Study area and plantation sites
This study took place in the Eastern Townships region of Quebec, Canada (Figure 2), across a regional climatic gradient corresponding to an elevation gradient from the St. Lawrence Lowlands to the foothills of the Appalachian Mountains. This gradient spans three ecological regions: region1a-T, at the lowest elevations and with the mildest climate, with sugar maple and hickory (Acer saccharum - Carya cordiformis) mature forest stands typical of mesic sites; region 2c-T, at mid-elevations, characterized by sugar maple - basswood (A. saccharum - Tilia americana) stands; and finally region 3d-M, at the highest elevations, characterized by sugar maple - yellow birch (A. saccharum - Betula alleghaniensis) stands. The landscapes of southern Quebec represent an excellent opportunity to study the multi-functional potential of tree plantations. Many regions have been largely deforested in the past due to agriculture; however, agricultural activities have since been largely concentrated into the St. Lawrence Lowlands, resulting in the abandonment of many former fields outside this region (Domon and Bouchard, 2007).
The present study uses a series of similar experimental hybrid poplar plantations, planted in May 2000 for a study on hybrid poplar establishment and growth at eight sites across the Eastern Townships (Truax et al., 2012). A subset of six sites with the best poplar canopy development was used for the present study. The site characteristics are summarized in Table 4. The plantations were all established on privately-owned fields that had been used for either grazing or crops, and subsequently abandoned. The vegetation before planting ranged from herbaceous to mixed herbaceous - shrub communities. Sites were prepared for planting in the fall of 1999 by ploughing and disking, and re-sprouting or germinating vegetation was eliminated in June 2000 by an application of glyphosate herbicide over the entire plantation area. Following this treatment, 2 m long rooted cuttings of nine different hybrid poplar clones were planted by hand at each site at a stem density of 833 ha-1 (4 m between rows and 3 m between stems along the row). Glyphosate was again applied in June 2001, but only between the rows.
After 8 seasons of growth, a wood volume production gradient had emerged across the sites (Truax et al. 2012). At certain sites, the plantations showed exceptional growth, while at others they showed average or poor growth. The majority of these plantations had, however, developed a closed canopy. In this study, understory native herbaceous plants were planted in the six most successful of these plantations.
Experimental design
Each plantation contains three replicate 12 m × 12 m plots of each of nine hybrid poplar clones in a randomized block design. For this study we used only two of these clones, namely clone 915303, a hybrid of Populus maximowiczii and P. balsamifera (M × B), and clone 131, a hybrid of P. deltoides and P. nigra (D × N), both developed in Quebec. Clone 3333, another D × N hybrid, was used in place of clone 131 where the latter showed high damage. At each site, three 12 m × 12 m woodlot plots were selected within a secondary forest stand close to the plantation. At the BED and HAM sites, the woodlot plots were located in two different forest stands, both close to the plantations. These forested areas had regenerated naturally after either the harvest of a previous stand, or the abandonment of agriculture. The age of each stand was estimated based on cores taken from a sample of the largest trees and interviews with landowners. The vegetation and environmental characteristics of these woodlot plots are described in Boothroyd-Roberts et al. (2013) (see also Table 4 for dominant tree species). Thus, the present study consisted of three woodlot plots per site and three plots per site in the M × B poplar plantations and in the D × N poplar plantations, for a total of 54 plots (6 sites × 3 stand types × 3 replicates). Four woodlot plots (three from the LAP site and one from the HAM site) were subsequently eliminated from the study due to damage from fallen trees or crushed deer exclusion cages.
We established experimental plantations of understory native herbaceous plants within each of the plots described above. We chose four species typical of mature broadleaf forests in the region, with conservation value and economic value, which could potentially serve as non-timber forest products (NTFPs). These are: Asarum canadense L., Maianthemum racemosum (L.) Link (syn. Smilacina racemosa), Sanguinaria canadensis L. and Trillium grandiflorum (Michx.) Salisb. All of these species, apart from Maianthemum, are officially designated as vulnerable species in Quebec, due to the threat of overexploitation of wild populations. Asarum is sought because of the essential oils contained in its rhizomes, while Sanguinaria is sought for its rhizome containing sanguinarine, a powerful alkaloid with known medicinal uses. Trillium and Asarum are both valued for their ornamental value in shaded gardens. Maianthemum is also used for horticulture to a lesser extent. No natural populations of any of the transplanted species were present in any plot; however, a patch of Asarum was observed near one of the plots in the Bromptonville woodlot.
Transplantation of understory herbs
Small plants (well-established seedlings) of these four species were purchased from Horticulture Indigo, a specialized nursery that cultivates native plants from seed, thus avoiding an impact on wild populations. This nursery is located in Melbourne, at the centre of our study area, avoiding any unnecessary transportation time that could compromise plant quality. We chose to use plants because introductions of forest perennial herbs have been shown to be more successful from transplants than from seed (Francis and Morton 1995, Primack 1996). We planted Asarum and Maianthemum in all 54 plots over the six sites. Due to the limited availability of plants, we planted Sanguinaria and Trillium at only three of the sites (27 plots; BED, BRO, LAP). In each plot, we planted 10 individuals (Trillium and Maianthemum) or 10 clumps (Sanguinaria and Asarum) of each species. Occasionally, 11 or 12 individuals of Trillium or Maianthemum were planted in cases in which it was impossible to separate two that had grown together. In total, 1636 plants were planted. The plants were introduced into small rectangular understory plantations from which all understory vegetation, large surface roots (to 15 cm depth), and litter were removed by hand. These understory plantations consisted of two rows per species and five plants per row with a spacing of 20 cm between plants and 20 cm between the rows (total areas of 120 cm × 180 cm for four species, and 120 cm × 100 cm for two species). The experimental plants were transplanted between May 25, 2009 and June 4, 2009. No weeding has been done since transplantation, but we protected the plants from larger herbivores (mostly white-tailed deer) with exclusion cages.
Understory herb measurements
The size of all transplanted plants was recorded following transplantation, between June 10 and 18, 2009, and again in the two following springs (2010 and 2011). Sanguinaria flowers were counted on April 22, 2010 and May 5, 2011 at the Bedford site; April 29, 2010 and May 7, 2011 at the Bromptonville site; and May 30, 2010 and May 14, 2011 at the La Patrie site. All remaining counts and measurements were taken between May 21 and June 3, 2010 and June 6 and June 9, 2011. Sanguinaria flowers were not counted the first year because the flowering season had finished before the time of transplantation.
For each species, we used different measurements, appropriate for the morphology and growth patterns of each, as proxies for the total biomass of the species in the plot. For Asarum and Sanguinaria, we measured the sum of the number of leaves and the number of flowers in each plot, while for Maianthemum and Trillium, we measured the sum of the leaf length of all leaves in the plot. Maianthemum and Trillium flowers were also counted.
Environmental characteristics
The environment within each plot was characterized during the 2009 growing season by its soil chemical and physical properties, stand basal area, light availability in the understory, and leaf litter biomass. Detailed methods and results of these measurements are presented in a previous article (Boothroyd-Roberts et al. 2013). Briefly, we collected five soil samples distributed systematically within each plot for chemical analysis, which were then combined into one composite. Samples were taken from the mineral soil at a depth of between 5 cm and 10 cm, corresponding to the principal rooting zone of herbaceous plants. We measured pH in a soil-water suspension. The available potassium (K), calcium (Ca), and magnesium (Mg) contents were determined through extraction with BaCl2 and detection by atomic absorption. The extractable phosphorus (P) content was measured using the Bray-2 method (Bray and Kurtz 1945) (modified by F. Lambert). The total nitrogen (N) and total carbon (C) were measured using dry combustion, high-temperature reduction of the combustion products, and thermo-conductometric detection. Soil moisture at a depth of 10 cm was sampled twice, in June and August, corresponding to dry periods with no significant rainfall events in the 48 hours prior, and measured gravimetrically after oven drying. Light availability at a height of 90 cm was measured using a digital hemispheric photograph, taken at the centre of the plot, to determine canopy openness and the average light received during the growing season. Leaf litter was collected after almost all the leaves had fallen from every tree from a 50 cm × 50 cm microplot and was subsequently dried and weighed.
Analyses
The responses of each understory herb species were analyzed using separate 2-way analysis of variance tests (ANOVAs) with stand type (M×B poplar plantation vs. D×N poplar plantation vs. woodlot), site and their interaction as fixed factors. The LAP site was excluded from ANOVAs because data was missing from all woodlot plots from this site. This left two remaining sites for Sanguinaria and Trillium, and five sites for Asarum and Maianthemum. The experimental unit was the plot (n = 3 per cell). All response measures were subjected to square-root transformations to improve the normality of ANOVA residuals and homoscedascicity; for presentation purposes the raw data are used in figures and tables. ANOVAs were done using the JMP software package (SAS Institute, Cary, NC).
To explore the relative influence of different biotic and abiotic environmental variables on the response of transplanted plants, we fit a series of linear mixed-effect regression models, which we then used for model selection and multi-model inference. Each understory herb response was fit separately and all response variables were square-root transformed prior to analysis. The replication level for all models was the plot, using all plots except for the four woodlot plots with missing data (n = 50 for Asarum and Maianthemum; n = 24 for Sanguinaria and Trillium). We began with an initial set of eight environmental variables as fixed factors and site as a random factor to account for the spatial clustering of plots. Site was not considered a fixed factor for this analysis since the question of interest was the relative effects of the different environmental variables rather than site effects. The initial set of environmental variables consisted of elevation, June soil moisture, soil C:N ratio, soil Ca, soil Mg, soil P, leaf litter biomass and availability of diffuse light. August soil moisture, soil C, soil N, soil K, soil pH and availability of direct light were excluded from the analyses because they were highly correlated with one of the selected variables (r > 0.6). For each response variable, we selected a single best model from all possible subsets of the initial set of environmental variables, using maximum likelihood and Aikaike’s Information Criterion for small samples (AICc: Burnham and Anderson 2002). We calculated a marginal R2 for these best models as an estimate of the variance explained by the fixed factors in the model (Nakagawa and Schielzeth 2013). Because there is often uncertainty in choosing the best model, multi-model inference was also done, using all models within 4 AICc of the best model. A relative importance value was calculated for each environmental variable as the sum of the Aikaike weights of all models in which the variable was included (Aikaike weights quantify the probability that a given model is the best model). Model-averaged regression coefficients were also calculated. Initial mixed-effects models were fit using the lme4 package (Bates et al. 2013) in R software (R Development Core Team 2011) and the MuMIn package (Barton 2013) was used for model selection and multimodel inference.