Sampling was conducted in two study areas, each centered on a large watershed in the central western Cascades of Oregon, USA. The Blue River study area has deeply dissected terrain characteristic of the eastern part of the western Cascade Range. It includes the 240-km2 Blue River watershed plus 33 km2 to the north. Elevation ranges from 316 to 1753 m, and most major ridges reach elevations above 1200 m. The Fall Creek study area is the eastern 300 km2 of the Fall Creek watershed. It has lower topographic relief, representative of the western part of the western Cascades. Elevation ranges from 254 to 1519 m, but only the highest ridges along the southern perimeter of the study area are higher than 1200 m.
Sampling was conducted in 124 stands: 71 in the Blue River study area and 53 at Fall Creek. Stands were located using a stratified random method that distributes sampling intensity throughout each study area while capturing most of the physiographic variation. Sample sites were identified by randomly selecting slope facets (areas of common aspect extending from ridgetop to valley bottom) across each study area and delineating the forested area at upper, mid, and lower slope positions of each facet. The forested area at each slope position was considered a stand, and a point was generated randomly within each stand to serve as the midpoint of a 120-m-long transect for sampling forest stand and age structure. Each transect consists of five, 0.02-ha circular plots at 30-m intervals. The entire transect is treated as a sample unit, providing one transect per stand and three stands per slope facet in most cases. Sampling was conducted in five plots along a 120-m transect rather than a single plot of equivalent area to increase the likelihood that age classes are spread over broad areas, and thus, are likely to represent effects of widespread disturbance rather than local tree-fall gaps.
To select sample sites, a 5 × 5-km grid was overlaid on each study area. Then at least one slope facet was randomly selected in each 25-km² cell, and the forest stands at upper, mid, and lower slope positions were delineated. Finally, the midpoint of one 120-m-long transect was located within the forest stand at each of the three slope positions. Slope facets were delineated by first applying a Gaussian filter to a 30-m Digital Elevation Model (DEM) to smooth features up to 240 m in extent. Then, aspect of the smoothed DEM was classified as northeast (0–90º), southeast (91–180º), southwest (181–270º), or northwest (271–360º), and areas of each aspect class smaller than 40 ha were combined with the more inclusive class (Fig. A1a). Slope facets were chosen for sampling by using a random number generator to identify one, 1-km² ell within each 25-km² cell, and selecting the slope facet that accounts for the majority of that cell.
Stands for sampling were located by generating a random point at the upper, mid, and lower slope positions of the selected slope facet. The slope positions were defined by dividing the elevation range of the slope facet into three equal portions (Fig. A1b). Aerial photos were used to draw polygons encompassing the forested area lacking 20th-century anthropogenic disturbance at each slope position, and a 100-m-wide buffer was applied to all roads, perennial streams, and harvested stands to avoid potential edge effects. Then, Universal Trans Mercator (UTM) coordinates were generated for a random point within the polygon at each slope position. This point served as the midpoint of a 120-m-long transect, and a GPS receiver was used to locate it in the field. The slope aspect was measured in the field, and the bearing to the additional plots was oriented perpendicular to the aspect (i.e., along the slope contour).
In addition to sampling along one slope facet in each 25-km² grid cell, topographic settings that were poorly represented in the data set (e.g., benches, cirques, and rocky ridges) were selected for sampling. Sample locations within these settings were identified by delineating the forest stand within the landform and randomly selecting a random point as the midpoint of the transect as described above.
The data set includes 32 slope facets where sampling was conducted in the stands at upper, mid, and lower slope positions (32 slope facets × 3 stands per facet = 96 transects). On the four largest of these slope facets, sampling at the three slope positions did not fully represent the diversity of topographic settings and stand conditions. A total of 6 additional transects was sampled on the four largest of these 32 slope facets, where the 3 slope positions do not fully represent the diversity of topography and stand conditions along the slope facet. These transects were located by delineating fine-scale landforms (e.g., benches or cirques) within the slope and generating a random point within the landform to serve as the midpoint of the transect, as described above. Due to the unavailability of forest lacking 20th-century anthropogenic disturbances, sampling was conducted in only two of the three slope positions along six additional slope facets (6 slope facets × 2 stands per facet = 12 transects). The remaining 10 transects were represented as single stands along 10 additional slope facets to include landforms, such as rocky ridges, benches, and cirques, that were otherwise poorly represented in the data set. Thus, sampling was conducted on a total of 48 slope facets, with at least three stands sampled on 32 of these facet, 2 stands sampled on 6 facets, and a single stand sampled on 10 facets.
Stand-structure data include the diameter at breast height (dbh) of all live and standing dead trees >15 cm dbh throughout each 0.02-ha plot and the number of saplings and shrubs 1.5–15.0 cm dbh by species in a 0.01-ha subplot within each plot. Evidence of fire, including charred bark and open catfaces (wounds usually extending to a height of 1–2 m from the base of the tree and commonly formed due to damage to the cambium by fire), was recorded for each tree.
Age-structure data were collected by coring a subset of the live trees >15 cm dbh (>10 cm dbh for Pacific yew [Taxus brevifolia Nutt.]). The subset was determined by dividing each 0.02-ha plot into four quadrants and coring the largest tree of each species in each quadrant. These criteria ensured that each species was sampled nearly proportional to its frequency in the transect. Selecting the largest tree increased the likelihood of sampling the oldest individuals of each species, but sampling one tree per species per quadrant ensured that smaller trees also were sampled. In all, 3277 trees were cored, representing an average of 27 trees per transect, or 76% of the live trees >15 cm dbh. Eighty-five percent of the cores were cross-dated. Establishment dates were estimated for 3038 trees, limited to cores that intersected the pith or where the inner ring formed a complete arc.
In the field, trees were cored as close to the ground as possible and the height of core extraction above the mineral soil surface was recorded for each tree. If trees had heartrot or were too large to reach the pith with an 81-cm-long increment borer, the area in close proximity to the transect was searched for trees of the same species and of similar size to substitute for the tree in the plot. If no suitable trees were available, an incomplete core from the tree in the plot was collected as a minimum estimate of tree age, but establishment dates were not estimated for these trees. Therefore, the incomplete core could contribute to the proportion of shade-intolerant and shade-tolerant trees that established before 1780 (i.e., if the innermost ring of the incomplete core was before 1780), but it did not contribute to the calculation of any of the other age-structure variables.
In the lab, tree cores were mounted and sanded until cell structure was visible under a binocular microscope, and the majority (85%) of cores was cross-dated. Cross-dating was conducted using skeleton plotting and the list-year method (Yamaguchi 1991, Stokes and Smiley 1996). Then, master ring-width chronologies were developed by measuring ring width in more than 200 Douglas-fir and 50 western hemlock cores to the nearest 0.001 mm using a Velmex sliding-stage micrometer. The remaining cores were cross-dated by visual comparison to the master chronologies. Fifteen percent of those cores that could not be cross-dated were angiosperm species, which were not included in the age-structure data set due to representation in a small portion (23%) of stands. The remaining non-cross-dated cores were primarily from shade-tolerant species that had suppressed growth in their first 100 years, which precluded cross-dating of the innermost rings. These cores were retained in the analyses because potential dating errors were assumed to be small relative to the timescale of stand development, and such errors were likely to have little influence on the age-structure variables used in the analysis.
Establishment dates were estimated for 3,038 trees (89% of the cores), limited to those that either intersected the pith, or where the innermost ring formed a complete arc. The geometric method of Duncan (1989) was applied to off-center cores to estimate an average of seven years to the pith. An additional four years, on average, was added to account for the time to reach coring height above the mineral soil surface by applying the equation of Morrison and Swanson (1990) to all species. This equation assumes that trees with wide radial growth of their innermost rings reach a given height sooner than those with narrower inner rings.
