Quantifying bark thickness and bark volume with alternative modeling procedures for eight species in the southeastern US

Abstract

Bark thickness is one of the important tree characteristics to assess plant survival following silvicultural treatments or natural disturbances. With an increased number of catastrophic wildfires under changing climate, a reliable approach for modeling bark thickness and bark volume is critical to monitor forest health and quantify carbon stocking in sustainable forest management. The objective of the study was to compare the efficacy of three alternative quantifying approaches and statistical models for both bark thickness and volume. In approach I, bark thickness is quantified from the differences of inside- and outside-bark diameters, which are predicted with two separate stem taper functions. The second approach combines a stem taper function and a bark thickness model, while in the third approach, bark thickness is quantified via a single predictive model with bark thickness at breast height included. The statistical models examined include segmented polynomial regression models (SEG), variable-exponent models (VEM) and generalized additive models (GAM). A total number of 71,187 observations of 2,932 trees were queried from the LegacyTree database, including eight ecologically and economically important species in the southeastern US.Results show that all combinations of approaches and models can quantify bark thickness and bark volume without bias for all eight species examined. Approach III yields higher accuracy and precision levels than approaches I and II. The performance of the models varies among quantifying approaches and species. GAM and VEM generally produce a more reliable quantification than SEG. Although GAM may not be the optimal model for a given species, such a model can provide preliminary quantitative information about bark thickness and bark volume when there is no existing model for the target species. The results of this work provide additional insights for forest ecologists and practitioners to select appropriate modeling approaches for quantifying bark thickness and volume.