Comparing tree-ring and permanent plot estimates of aboveground net primary production in three eastern US forests

Forests account for a large portion of sequestered carbon, much of which is stored as wood in trees. The rate of carbon accumulation in aboveground plant material, or aboveground net primary productivity (aNPP), quantifies annual to decadal variations in forest carbon sequestration. Permanent plots are often used to estimate aNPP but are usually not annually resolved and take many years to develop a long data set. Tree rings are a unique and infrequently used source for measuring aNPP, and benefit from fine spatial (individual trees) and temporal (annual) resolution. Because of this precision, tree rings are complementary to permanent plots and the suite of tools used to study forest productivity. Here we evaluate whether annual estimates of aNPP developed from tree rings approximate estimates derived from colocated permanent plots. We studied a lowland evergreen (Howland, Maine), mixed deciduous (Harvard Forest, Massachusetts), and mixed mesophytic (Fernow, West Virginia) forest in the eastern United States. Permanent plots at the sites cover an area of 2-3 ha, and we use these areas as benchmarks indicative of the forest stand. We simulate random draws of permanent plot subsets to describe the distribution of aNPP estimates given a sampling area size equivalent to the tree-ring plots. Though mean tree-ring aNPP underestimates permanent plot aNPP slightly at Howland and Fernow and overestimates at Harvard Forest when compared with the entire permanent plot, it is within the 95% confidence interval of the random draws of equal-sized sampling area at all sites. To investigate whether tree-ring aNPP can be upscaled to the stand, we conducted a second random draw of permanent plot subsets simulating a twofold increase in sampling area. aNPP estimates from this distribution were not significantly different from results of the initial sampling area, though variance decreased as sampling area approaches stand area. Despite several concerns to consider when using tree rings to reconstruct aNPP (e.g., upscaling, allometric, and sampling uncertainties), the benefits are apparent, and we call for the continued application of tree rings in carbon cycle studies across a broader range of species diversity, productivity, and disturbance histories to fully develop this potential.