Site Description
The study site is located at the Rogers Research Site (RRS) (42.236679 N, -105.344440 E) in the North Laramie Mountains in southeastern Wyoming. RRS is managed by the Wyoming Agricultural Experiment Station and owned by the University of Wyoming (UW). Elevation ranges between 2000–2200 m at the site and is approximately 130 ha in size with moderate to steep slopes (5–50%). Mean annual precipitation is 37.6 cm with mean annual temperature ranging between 14 ̊C to less than 0 ̊C.
Understory vegetation associated with ponderosa pine forests in the Laramie Mountains prior to wildfire disturbance were shrubs–primarily in the Rosaceae family like antelope bitterbrush (Purshia tridentata), choke cherry (Prunus virginiana), service berry (Amelanchier alnifolia), and Wood’s rose (Rosa woodsii). Forbs included cinquefoil (Potentilla spp.), fringed sagewort (Artemisia frigida), geraniums (Geranium spp.), milkvetch (Astragalus spp.), and yarrow (Achillea millefolium). The dominant grasses and sedges that occur were bluebunch wheatgrass (Pseudoroegneria spicata), wild Idaho fescue (Festuca idahoenis), prairie junegrass (Koeleria macrantha), and Geyer’s sedge (Carex geyeri) (Howard 2003).
RRS soils are moderately deep (50–100 cm) and coarse textured on hillsides and ridges; where the water table is high, thick, dark and fine-textured soil occurs. Soil pH ranges from 7.2 in the top 10 cm and 6.4 below 10 cm (Williams and Waggener 2017; Wilkin et al. 2019). Soils at the study site are characterized as moderately developed Alfisols and shallow Entisols with low fertility and low water-holding capacity formed from granitic weathering. Alfisols and Entisols are classified as fine-loamy, mixed, superactive, frigid Typic Haplustalfs moderately deep and loamy-skeletal, mixed, micaceous, frigid Lithic Ustorthents shallow, respectively (Munn et al. 2018).
The 2012 Arapaho Fire occurred during one of the driest years on record in the state (Scasta 2015). Temperatures for the Arapaho Fire at RRS reportedly ranged from 200–500 ºC based on black and white soil surface ash color post-fire, ensuring complete consumption of the organic soil horizon in some areas (Wilkin et al. 2019).
According to (Seymour, Driese, and Waggener 2017) ponderosa pines covered approximately 80% of the RRS with trees in different age classes. After the Arapaho Fire, approximately 5% of the trees remained, converting a forested landscape to a shrub and forb dominated landscape.
Study design
A block design was implemented in the summer of 2015 to determine the best combination of management treatments for ponderosa pine restoration following a high-severity wildfire (Herget et al. 2018). Four blocks were established at RRS, each block comprising of 18 plots of 50 x 50 m (0.25 ha) in size.
Each plot received different combinations of a pine introduction method (natural regeneration, broadcast seeding, and planted seedlings), a logging method (no logging, bole only removal, whole tree removal), and an erosion control seeding (no erosion seeding and seeding with a native grass mix) for a total of 72 plots across all four blocks (Fig. 1). To account for edge effect all measurements and surveys were conducted in a subplot of 27 x 27m (0.07 ha) in the center of each plot. From here on logging methods ‘bole only removal’ and ‘whole tree removal’ will be referred to as ‘bole only’ and ‘whole tree’.
In total 2,400 one year old, nursery grown ponderosa pine seedlings were planted in all ‘planted seedling’ pine introduction treatments with 100 seedlings planted per plot 3 m apart in a grid system within the inner subplot. Nursery stock came from Colorado State Forestry Nursery in Fort Collins, Colorado. Seeds used in the ‘broadcast seeding’ plots originated from the Roosevelt National Forest in north-central Colorado. Seeds were kept in cold storage and had 70% germination viability. Seeds were dispersed using a hand-held broadcast seeder at 158 grams per subplot.
The erosion treatment was implemented by broadcast seeding native grass species; included in the erosion grass seed mix were ‘Bromar’ mountain bromegrass (Bromus marginatus), ‘Lodorm’ green needlegrass (Nassella viridula), ‘Pryor’ slender wheatgrass (Elymus trachycaulus), and ‘Winchester’ Idaho fescue (Festuca idahoensis) with 92% germination viability. Treatments were randomly assigned to plots based on the feasibility of a treatment establishment as determined by slope and topography of a plot (Herget et al. 2018).
Seedling surveys
Ponderosa pine seedling surveys were done for all pine introduction treatments in the summer of 2017, while ‘planted seedlings’ plots were also surveyed in 2015, one to two months after seedlings were planted and in the fall of 2016. Surveys were conducted by walking each subplot in 3 x 3 m grids and counting all ponderosa pine seedlings and marking each as live or dead in planted treatments while seedling presence in broadcast-seeded and natural regeneration treatments were counted and marked live. All seedlings were photographed and marked with a Garmin global positioning system (GPS). Seedling numbers per plot were converted to stems ha− 1 for statistical analysis.
Vegetation Surveys
Vegetation surveys were done in the summer of 2017 (June and July) for all plots. Starting at the northeast corner of each plot, five 0.5 m2 quadrats were read at 15, 20, 25, 30, and 35 m along a 50 m transect within the subplot using the Daubenmire method for vegetation cover. Percent bare ground, ground cover (rock, lichen, litter, and woody litter), and both native and non-native plant species cover were recorded in each quadrat (Coulloudon et al. 1996).
Statistical analyses
We were interested in the mean of ponderosa pine seedlings, vegetation functional groups, and ground cover among the three pine introduction methods and the three logging methods, and the combination of these factors at the Rogers Research Site. Comparisons of interest were carefully planned and defined before any data was collected, therefore no multiple comparison adjustments were used. All statistical analyses were done in R version 4.0.3 (R Core Team 2020).
We fitted a Bayesian linear mixed-effect model with a negative binomial distribution using ponderosa pine seedling counts as the response variable, pine introduction method and logging method and their interaction as the fixed effects, and block as the random effect using the blme package (Chung et al. 2015) and lme4 package (Bates et al. 2014).
We fitted a generalized linear mixed-effect model (GLMM) with a Tweedie distribution to estimate differences in the mean vegetation functional group cover and ground cover between pine introduction and logging methods at RRS using the glmmTMB package (Brooks et al., 2017). The mean functional group cover and ground cover were the response variable, pine introduction and logging methods were the fixed effects, and block was the random effect.
The model with the lowest AIC score was chosen to determine what fixed effects were most important to include in our model for vegetation functional group cover and ground cover. Erosion control seeding was a fixed effect in models with the highest AIC score, therefore we removed it from our models. However, to determine if erosion control seeding did affect the mean percent cover of the four grass species in the erosion control seed mix, invasive species, and bare ground, a separate GLMM was fit with the pine introduction method, logging method, and erosion control seeding as the fixed effects and block as the random factor.
Residuals from the fitted models were graphically checked with the DHARMa package (Hartig 2021) and model assumptions of constant variance and normality were reasonably met. The blocks in the study were assumed to be independent of one another. The estimated marginal means and contrasts for each model was derived using the emmeans package (Lenth 2021). To test for an overall treatment effect a Wald Chi-Square test was performed on the chosen models to determine the degrees of evidence against the null hypothesis for the pine introduction method, logging method, and erosion control seeding.
A permutational multivariate analysis of variance (PERMANOVA) with a Bray-Curtis dissimilarity index was used to determine the effect of logging and pine introduction methods and erosion seeding on the vegetation functional groups using the adonis function in the vegan package. (Oksanen et al. 2020). Permutations were constrained using block. Pairwise comparisons with a Bonferroni correction was done among logging methods and pine introduction methods using the pairwise.perm.manova function in the RVAideMemoire package (Hervé 2021).