For the primary experiment we used aerial photographs (1946-1997) and field reconnaissance to delineate a 16-ha area with recent, as well as older, conifer invasion. We established nine 1-ha (100 x 100 m) experimental plots in June 2003. Each plot contained meadow openings and forest patches of varying age and structure. Plots were randomly assigned to one of three treatments (n = 3): (1) control—no trees removed, (2) unburned—all trees removed and logging residues piled and burned leaving most of the ground surface unburned, and (3) burned—all trees removed and logging residues broadcast burned.
Tree removal occurred in January-February 2006 on deep, compacted snow to minimize soil disturbance. Larger trees were cut with chainsaws and smaller trees with a mechanical faller. Rubber-tired and tracked skidders were used to yard tree boles to an off-site landing. To reduce fuel accumulation, trees were removed with limbs attached (to the extent possible). In unburned plots, slash piles (~2 m tall, 2-4 m in diameter) were constructed by hand in June 2006. Piles were dispersed through each plot in locations not sampled for vegetation. Piles were ignited on 2 November 2006 after an extended dry period and burned to completion (95-100% consumption) within two days. In burned plots, narrow fire lines supplemented by a system of fire hoses were constructed around each plot perimeter. Slash was broadcast burned on 28 September 2006. Fine-fuel loadings (1- to 100-hr) averaged 53-69 Mg/ha. Plots burned to completion within 2 hours; flame length averaged 1-2 m and consumption of fine fuels averaged 67-87%.
Responses to tree removal and broadcast burning.— Prior to experimental treatment (June 2003), we established a permanent grid in each of the nine plots to create a system of 10 x 10 m subplots (100 per plot). Edge subplots were treated as a buffer and excluded from vegetation sampling. All remaining subplots (n = 64) were sampled in four of the plots (two broadcast burned, two unburned)—those used to reconstruct tree invasion history and associated changes in vegetation. In the remaining five plots, alternate subplots (n = 32) were sampled. For the unburned treatment, burn piles were restricted to subplot boundaries, thus avoiding areas used for vegetation sampling.
To characterize overstory composition and structure prior to treatment, we identified and measured all trees >1.4 m tall for diameter within each 10 x 10 m subplot (sampling unit). Within a subset of the subplots (32 or 64 per experimental plot), ground vegetation was sampled in four 1 x 1 m quadrats spaced at fixed distances. We estimated cover of (1) mineral soil (bare ground), (2) recent gopher disturbance (mounds and tunnel castings; post-treatment only), and (3) each vascular plant species. We also tallied all conifer seedlings (post-treatment only). Ground-surface conditions and vegetation were sampled between early July and mid-August in 2004 (pre-treatment), and at multiple points following tree removal and burning (2007, 2009, and 2013). Post-treatment samples represent 2, 4 and 8 years after tree removal and 1, 3 and 7 years after broadcast burning.
In 2013, two additional types of data were collected. First, we made a full tally of species for each 10 x 10 m subplot within each experimental plot. Second, data on ground-surface conditions and plant species composition were collected from a total of 117 “reference meadow” transects. Transects were distributed across the study site in open meadows bordering the experimental plots. Locations were chosen in areas distant from any tree influence. We used the same method to sample as in the experimental subplots: four 1 x 1 m quadrats spaced at fixed distances along a transect. At each location, the start point of each transect was chosen haphazardly.
Post-treatment soil sampling.—Following tree removal and burning, soils were collected from 15 randomly selected subplots in each plot (August 2007 and 2009). For bulk density, one soil core (0-10 cm depth, 137 cm3) was collected from the center of each subplot using a bulk density sampler; litter was first removed from the soil surface. For soil chemistry, cores (0-10 cm depth, 35 cm3) were taken with a tube-type sampler (Oakfield Apparatus, Inc., Oakfield, Wisconsin, USA). After removing surface litter, cores were extracted from two points adjacent to each of the four vegetation quadrats; the eight subsamples per subplot were composited as they were collected. On the same day, samples were set out to air dry (less than 25°C) for ~72 hours.
Local effects of pile burning.— Burn-pile scars were assessed for changes in soils and vegetation independent of the treatment comparisons. In July 2007 (1 year after pile burning) we first estimated the total cover of burn scars in each plot using the line-intercept method. Ten 80-m transects were run from random points along the southern edge of each plot (inside the buffer). Cover was estimated from the proportion of total transect length intersected by burn scars.
Vegetation was sampled in a random subset of burn scars (10 per plot, 30 total) using a design that tested for variation in response relative to the distance and direction from the edge of the burn scar. From the center of each burn scar we established a transect in a random direction across the edge of the scar into unburned vegetation. Permanent quadrats (0.2 x 0.5 m) were placed along each transect at four locations: (C) scar center (white ash or reddened mineral soil), (E) burned edge (blackened duff or charcoal), (U1) unburned edge, and (U2) unburned vegetation at a distance from U1 equal to the distance between C and E (0.5-1.7 m). Quadrats were sampled at the same time (post-treatment only), and for the same variables, as in the larger experiment.
Soils were collected from half of the burn scars sampled for vegetation (5 per plot, 15 total). One bulk density sample (10 cm depth, 137 cm3) was collected adjacent to quadrats C, E, and U2. Samples for soil chemistry were taken with a tube-type sampler (as above) at six locations adjacent to each quadrat; subsamples were composited as they were collected and were processed as in the larger experiment.
The PI trained the field crews in plant identification, plant and soil sampling, and estimating cover. Prior to data collection, crews work with the PI to calibrate cover estimates, repeating the calibration periodically. Data were collected on field forms that document sampling locations and definitions of ground-surface variables and seedling height classes. Clipboards included complete lists of species and species codes and reference materials for difficult-to-identify taxa (including line drawings and key characters). Unknown plants were collected adjacent to quadrats, labeled, and cross-referenced to field forms. Sampling checklists were used to ensure that the correct subplots are sampled. Field sheets were collected daily by the PI or crew leader and checked for legibility and completeness. Problem records (incorrect codes, illegible or missing data, and unknown plants) were noted and resolved in the office or field. Once corrected, field forms were scanned to pdf files for back-up.
Field data were entered into Excel spreadsheets using a double-entry process. The PI managed all aspects of data-entry, verification, and file management including training, distributing/receiving data-entry assignments, compiling and cross-checking double-entered files, and reconciling any differences against field sheets. Following reconciliation, corrected files were compiled by data type and subjected to quality assurance procedures (nulls, duplicates, numeric ranges, legitimate species codes, and other relational checks using rules specific to each data type).
Soil bulk density and chemistry were analyzed at the UW Analytical Services Center, Seattle. The Center follows EPA standards, including frequent calibration, and is certified by the Washington Department of Ecology. Total C and N were determined with a CHN analyzer and available NH4-N and NO3-N, were determined with an auto-analyzer. Data were output in electronic form and referenced by lab number and field location (plot and subplot or plot and burn-scar/position).
Bunchgrass Ridge is a gently sloping plateau situated along the western slope of the High Cascades in Oregon. The study area (1350 m elevation) supports a 100-ha mosaic of meadows and coniferous forests of varying size, age, and structure, reflecting nearly two centuries of tree invasion. Meadows support diverse and well-developed communities of mesic- and dry-site graminoids and herbs. Forests are dominated by Abies grandis and Pinus contorta, with herbaceous understories that vary in composition with forest age and structure.
Soils are deep (>1.7 m), fine to very fine sandy loams derived from andesitic basalt and tephra deposits with varying amounts of glacially derived rock. They grade from Vitric Melanocryands in open meadows to Aquic Vitricryands in older forests. Soil profiles indicate presence of grassland vegetation for centuries (possibly millennia), even in areas that currently support older trees. In areas of open meadow, the burrowing activities of the western pocket gopher (Thomomys mazama) contribute to considerable mixing and exposure of mineral soil.
The climate is maritime, with cool, wet winters and warm, dry summers. At Santiam Pass (1,488 m), 17 km to the north, temperatures average -6.9 deg C (minimum) and 0.7 deg C (maximum) in January, and 6.1 deg C and 27.8 deg C in July. Annual precipitation averages ~220 cm, but is highly seasonal, producing frequent summer drought. Annual snowfall averages ~11.5 m, resulting in a deep snowpack that can persist into May or early June.
