1. Invasion by generalist tree species can cause biotic homogenization and such community impoverishment is likely more important in rare forest types.  We quantified changes in tree diversity within Carolinian (range in Central Hardwoods), northern (range reached Northern hardwood-conifer/Boreal-spruce-fir) and central species (range in Central Hardwood region and Northern hardwood-conifer) in an old forest in southern Canada at points surveyed 24 years apart.

2. We asked: How did mature tree and sapling composition and abundance change for the 3 species’ groups? Did those changes lead to biotic homogenization? Can species’ changes be explained by community traits? We tested for differences in temporal and spatial tree β-diversity, as well as forest composition and structure, using univariate/multivariate analyses and a community trait-based approach to identify drivers-of-change.

3. Major increases occurred in abundance for mature Acer rubrum (northern), while others decreased (Fraxinus americana, Populus grandidentata); declines were found in A. saccharinum (central) and Cornus florida (Carolinian).  Species composition of saplings, but not mature trees, changed due to replacement; no evidence for biotic homogenization existed in either cohort.  As a group, northern mature tree species increased significantly, while central species declined; saplings of Carolinian species declined. 

4. Shade-tolerance in mature trees increased, reflecting successional changes, while drought-tolerance decreased perhaps due to changing temperatures, altered precipitation or ground water levels.  Saplings showed declines in all traits, probably because of compositional change.  

5. Our results demonstrated that saplings can more closely reflect change in forest dynamics than mature trees, especially over short time periods.  Based on sapling trends, this remnant could ultimately transition to a mesophytic hardwood stand dominated by A. rubrum and other shade-tolerant species, creating a more homogeneous forest. 

6. While encouraging regeneration for Carolinian and central tree species could ensure high levels of diversity are conserved in the future, it is important that this is balanced with the primary management goal of maintaining the older-growth characteristics of the forest.

Data were collected in Backus Woods, Norfolk County, Ontario, Canada (42^o 40 N and 80^o 29 west), within the Lake Erie Lowland Ecoregion.  A permanent grid of iron stakes was established by the Long Point Region Conservation Authority (LPRCA) in 1983; stakes were installed from east to west (bearing 240^o) at 50 m intervals along a series of parallel rows, 100 m apart.  Trees were surveyed at 397 grid points (totalling 3,176 sampled mature trees and saplings) in 1985 and 2009. 

Each of the permanent grid points was used as a starting point to quantify tree species abundance using the point distance sampling method. Point distance methods involved first, dividing the area around each grid point into four quadrants in each compass direction.  Within each of these quarters the closest tree was located, identified to species, and the distance of the tree from the sampling point and its diameter at breast height measured (dbh 1.3 m).  We treated mature trees (≥10 cm dbh) and saplings (<9.9 cm dbh) separately.  

Mature trees and saplings were identified to species. A total of 38 tree species was identified during the two surveys. We pooled the hybrid Acer x freemanii with A. saccharinum because we suspected observer bias; we also pooled Fraxinus profunda with F. pennsylvanica as the former was not recognized as a distinct species until 1992.  Because we were also interested in analyzing changes in abundance of rare tree species, we had to account for potential biases in our sampling due to differences in grid point locations, and so eliminated extremely rare species that occurred at less than 1% of grid points. This selection retained 28 species.

Some minor differences between years occurred in the locations used to initiate sampling. In 1985, surveyors tried to avoid disturbance to vegetation caused by the installation of stakes and so they began the point quarter distance sampling at a point away from the stake (mean distance 7.41 m).  In 2009, the stake or grid point itself was used as the starting point for the point quarter distance sampling. To minimize variation due to these slight differences in vegetation sampling locations between years, we therefore examined the community as a whole by combining multiple stakes, not analyzing individual stakes. To do so, we constructed a fishnet overlay (cells of grid points) to group sampled grid points in ArcGIS. We optimized the location and the mesh size of the fishnet to contain the maximum number of grid points per cell with a minimum criterion of 10 grid points per cell. We used this optimized fishnet for both years so that despite slight variation in the location of sampled grid points (mostly 6 m offsets), grid points belonged to the same cell in both years. These new 20 cells, containing between 11 and 18 sampled grid points, were used as sample units in all subsequent analyses.

One file is attached (Backus Woods data) with 6 data sheets:

1) Sheet 1 = Backus_data - all tree data 

2) Sheet 2 = bw_saplings - sapling data

3) Sheet 3 = bw_groups - tree species classified as 'Carolinian', 'Central' and 'Northern' (see Methods above)

4) Sheet 4 = bw_traits - TolS (shade tolerance), TolD (drought tolerance), TolW (waterlogging tolerance), SeedW (seed mass)

5) Sheet 5 = block_id (cell to which each gridpoint belonged)

6) Sheet 6 = plotcells_all