects there may be other site-specific factors influencing weevil estimates that we haven’t accounted for (e.g., size and isolation of infestation). It is also important to note that based on our current time series, toadflax and weevil populations on average do not appear to undergo large boom-bust cycles (unstable behavior). The population time series analyzed so far are relatively stable (e.g., small year-to-year variation in fluctuations around the mean, even though they may be declining). It is possible that weevil immigration from surrounding areas into our sites helps to prevent weevil populations from going locally extinct and contributes to the observed low variation in weed abundance over time. These interpretations may change as we monitor sites for longer periods of time.

In conclusion, declines in Dalmatian toadflax at our sites in Idaho have been widespread since the introduction of M. janthiniformis. Our population modeling approach strongly indicates that the observed significant declines in toadflax abundance are partly due to the impact from M. janthiniformis. Despite the consistency in the reduction of toadflax abundance observed among sites, there was a high amount of site-level variation in the responses of the surrounding plant community following weevil releases. In general, however, the monitoring program is documenting regional increases in native grass cover but also cover of other exotic weeds. Since the native plant community has shown a fairly weak response to reductions in Dalmatian toadflax additional management (restoration and integrated management of co-occurring weeds) is likely needed.

Notes

Acknowledgements

We thank the monitoring personnel, Carol Randall, and Paul Brusven for their instrumental role in data collection and anonymous reviewers for the valuable suggestions to the manuscript. We also would like to thank Richard Reardon for his continued support of our research. This research was funded by USDA USFS Special Technology Development Project (STDP) Grant 10-CA-11010000-013 to M.S.

Supplementary material

10526_2017_9848_MOESM1_ESM.docx (118 kb)
Supplementary material 1 (DOCX 118 kb)

References

  1. Alex J (1962) The taxonomy, history, and distribution of Linaria dalmatica. Can J Bot 40:295–307CrossRefGoogle Scholar
  2. Barton K (2012) MuMIn: Multi-Model Inference. R package version 1.15.6. https://CRAN.R-project.org/package=MuMIn. Accessed 3 Oct 2017
  3. Berryman AA (2003) On principles, laws and theory in population ecology. Oikos 103:695–701CrossRefGoogle Scholar
  4. Blumenthal DM, Norton AP, Cox SE, Hardy EM, Liston GE, Kennaway L, Booth DT, Derner JD (2012) Linaria dalmatica invades south-facing slopes and less grazed areas in grazing-tolerant mixed-grass prairie. Biol Invasions 14:395–404CrossRefGoogle Scholar
  5. Briese D, Pettit W, Walker A (2004) Evaluation of the biological control agent, Lixus cardui, on Onopordum thistles: experimental studies on agent demography and impact. Biol Control 31:165–171CrossRefGoogle Scholar
  6. Buckley YM, Hinz HL, Matthies D, Rees M (2001) Interactions between density-dependent processes, population dynamics and control of an invasive plant species, Tripleurospermum perforatum (scentless chamomile). Ecol Lett 4:551–558CrossRefGoogle Scholar
  7. Burnham KP, Anderson DR (2002) Model selection and multimodel inference: a practical information-theoretic approach, 2nd edn. Springer, New YorkGoogle Scholar
  8. Carney VA (2003) Ecological interactions of biological control agent, Mecinus janthinus Germar, and its target host, Linaria dalmatica (L.) Mill., MS Thesis, University of LethbridgeGoogle Scholar
  9. Crawley MJ (1989) Insect herbivores and plant population dynamics. Annu Rev Entomol 34:531–562CrossRefGoogle Scholar
  10. De Clerck-Floate R, Harris P (2002) Linaria dalmatica (L.) Miller, broad-leaved Dalmatian toadflax (Scrophulariaceae). In: Mason P, Huber J (eds) Biological control programmes in Canada 1981–2000. CABI Publishing, Wallingford, pp 368–374Google Scholar
  11. De Clerck-Floate R, Miller V (2002) Overwintering mortality of and host attack by the stem-boring weevil, Mecinus janthinus Germar, on Dalmatian toadflax (Linaria dalmatica (L.) Mill.) in western Canada. Biol Control 24:65–74CrossRefGoogle Scholar
  12. Goulet EJ, Thaler J, DiTommaso A, Schwarzländer M, Shields EJ (2013) Impact of Mecinus janthinus (Coleoptera: Curculionidae) on the growth and reproduction of Linaria dalmatica (Scrophulariaceae). Great Lakes Entomol 46:90–98Google Scholar
  13. Jamieson MA, Knochel D, Manrique A, Seastedt TR (2011) Top-down and bottom-up controls on Dalmatian toadflax (Linaria dalmatica) performance along the Colorado Front Range, USA. Plant Ecol 213:185–195CrossRefGoogle Scholar
  14. Jeanneret P, Schroeder D (1992) Biology and host specificity of Mecinus janthinus Germar (Col.: Curculionidae), a candidate for the biological control of yellow and Dalmatian toadflax, Linaria vulgaris (L.) Mill. and Linaria dalmatica (L.) Mill. (Scrophulariaceae) in North America. Biocontrol Sci Technol 2:25–34CrossRefGoogle Scholar
  15. Nowierski RM (2004) Mecinus janthinus. In: Coombs EM, Clark JK, Piper GL, Cofrancesco AF (eds) Biological control of invasive plants in the United States. Oregon State University Press, Corvallis, pp 392–394Google Scholar
  16. Peterson RKD, Sing SE, Weaver DK (2005) Differential physiological responses of Dalmatian toadflax, Linaria dalmatica L. Miller, to injury from two insect biological control agents: implications for decision-making in biological control. Environ Entomol 34:899–905CrossRefGoogle Scholar
  17. Pinheiro J, Bates D, DebRoy S, Sarkar D (2011) nlme: Linear and nonlinear mixed effects models. R package version 3.1-127. http://CRAN.R-project.org/package=nlme. Accessed 3 Oct 2017
  18. Price PW, Denno RF, Eubanks MD, Finke DL, Kaplan I (2011) Insect ecology: behavior, populations and communities. Cambridge University Press, CambridgeCrossRefGoogle Scholar
  19. Robocker W (1974) Life history, ecology and control of Dalmatian toadflax. Tech Bull Wash Agric Exp Stn 79:1–20Google Scholar
  20. Schat M, Sing S, Peterson R, Menalled F (2011) Growth inhibition of Dalmatian toadflax, Linaria dalmatica (L.) Miller, in response to herbivory by the biological control agent Mecinus janthinus Germar. J Entomol Sci 46:232–246CrossRefGoogle Scholar
  21. Sing S, Weaver D, Nowierski RM, Markin GP (2008) Long-term field evaluation of Mecinus janthinus releases against Dalmatian toadflax in Montana (USA). In: Julien MH, Sforza R, Bon MC, Evans HC, Hatcher PE, Hinz HL, Rector BG (eds) Proceedings of the XII international symposium on biological control of weeds, La Grande Motte, France, 22–27 April, 2007. CABI Publishing, Wallingford, UK, pp 620–624Google Scholar
  22. Sing SE, De Clerck-Floate R, Hansen RW, Pearce H, Randall CB, Tosevski I, Ward SM (2016) Biology and biological control of Dalmatian and yellow toadflax. USDA Forest Service, Forest Health Technology Enterprise Team. FHTET-2015-03, Morgantown, West VirginiaGoogle Scholar
  23. Toševski I, Caldara R, Jović J, Hernández-Vera G, Baviera C, Gassmann A, Emerson BC (2011) Morphological, molecular and biological evidence reveal two cryptic species in Mecinus janthinus Germar (Coleoptera, Curculionidae), a successful biological control agent of Dalmatian toadflax, Linaria dalmatica (Lamiales, Plantaginaceae). Syst Entomol 36:741–753CrossRefGoogle Scholar
  24. USDA-NRCS (2017) The PLANTS Database (http://plants.usda.gov, 31 March 2017). National Plant Data Team, Greensboro, NC 27401-4901 USA
  25. van Hezewijk BH, Bourchier R, De Clerck-Floate R (2010) Regional-scale impact of the weed biocontrol agent Mecinus janthinus on Dalmatian toadflax (Linaria dalmatica). Biol Control 55:197–202CrossRefGoogle Scholar
  26. Vujnovic K, Wein RW (1997) The biology of Canadian weeds. 106. Linaria dalmatica (L.) Mill. Can J Plant Sci 77:483–491CrossRefGoogle Scholar
  27. Weed AS, Schwarzländer M (2014) Density dependence, precipitation and biological control agent herbivory influence landscape-scale dynamics of the invasive Eurasian plant Linaria dalmatica. J Appl Ecol 51:825–834CrossRefGoogle Scholar
  28. Wilson L, Sing SE, Piper GL, Hansen RW, De Clerck-Floate R, MacKinnon DK, Randall CB (2005) Biology and biological control of Dalmatian and yellow toadflax. FHTET-05-13, USDA Forest Service, Morgantown, West VirginiaGoogle Scholar

Copyright information

© International Organization for Biological Control (outside the USA) 2017

Authors and Affiliations

  • Aaron S. Weed
    • 1
  • Joseph Milan
    • 2
  • Mark Schwarzlaender
    • 3
  1. 1.Northeast Temperate Inventory and Monitoring NetworkNational Park ServiceWoodstockUSA
  2. 2.US Bureau of Land ManagementBoiseUSA
  3. 3.Department of Plant, Soil, and Entomological SciencesUniversity of IdahoMoscowUSA

Personalised recommendations