HR: 1340h
AN: GC33A-0954 [Abstracts]
TI: A new method for biometrically-based estimation of pre-settlement forest carbon stocks in the
United States
AU: * Bouldin, J
EM: jrbouldin@ucdavis.edu
AF: University of California, Davis, Plant Sciences Dept., Davis, CA 95616, United States
AB:
I present a new method to provide a historical baseline estimate of pre-settlement (19th century) forest carbon
stocks over much of the United States. Land use changes are important drivers of land-atmosphere carbon
dynamics. In the United States, quantification of terrestrial carbon loss due to alterations of pre-settlement
forests has been very imprecise, because the only data set suitable for the purpose, the General Land Office
(GLO) bearing tree data, was collected for non-ecological purposes. However, one crucial parameter necessary
for the estimation of tree density was not collected. Fortunately, other information from which that parameter can
be estimated, was collected. Tee size data was also collected, so that once the missing parameter is estimated,
tree density and size data allows for the estimation of derivative variables such as tree height, biomass, and
carbon, using allometric equations. The missing parameter is the rank order of the distance of sampled trees
from nearby survey points. The method is based on spatial simulations of tree locations and survey points, and a
numerical, maximum-likelihood based estimation procedure. The only simplifying assumption required is that
trees are more or less randomly arranged around each sample point. Tree patterns at larger spatial scales can
vary from random to highly aggregated. Estimation of density relies on a robust formula applicable to either
random or non-random spatial patterns, as developed by Morisita in 1957. Estimation of the ranked distances
involves the comparative analysis of 15 independent statistics based on the ratios of distances of pairs of trees
from survey points. The tree pairs can take on rank order values of up to 20, creating 190 independent
distributions for each statistic. The same statistics computed from actual GLO data are then compared to the
distributions from the simulated data, using a type of maximum likelihood estimation, to estimate the ranked
distance for both trees in each pair. The accuracy and precision of density estimates based on the univariate and
multivariate distributions of the various statistics is explored via further simultation.
DE: 1632 Land cover change
DE: 1851 Plant ecology (0476)
DE: 4806 Carbon cycling (0428)
DE: 4815 Ecosystems, structure, dynamics, and modeling (0439)
SC: Global Environmental Change [GC]
MN: 2007 Fall Meeting