Distributional shifts of regional forest communities in the eastern United States

Listed in Datasets

By Jonathan Knott1, Michael A. Jenkins1, Christopher M. Oswalt2, Songlin Fei1

1. Purdue University 2. USDA Forest Service

This dataset contains the distribution of forest communities across the eastern United States at T1 (1980-1995) and T2 (2015-2017) and the species comprising these communities.

Version 2.0 - published on 10 Dec 2019 doi:10.4231/831Y-Z846 - cite this Content may change until committed to the archive on 10 Jan 2020

Licensed under CC0 1.0 Universal



Forest ecosystems across the globe have been impacted by climate change, resulting in species-level tree migration.  However, little is known about how forest communities are responding to climate change, which is essential to understanding the sustainability of forests and the services they provide. 


Eastern United States

Major taxa studied

Forest tree species


Using a region-wide forest inventory dataset from the USDA Forest Service’s Forest Inventory and Analysis Program with over 70,000 plots, we identified regional forest communities using the Latent Dirichlet Allocation method. We assessed community-level spatial shifts over the last three decades by quantifying directional movement in community centroid and changes to community area of coverage.  We quantified forest community changes with Jensen-Shannon distance and modeled changes as a function of climate and non-climate related variables using generalized linear mixed-effects models (GLMMs). We utilized the frequency distribution of forest communities across the various climate conditions to predict where the communities were expected to migrate to during the study period and compared the climate-predicted shifts to the observed shifts.


We identified 12 regional forest communities of the eastern United States. Most communities experienced a relatively short yet significant shift in their distribution (median = 8.0 km dec-1), but no dominant directional shift was observed during the study period. In addition, we observed a wide range of contractions and expansions of the communities’ area of coverage.  Initial climate conditions prior to the study period were significant predictors of community change over time, but the only significant climate change predictors were related to seasonal temperature variability. Climate predicted well where the communities were located, but failed to predict the magnitude of observed shifts.

Main conclusions

Forest communities have shifted their distributions over the last three decades, but our results revealed weak associations between climate change and community responses. Although climate predicted where the communities were located, climate change failed to predict the shifts in communities over the study period. The weak relationship between climate change and forest community responses may indicate a potential lag between climate change and responses of regional forest communities or the inability of forest communities to keep up with changing climate.

Keywords: Climate change, forest communities, Forest Inventory and Analysis, Latent Dirichlet Allocation, spatial shifts, tree migration

Cite this work

Researchers should cite this work as follows:



This version contains updated files from the previous version. In Version 1.0, the species composition of each community was kept constant between T1 and T2, and the data was aggregated to the 1452 km2 hexagon level before running the Latent Dirichlet Allocation (LDA) model. In Version 2.0, the species composition of each community was allowed to vary from the T1 starting values to better fit the data. The data was also aggregated after running the LDA model, which allowed associations between species to be defined at the plot level as opposed to the larger hexagon level.

The Purdue University Research Repository (PURR) is a university core research facility provided by the Purdue University Libraries, the Office of the Executive Vice President for Research and Partnerships, and Information Technology at Purdue (ITaP).