Characterizing Stand and Biomass Tables from Diameter Distribution Models: A Case Study for Mixed-Hardwood Forests in Eastern Tennessee, USA

Abstract

Accurately assessing forest structure and productivity is critical to making timely management decisions and monitoring plant communities. This study aims to evaluate the prediction accuracy of site-level stand and biomass tables from the diameter distribution models. The efficacy of the single Weibull function and two finite mixture models was compared for six species groups on three mixed-hardwood sites in eastern Tennessee, USA. To evaluate model performance, two types of stand/biomass tables were generated. The first type was constructed from all species on a given site (site-specific), whereas the second type was built for a single species from all sites (species-specific). Results indicate that both types of stand and biomass tables were consistently well quantified by the two-component mixture model in terms of goodness of fit, parsimony and robustness. The two-component mixture model better characterized the complex, multimodal diameter distributions than the single Weibull model, which underpredicted the upper portion of the distributions. The three-component model tends to overfit the data, which results in lower prediction accuracy. Among the three models examined, the two-Weibull mixture model is suggested to construct site-level stand/biomass tables, which provides more reliable and accurate predictions to assess forest structure and product class. Study Implications Compared to pine monocultures, diameter distribution models for upland mixed-hardwood forests in the Southeastern United States have not been widely explored. Mixed-hardwood forests not only supply high-quality timber for domestic and international uses, but also provide various ecosystem services and essential habitats for wildlife. The finite mixture model has been proposed for characterizing the irregular forms of diameter distribution curves, but the reliability of this method has not been explicitly examined for a wide variety of species. This study provided insights for natural resources managers to select appropriate models when modeling stand and biomass tables for mixed-hardwood forests.

mixR: An R package for Finite Mixture Modeling for Both Raw and Binned Data

“…mixR is actively used by researchers and practitioners in various fields (Buckland et al, 2021;Büchel & Corman, 2021;Jung et al, 2020;Korne et al, 2021;Ogana, 2020;Sylvestre et al, 2020;S.-I. Yang et al, 2021;Z.…”