Suppression of reed canarygrass by assisted succession: A sixteen-year restoration experiment

Assisted succession could enable long-term restoration of invaded areas where successional trajectories have stalled due to competition from invasive species. Many invasives are shade-intolerant, therefore interventions that reduce light availability should suppress invasion and simultaneously re-establish successional processes. However, restoration success also depends on identifying critical system thresholds, e.g., invader abundances below which regeneration of desired species is possible. We report the successful use of assisted succession to restore a swamp forest invaded by Phalaris arundinacea (reed canarygrass; hereafter Phalaris), initiated by a high-density planting of woody species to outcompete the invader by reducing light availability.

We established five pre-planting treatments in a Phalaris near-monoculture in Wisconsin, USA: herbicide-only, herbicide+plow, herbicide+burn, herbicide+mow, and control. In 2003 we planted 23 tree and shrub species at high densities, then in 2019 we censused the site to: (1) evaluate the effect of our interventions on community composition, (2) document trends in community change over time, and (3) determine light availability thresholds that influence community composition.

We found no differences among pre-planting invader removal treatments. Late fall glyphosate application suppressed Phalaris long enough that a dense canopy of native woody species could establish and eventually out-shade it. Overstory densities of 0.071/m2 suppressed Phalaris to 50% cover, but, due to nonlinearities, much higher densities were needed to reduce light availability and thus Phalaris cover enough to shift the system from being invader-dominated. Regeneration of the woody species we had initially planted suggests long-term restoration success.

Synthesis and applications. An empirical understanding of long-term community dynamics can help manage invasive species and restore target plant communities. We show a cost-effective restoration strategy for forests invaded by shade-intolerant invaders that arrest succession. Our data indicate that establishing a dense canopy of woody species through assisted succession can re-introduce feedbacks enabling long-term ecosystem recovery. We also illustrate the value of identifying critical thresholds influencing the abundance and impact of key invasive species.

Methods

Study area and treatments: This study was conducted in southeastern Wisconsin, USA at the Huiras Lake State Natural Area (Lat: 43.5136095, Lon: -87.9828698). In 2002, 0.50 ha of a near-monoculture field of Phalaris was divided into 50 plots ranging from 30.3 to 154 m2. Except for two large blocks left as eight untreated control plots, the entire site was sprayed with 0.7% active ingredient glyphosate solution in early November 2002; such late-season herbicide application is recommended for Phalaris (Adams & Galatowitsch, 2006). Four pre-planting invader removal treatments were established in sprayed plots between fall 2002 and spring 2003 (Hovick & Reinartz, 2007): 1) herbicide-only (H) – no additional pre-planting treatment; 2) mow+herbicide (MH) – Phalaris biomass mowed in early fall 2002, before herbicide application; 3) herbicide+burn (HB) – herbicide application followed by burning Phalaris litter in spring 2003; and 4) herbicide+plow (HP) – herbicide application followed by plowing in spring 2003 (Figure S1).  

In spring 2003, 22 tree and shrub species were planted by hand at a density of 9,500/ha (0.95/m2) across all plots (4668 individual plants; 29 to 146 per plot, depending on plot size). Most species were bareroot stock, although American elm, yellow birch, and basswood were small, rooted plugs and willows were live stakes. Initial planting numbers varied widely among species, due to availability (Table 1). No additional Phalaris management occurred post-planting. 

Sampling design: We surveyed vegetation in July 2019, complementing this data set with survival data from summer 2003 and 2004 (Hovick & Reinartz, 2007). The 2019 survey had three main components: 

1. Groundcover - We quantified Phalaris cover and that of any non-Phalaris herbaceous species, recording visual estimates of each group in two 1m2 quadrats per plot. We used modified Daubenmire cover classes: <5%, 5-24%, 25-49%, 50-74%, and 75-100% (Daubenmire, 1959), allowing the sum of Phalaris and non-Phalaris cover to exceed 100%. For each quadrat, we also estimated cover of litter, vegetation, or bare soil, constraining the summed estimates of these components to equal 100%. Prior to analysis, cover estimates were transformed to the midpoint of each cover class (restricting midpoints to minimum 2.5%, maximum 87.5%) and averaged to yield a single estimate per plot.  

2. Tree and shrub abundance – To characterize the woody plant community, we separately quantified species-specific abundances in the overstory, sapling layer, and seedling layer. In the overstory, we recorded all woody species in the plot ≥2.5 cm diameter at breast height (DBH). Due to widespread mortality in some species (most notably green and black ash [Fraxinus americana and F. nigra]), we noted whether overstory individuals were alive (foliage green), dead (no green foliage), or resprouting (most foliage dead, but some recent growth with green leaves). We sampled saplings within two 2×2 m sampling quadrats per plot (each of which contained a groundcover sampling plot at one corner), counting as saplings all woody species >1 m tall and <2.5 cm DBH. We sampled seedlings using four 1 m2 quadrats per plot, positioning two quadrats at opposite corners within each sapling quadrat. We counted as seedlings all woody species <1 m but >30 cm in height. Individuals in each stratum were identified to species whenever possible, except for lumping together ashes (as Fraxinus spp.) and dogwoods (Cornus sericea and C. stolonifera as Cornus spp.).  

3. Light availability - Total light availability above the herbaceous level was estimated as percent canopy openness using hemispherical photos. We took two pictures per plot, centered in each sapling quadrant, and analyzed with Gap Light Analyzer, Version 2.0 (Frazer et al., 1999). Data were averaged to yield a single estimate per plot.  

Statistical analysis: Using census data from three years (2003, 2004, 2019), we quantified density and survival trajectory. Density is per-plot abundance of a given species divided by plot area (overstory) or sampling area (saplings, seedlings). Survival trajectory is per-species density, relative to the number of planted individuals (2003) or relative to density in the preceding survey (2004, 2019). Survival trajectory in 2019 was divided by fifteen to compare all density changes on a per-year basis.  

As preliminary tests of differences among removal treatment types, we compared light availability, Phalaris cover, and overstory density across treatments using linear mixed models (R Core Team, 2021) lme4::lmer, (Bates et al., 2015); car::Anova, (Fox & Weisberg, 2019) with random effects, included to account for spatial nonindependence among plots within blocks and among blocks themselves (Figure S1). All pairwise comparisons among removal treatments and controls were then contrasted with Tukey’s post-hoc tests. Seeing negligible differences among removal treatments themselves, as expected, we shifted our focus to be on Phalaris removal vs. control differences in density, survival trajectory and groundcover percentages. We present linear mixed models followed by Dunnett’s post-hoc tests comparing each treatment to the control (emmeans, (Lenth et al., 2020)). Most response variables were nonnormal, but model residuals were largely normally distributed, supporting model fits. 

We estimated nonlinear relationships at the plot scale (n=50), combining data from all pre-planting treatments to describe 1) the ability of increased overstory densities to decrease light availability and (by extension) Phalaris cover, 2) the ability of decreased light to suppress Phalaris, and 3) the impact of Phalaris cover in suppressing sapling density. These models allowed us to estimate, for example, the overstory density or reduction in light availability required to suppress Phalaris to a target percent cover. Preliminary analyses indicated no support for treatment-specific models and that these relationships were best characterized using log-logistic curves (drc::drm (Ritz et al., 2015)), with slope and inflection point the estimated parameters. We fixed lower and upper limits to realistic ranges of y: 0 to 100 for Phalaris percent cover or light availability, and 0 to 3.125 saplings/m2 (the maximum observed value) for sapling density. We used these curves to estimate predictor values required to suppress Phalaris to 50%, 25%, and 10% cover. For the three models with Phalaris cover as a variable, we achieved estimates directly by substituting the target percentage and solving for the corresponding value. We used model-estimated light availability corresponding to 50%, 25%, and 10% Phalaris cover to predict overstory densities required to achieve those light levels. These estimated nonlinear relationships imply causal direction, but to examine other relationships lacking clear causality we calculated Pearson correlations (stats::cor, stats::cor.test, (R Core Team, 2021).