College of Food, Ag & Nat Res Sci
Allometric models are routinely used to estimate biomass at the individual tree scale, which forms the basis for estimating forest carbon stocks within national greenhouse inventories. However, these models may be highly uncertain, owing to a lack of comprehensive “felled tree” datasets for even common tree species. Further, this error is frequently ignored when scaling up to develop biomass estimates at the national scale. This project applies a Bayesian hierarchical modeling framework for predicting aboveground tree biomass and propagating resulting uncertainty into national stock estimates with the US Forest Service Forest Inventory and Analysis (FIA) data. These researchers take advantage of a large, multispecies felled tree database to fit their models and to quantify resulting uncertainty when applied to observations within FIA.
This work involves generating posterior predictive distributions of biomass for many individual trees (~3,000,000 individual records within FIA) via Markov chain Monte Carlo (MCMC) procedures. Derivation of these posterior distributions, as well as subsequent processing of the results into areal estimates for the whole United States, is very time-consuming on standard computers.