Ecological Archives A 020 - 020 - A 5

I. Height allometries Since 1999, FIA data include height measurements. However, pre-1999 FIA data (which currently comprise most of the available plot re-measurements) do not include height data for most of the U.S. Therefore, incorporating tree height into analyses of pre-1999 FIA data requires height-dbh allometries, which we estimated using the post-1999 FIA data. Because we wished to treat our field data the same as we would treat FIA data, we used these same allometries in the analyses presented in this paper. In December 2006, we downloaded all available FIA data from www.fia.fs.fed.us/, and we estimated height-dbh allometries for all species listed in FIA documentation. We excluded data from plantations and plots with any documented harvesting. To reduce the computational requirements for the analysis, we randomly subsampled from the data for species with unnecessarily large sample sizes (see Subsampling algorithm below). To constrain the height allometry parameters for rare species, we adopted a Bayesian hierarchical modeling approach (Gelman et al. 2004, Clark 2005, Ogle and Barber 2008) with the following nested taxonomic levels: species, genus, family, order, division, and base, where the base is the ‘root’ of the taxonomic hierarchy. Our approach is ‘hierarchical’ both in the statistical sense (i.e., each allometric parameter was assumed to be a sample from a probability distribution) and in the taxonomic sense (i.e., we used the taxonomic hierarchy to structure the analysis). This approach allows rare taxa to borrow support from common taxa, particularly those that are closely related. We assumed a power law relationship between height and dbh, which implies a linear model in log-space: