Modelling sapling density for sugar maple-dominated mixed stands of eastern Canada

Abstract

Saplings (tree species whose diameter at breast height ranges from 1.1 to 9.0 cm) are important in forest growth and development. Their abundance and density can lead to specific forest successional trajectories and ecosystem characteristics. Yet, the consideration of saplings in forest management planning is a relatively new topic and is still rarely included in forest growth models. We developed sapling density models for 10 species groups with the objective of providing forest managers with additional tools to support the development of more precise prediction systems. We used dendrometric and environmental variables to model sapling density and species-wise density ratios. We evaluated Poisson and gamma regressions for the modelling of sapling density and zero-inflated logistic regressions to model species ratios. We used repeated measurements (from 1982 to 2019) from permanent sample plots located in hardwood forests in northeastern North America. Our results show that the gamma regression was superior to the Poisson regression. The cross-validated mean bias using gamma regression was 55 stems·ha−1 with a relative percentage error of 2.5% and an R2 of 0.43. The species-wise sapling density ratio model had an overall R2 of 0.93, and the species-wise mean R2 ranged between 0.90 and 0.96. Among the examined model covariates, stem density of merchantable-sized trees and latitude were significant in both models. We believe that the models we developed can be useful for forest management planning and sustainable merchantable production.