Regional Scale Uncertainty Estimates from Fine-Scale Forest Inventory: Stored and Accumulated Forest Carbon in the Eastern US.

Forests are an important sink in the global carbon budget, so understanding potential consequences of climate change on forest biomass and productivity is of key concern. However, understanding how interactions between different dimensions of climate change and local conditions like soil and topography affect forest biomass requires proper treatment of uncertainty, which derives from multiple sources. We show that uncertainty in from sampling forest biomass and productivity should be quantified by modeling what is actually observed– the numbers of trees on a plot, their size, and the species composition. Accommodating these sources of variation is complicated by two issues.  First, large-scale forest inventories provide noisy data, due to small plot size, with the important, large individuals that contribute most to biomass being especially sporadic.  Second, the numbers of individuals, their sizes, and species compositions are not independent.  FIA plots of 1/16 ha in eastern US can be used to estimate uncertainty only at the 1/16 ha scale—the observation scale.  Uncertainty at this scale is hardly relevant for continent-wide inventory estimates.  More valuable would be estimates at, say, the 1-ha scale, a scale that is not observed, but can be linked to observations through models.

We developed a Bayesian statistical model framework for predicting hectare-scale forest biomass and productivity, with uncertainty in both, using information on climate, soil, and topography. This model is novel in the way interactions between densities of individuals, sizes of individuals, and species composition are handled.  The dependence on tree number and size structure provides a link to area, because the numbers and sizes of trees, rather than plot area per se, are the principle source of uncertainty that can vary across scales. We applied this model to predict 1.5 million pixels of forested land in the eastern United States and found 45.0 ± 0.4 billion trees greater than 12.7 cm in diameter contain an aboveground biomass of 13.7 ± 0.14 Pg. Sensitivity calculations of biomass productivity (t/ha/yr) from the model indicate widespread interactions between stand age, climate, soil, and topography. This method not only offers not only out-of-sample biomass and productivity estimates with meaningful uncertainty at the hectare scale, but also provides insight on the consequences of future climate change to in-situ forest biomass.