Estimating and predicting bark thickness for seven conifer species in the Acadian Region of North America using a mixed-effects modeling approach: comparison of model forms and subsampling strategies

• Published in: European journal of forest research 2011-03, Vol.130 (2), p.219-233
• Main Authors: Li, Rongxia, Weiskittel, Aaron R.
• Format: Article
• Language: English
• Subjects: Bark; Biomedical and Life Sciences; Coniferous trees; Estimating techniques; Forestry; Life Sciences; Mathematical models; Original Paper; Plant Ecology; Plant growth; Plant populations; Plant Sciences; Trees
• ISSN: 1612-4669; 1612-4677
• DOI: 10.1007/s10342-010-0423-y

In many situations, information on stem diameters inside bark (dib) are more desirable than on diameters outside bark (dob). However, obtaining dib measurements is usually expensive, time-consuming, and prone to significant measurement errors when done on standing trees. Many bark thickness equations have been proposed to estimate the dibs of standing trees. In this study, we compared several commonly used bark thickness equations for seven conifer species in the Acadian Region of North America. Mixed-effects modeling techniques were employed to fit linear and non-linear bark thickness equations. We found the equation proposed by Cao and Pepper (South J Appl Forestry 10:220–224, 1986; Eq. 5) performed significantly better than other equations for most of our study species. The Cao and Pepper (South J Appl Forestry 10:220–224, 1986 ) equation is a function of dob, relative height in the stem, tree height, and the ratio of dib to dob at breast height. The mean absolute bias was found to be reduced up to 74% compared with using a fixed ratio approach employed in the widely used Northeastern variant of the Forest Vegetation Simulator (FVS-NE) growth and yield model. Leave-one-out cross validation was further performed to determine the location of suitable prior measurements in the prediction process for three of the most well-behaved equations. Results show that no unified prior measurement can provide best predictive abilities across all species as the choice of prior dib measurements depends on both species and bark thickness equations.