A Comparison of Absolute and Relative Stand Density Measures Used in the Northeastern United States

The quantification of observed stand density relative to a desired density, a measure known as relative density, is a critical component of many silvicultural treatments. Calculating stand density is more complicated when there are multiple species involved. I compared four regionally appropriate relative density measures to frequently used absolute density measures such as biomass, basal area, trees per area and stand density indices. I found absolute measures inferior to relative measures of stand density in that they lack an accepted reference point that allows for the comparison of one stand to the next in a meaningful and biologically accurate way. Focusing on the four relative density measures I explored the effects of species groups, specific gravity, plot size and definition of maximum density in mixed-species forests of the northeastern United States (New England and New York). The comparison of the relative density measures considered here resulted in conclusions similar to Curtis (1970) in that the choice among the measures is, in part, a matter of available information and convenience of computation. The cluster analysis implied measures form clusters based on the lumping viewpoint versus splitting viewpoint. Rater agreement analysis, used as a novel method of comparing relative density models, suggests that Ducey and Knapp (2010) density estimates fall in between other density model estimates and thus if a single relative density model needs to be used in this region the Ducey and Knapp one seems most appropriate. Although on average FOXDEN2.1 (Desmaris, 2001) and Stout and Nyland (1986) provide higher density estimates than Ducey and Knapp (2010) and Woodall et al. (2006) when density estimates are assigned to categories, Ducey and Knapp (2010) and Woodall et al. (2006) consistently place more plots in higher density categories than FOXDEN2.1 (Desmaris, 2001) and Stout and Nyland (1986). The small FIA plot size used to estimate model coefficients may explain why Ducey and Knapp (2010) and Woodall et al. (2006) behave this way