Long-term forest dynamics in the North Carolina Piedmont: A real-time evaluation of forest succession using permanent-sample plots

Background/Question/Methods

For more than a century, succession has been a focal point in plant ecology. Long-term studies are uncommon but critical for understanding forest succession due to the complexity of ecological change through time. We used 80 years of permanent-sample plot data in the Duke Forest to examine in depth the long-term, real-time patterns of population and community dynamics of trees across a long-term successional sequence. Specifically, we examined the long-term trends of species composition and whether successional trajectories empirically support classical predictions about the transition from even-aged, old-field pine stands to mature, mixed-age hardwood forests dominated by oaks and hickories. Our permanent plots document species identity, height, and diameter of every tree in thirty-four 400-1000 m²plots every 5 years spanning a successional sequence from 1934-2013. Mortality and ingrowth were also recorded each sampling period. We calculated basal areas for each species, plot and year and used these data to map samples into a reduced ordination space using Nonmetric Multidimensional Scaling. A joint plot with species ordination scores was computed with weighted averaging of sample scores. We used agglomerative hierarchical clustering to group the data, and we added change vectors to trace individual plots across time in the ordination space.

Results/Conclusions

A scree plot from the step down process indicated that 2 axes optimized the explanatory power of the ordination. Additional permutations resulted in an optimized ordination with a stress of 0.245. After rotating the ordination axes using PCA, the 1^st and 2^nd axes had R² values of 0.527 and 0.255 respectively and collectively explained 78% of the variation among the samples. Cluster analysis illustrated 4 groups that can be described using species ordination scores from the joint plot: early successional pine plots dominated by Pinus taeda; late-successional mesic plots associated with maples, elms and beeches; mature oak-hickory plots; and xeric upland forests associated with Quercus stellata. The change vectors indicated that the pine stands have succeeded predominantly toward a more ‘mesic’ forest type and not toward the predicted oak-history cluster, seemingly contradicting the predictions of classical models. Alternatively, this may indicate a regional shift in successional patterns, necessitating further correlational studies to indicate what is driving this trend. This study provides an updated, directly-informed understanding of long-term successional patterns in Southeastern U.S. forests to allow for continued and enhanced utility of succession as a predictive model in the face of land-use and climate change.