Research Article |
Corresponding author: Christian D. Larson ( christian.larson@montana.edu ) Academic editor: Johannes Kollmann
© 2021 Christian D. Larson, Fredric W. Pollnac, Kaylee Schmitz, Lisa J. Rew.
This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Citation:
Larson CD, Pollnac FW, Schmitz K, Rew LJ (2021) Climate change and micro-topography are facilitating the mountain invasion by a non-native perennial plant species. NeoBiota 65: 23-45. https://doi.org/10.3897/neobiota.65.61673
|
Mountainous areas and their endemic plant diversity are threatened by global climate change and invasive species. Mountain plant invasions have historically been minimal, however, climate change and increased anthropogenic activity (e.g. roads and vehicles) are amplifying invasion pressure. We assessed plant performance (stem density and fruit production) of the invasive non-native forb Linaria dalmatica along three mountain roads, over an eight-year period (2008–2015) in the Greater Yellowstone Ecosystem (GYE), USA. We evaluated how L. dalmatica performed in response to elevation, changed over time, responded to climate and how the climate of our sites has changed, and compared elevation, climate, micro-topography (slope aspect and angle), and fruit production among sites with differing temporal trends. Linaria dalmatica stem density and fruit production increased with elevation and demonstrated two temporal groups, those populations where stem densities shrank and those that remained stable or grew over time. Stem density demonstrated a hump-shaped response to summer mean temperature, while fruit production decreased with summer mean maximum temperature and showed a hump-shaped response to winter precipitation. Analysis of both short and long-term climate data from our sites, demonstrated that summer temperatures have been increasing and winters getting wetter. The shrinking population group had a lower mean elevation, hotter summer temperatures, drier winters, had plots that differed in slope aspect and angle from the stable/growing group, and produced less fruit. Regional climate projections predict that the observed climate trends will continue, which will likely benefit L. dalmatica populations at higher elevations. We conclude that L. dalmatica may persist at lower elevations where it poses little invasive threat, and its invasion into the mountains will continue along roadways, expanding into higher elevations of the GYE.
Alien Plant Invasion, Climatic Effects, High-elevation Refugia, Linaria dalmatica, Species Range Shift
Mountainous areas are often regarded as resistant to invasive plant species due to the harsh climate and a low human footprint (
Mountain invasions by non-native plant species generally follow a process called directional ecological filtering (
The Greater Yellowstone Ecosystem, USA (GYE) is a mountainous region of high conservation value for its endemic native biodiversity and species of concern. Yellowstone National Park is a conservation area in the region and has management goals of protecting endemic and rare plant species, native biodiversity and limiting and preventing the spread of non-native species (
The invasive Eurasian perennial forb Linaria dalmatica is a species of concern within the region because of its impacts on forage, land values and native plant communities (
The importance of roadways, climate, elevation and topography for non-native species, such as L. dalmatica, laid the groundwork for our study, which investigated the temporal and spatial abundance patterns of the L. dalmatica along roadways within the mountains of GYE. Specifically, we assessed the environmental constraints shaping L. dalmatica’s abundance and propagule pressure along three roads in the mountains of GYE seven times (2008–2013, 2015) over an eight-year period (2008–2015). Our research questions were: 1) Does L. dalmatica performance (stem density and fruit production) differ along an elevation gradient? 2) How has L. dalmatica stem density changed over the course of the study? Our remaining questions investigated mechanisms driving the observed elevation and temporal trends: 3a) What climate variables best describe L. dalmatica’s performance and 3b) how have these climate variables changed over the course of the study and over a longer term (1980–2016)? 4) Do climate (using the variables that best explained stem density and fruit production), elevation, micro-topography (slope angle and aspect) and fruit production, differ between sites where L. dalmatica populations are stable/growing and where they are shrinking?
In 2008, L. dalmatica presence was surveyed along three predominately tarred roads near Gardiner, MT, USA (45°01'60"N, 110°42'33"W), one within Yellowstone National Park and the other two just north of the park boundary (Suppl. material
At each site, ten 1 m2 monitoring plots were placed randomly in patches of L. dalmatica. During the years of the study (2008–2013, 2015), L. dalmatica stem density (stems m-2) and fruit production (seed capsules m-2) were recorded at the end of each growing season (late August) in each monitoring plot. Additionally, in each plot, slope angle was taken with an inclinometer and slope aspect was taken with a compass. We had to remove one site from the study because it was located within a riparian zone that became a construction zone for a period during the project, thereby creating unrepresentative outliers.
Climate data (4 km resolution) used in the analyses were obtained from the University of Idaho’s Climate Engine, which uses gridded surface meteorological data, based on PRISM and NLDAS2 (
To address our first question, we first averaged performance data from the ten subplots at each site and created linear mixed effects models by assessing stem density and fruit production across the elevation gradient, with elevation as a fixed effect and site as a random effect to account for repeated measures through time.
To address our second question, we first assessed how stem density changed at each site over time using a linear mixed effects model with stem density as the response variable, year as a fixed effect and plot as a random effect to account for repeated measures through time. In this analysis, we observed sites with shrinking stem densities (n = 9, Suppl. material
To address the first part of our third question, we evaluated the effects of 24 climate variables on L. dalmatica stem density and fruit production. To do this, following similar methodology as previous studies (
For the second part of our third question, we used separate linear mixed effects models for the years of the study (2008–2013, 2015) and over a longer period (1980–2016) using climate variables that were significant in the stem density and fruit production models as responses, year and population group as fixed effects and site as the random effect to account for repeated measures through time.
To address our fourth question, we compared the environment of the shrinking population group with the environment of the stable/growing population group using permutational multivariate analysis of variance with a Euclidean distance (“adonis” in the R-package “vegan”;
The linear mixed effects models’ assumptions of normality and homoscedasticity were assessed by visually inspecting model residuals and fruit production and temperature data were log normally transformed. Significant relationships between predictor (fixed effects) and response variables at the P < 0.05 level were calculated from F statistics, based on Satterthwaite’s approximations of degrees of freedom. All statistics were performed in the statistical software R, version 3.6.0 (
Elevation had the strongest relationship with axis 1 of our principal components analysis, which explained 80% of the variance (Table
A relationship of Linaria dalmatica stem density and B fruit production with elevation. Fruit production regression line is back transformed predicted values derived from the mixed effects model.
The primary environmental variables of the study sites. Results from the linear regression models used to assess the strength and direction of the relationship between PCA axis 1 and the environmental variables used to ordinate the 17 study sites.
Variable | Coefficient | SE | P | r 2 |
---|---|---|---|---|
Elevation | 0.03 | < 0.001 | < 0.001 | 0.91 |
Annual precip. | 0.03 | 0.001 | < 0.001 | 0.82 |
Summer mean temp | -1.80 | 0.11 | < 0.001 | 0.71 |
Winter precip. | 0.10 | 0.006 | < 0.001 | 0.70 |
Autumn mean max temp. | -1.72 | 0.10 | < 0.001 | 0.70 |
Summer mean min temp. | -2.10 | 0.13 | < 0.001 | 0.68 |
Annual mean max temp. | -1.80 | 0.12 | < 0.001 | 0.67 |
Summer mean max temp. | -1.42 | 0.09 | < 0.001 | 0.66 |
Spring precip. | 0.06 | 0.004 | < 0.001 | 0.65 |
Autumn mean temp. | -2.03 | 0.14 | < 0.001 | 0.65 |
In nine of 17 sites, stem density decreased over the eight-year study (shrinking populations), while stem densities at eight sites remained stable (6) or demonstrated a growing (2) trend (Suppl. material
The estimated marginal means for stem density of those sites with shrinking (dark grey) and stable/growing populations (light grey) across the years of the study (2008–2013, 2015).
Assessing Linaria dalmatica stem density (stems m-2) among years for the shrinking and stable/growing populations, using estimated marginal means with Tukey pairwise comparisons. Contrast indicates the years being compared, while estimate is the difference between the mean value of the first year compared with the second.
Population group | Contrast | Estimate | SE | df | T. ratio | P |
---|---|---|---|---|---|---|
Shrinking | 2008–2012 | 4.1 | 1.1 | 89 | 3.6 | 0.01 |
2008–2013 | 5.0 | 1.1 | 89 | 4.4 | < 0.001 | |
2008–2015 | 6.1 | 1.1 | 89 | 5.3 | < 0.001 | |
2009–2010 | 3.6 | 1.2 | 89 | 3.1 | 0.044 | |
2009–2011 | 3.7 | 1.2 | 89 | 3.1 | 0.036 | |
2009–2012 | 5.1 | 1.2 | 89 | 4.3 | < 0.001 | |
2009–2013 | 6.0 | 1.2 | 89 | 5.1 | < 0.001 | |
2009–2015 | 7.0 | 1.2 | 89 | 6.0 | < 0.001 | |
2010–2015 | 3.4 | 1.1 | 89 | 3.0 | 0.054 | |
Stable/growing | 2008–2011 | -5.8 | 1.2 | 89 | -4.80 | < 0.001 |
2008–2012 | -4.5 | 1.2 | 89 | -3.7 | 0.006 | |
2008–2013 | -4.7 | 1.2 | 89 | -3.8 | 0.004 | |
2008–2015 | -4.5 | 1.2 | 89 | -3.7 | 0.006 |
Assessing Linaria dalmatica stem density (stems m-2) between shrinking and stable/growing populations in each year of the study, using estimated marginal means with Tukey pairwise comparison. Estimate is the difference between the mean value of the shrinking group and the stable/growing group.
Contrast | Year | Estimate | SE | df | T.ratio | P |
---|---|---|---|---|---|---|
Shrinking : growing | 2008 | -0.2 | 2.5 | 22.6 | -0.1 | 0.933 |
2009 | -2.3 | 2.5 | 23.2 | -0.9 | 0.365 | |
2010 | -5.7 | 2.5 | 22.6 | -2.3 | 0.034 | |
2011 | -8.7 | 2.5 | 22.6 | -3.5 | 0.002 | |
2012 | -8.8 | 2.5 | 22.6 | -3.5 | 0.002 | |
2013 | -9.9 | 2.5 | 22.6 | -4.0 | < 0.001 | |
2015 | -10.8 | 2.5 | 22.6 | -4.3 | < 0.001 |
Summer mean temperature best described L. dalmatica’s stem density response to climate, with stem density demonstrating a hump-shaped response with a peak at 12.5 °C (Table
Over the years of the study (2008–2013, 2015), summer mean and summer mean maximum temperatures increased (P < 0.001, P < 0.001, respectively). Consistent with this short-term trend, between 1980–2016, summer mean temperatures increased by an average of 0.28 °C per decade (P < 0.001) and summer mean maximum temperatures increased by an average of 0.2 °C per decade (Fig.
Linaria dalmatica stem density response to A summer mean temperature and L. dalmatica fruit production response to B summer mean maximum temperature and C winter precipitation. Fruit production regression line is back-transformed predicted values derived from the mixed effects model.
Linaria dalmatica response to climate. Results of the linear mixed effects models assessing L. dalmatica stem density (stems m-2) and fruit production (seed capsules m-2) in response to climate variables. All response variables underwent a second order polynomial transformation.
Fixed effects | Random effects | |||||||
---|---|---|---|---|---|---|---|---|
Response | Predictor | Estimate | SE | df | T value | P | Variance | |
Site | Residual | |||||||
Stem density | (Intercept) | 9.71 | 1.22 | 15.77 | 7.94 | < 0.001 | 24.04 ± 4.90 | 9.49 ± 3.08 |
Poly(su mean)1 | -21.33 | 7.14 | 103.57 | -2.99 | 0.004 | |||
Poly(su mean)2 | -10.21 | 5.18 | 114.05 | -1.97 | 0.051 | |||
Fruit production | (Intercept) | 2.55 | 0.36 | 11.07 | 7.03 | < 0.001 | 2.13 ± 1.46 | 0.78 ± 0.88 |
Poly(wint precip)1 | 7.83 | 1.57 | 113.99 | 5.00 | < 0.001 | |||
Poly(wint precip)2 | -5.03 | 1.15 | 101.80 | -4.38 | < 0.001 | |||
Poly(su max)1 | -12.74 | 1.87 | 108.41 | -6.80 | < 0.001 | |||
Poly(su max)2 | -3.58 | 1.39 | 110.50 | -2.57 | 0.012 |
Our permutational multivariate analysis of variance indicated that environments (climate, elevation, slope angle and aspect) differed between the sites of the shrinking populations and the sites of the stable/growing populations (F = 39.0, r2 = 0.25, P = 0.001). Results of our principal components analysis similarly demonstrated a difference in environments between the two population groups (Fig.
Comparisons of climate trends (1980–2016) between the sites of the shrinking populations (dark grey) and the stable/growing populations (light grey) for A summer mean temperature B summer mean maximum and C winter precipitation. Dashed vertical line denotes the beginning of the study (2008).
Principal components analysis among the 17 study sites, based on 27 environmental variables (24 climate variables, elevation, slope angle and aspect). PCA axis 1 explained 80% of the variation and PCA axis 2 explained 13% of the variation. Shrinking sites in each year are dark grey, while stable/growing sites in each year are light grey. Ellipses are standard deviations centred on the mean of each group.
High elevation and mountainous regions have historically presented more obstacles to plant invasions than lowlands (
Studies have found non-native roadside mountain invasions to be characterised by species with high temperature affinities (
It has been suggested that non-native species that are shifting upwards in elevation are not in equilibrium with their environment and their expansion is not due to climate change, but is due to filling of potential niches (
Biotic characteristics of ecosystems affect non-native plant invasions (
The observed upward shifts in elevation by non-native species have so far been closely associated with anthropogenic sources of propagules and roadways (
Interestingly, our L. dalmatica abundance and reproduction findings are consistent with a recent L. dalmatica species distribution model that found elevation and summer maximum temperature to be key drivers of its regional distribution in Montana, Wyoming and Colorado (
Fruit production has not been evaluated in other regional studies and we observed an increase with elevation and one of the most important climate variables was winter precipitation. Initially, this is ecologically perplexing; however, L. dalmatica growth and abundance responds positively to snowfall and precipitation (
Mountainous regions provide a strong environmental filter, but are also topographically complex at a local scale (e.g. slope angle and aspect), which alters climate and growing conditions (
Our results present evidence that the climate constraints of a mountainous region are weakening with a warming climate (
Our results also suggest that the predicted warming pattern will be detrimental to the L. dalmatica trailing edge populations found at lower elevations. The lower elevation L. dalmatica populations may be in an extinction lag phase, where extinction (at these locations) has been delayed, but is imminent (
Interestingly, regional climate predictions suggest precipitation will increase by mid-century (
In the mountains of the Greater Yellowstone Ecosystem, high elevations, especially on steep south-facing slopes, appear more suitable for L. dalmatica, with conditions of many low sites being too warm and dry for the species. Under future climate scenarios, as a warming climate weakens climate restrictions, L. dalmatica will likely expand its range along roads into the mountains of GYE, as has been suggested for other non-native species under climate change (
These findings are important for land managers because they show that high elevation L. dalmatica populations are producing seed and potentially acting as source populations. Therefore, we suggest the focus of L. dalmatica management be on monitoring current populations at high elevations, surveying for new populations on suitable landscape positions and containing high elevation roadside populations before the invasion expands into interior native communities under a changing climate.
Funding for this project was provided by the National Research Initiative, project number 2009-55320-05033. LJR is supported by the National Institute of Food and Agriculture, U.S. Department of Agriculture Hatch MONB00363. We would like to thank the United States Forest Service and the National Park Service for their cooperation. We would also like to thank Adam Gebauer, Barbara Keith, Alexandre Wing, Kimberley Taylor, Landon Zimmerman, Jordan Meyer-Morey and Mel Cheeseman for assistance in the field. Finally, we would like to thank the reviewers and the Subject Editor, Johannes Kollmann, for their invaluable insight and comments, which made this manuscript possible.
Figure S1. Location of study sites (17) by (A) road, (B) climate cell, and (C) population grouping
Data type: occurrence
Figure S2. Linaria dalmatica stem density (stems m-2) over the course of the study (2008–2013, 2015) at those sites with shrinking populations
Data type: occurrence
Explanation note: Each site is identified by road and elevation.
Figure S3. Linaria dalmatica stem density (stems m-2) over the course of the study (2008–2013, 2015) at those sites with stable/growing populations
Data type: occurrence
Explanation note: Each site is identified by road and elevation.
Table S1. Correlation matrix
Data type: statistical data
Explanation note: Correlation matrix for elevation (m), slope aspect (°) and angle (°), and the temperature (°C) and precipitation (mm) variables from which climate models were built. ‘Mean’, ‘maximum’, ‘minimum’ refer to temperature (°C).