Bayesian inference of biomass growth characteristics for sugi (C. japonica) and hinoki (C. obtusa) forests in self-thinned and managed stands

Abstract

Background: Forests are an important sink for atmospheric carbon and could release that carbon upon deforestation and degradation. Knowing stand biomass dynamic of evergreen forests has become necessary to improve current biomass production models. The different growth processes of managed forests compared to self-managed forests imply an adaptation of biomass prediction models. Methods: In this paper we model through three models the biomass growth of two tree species (Japanese cedar, Japanese cypress) at stand level whether they are managed or not (self-thinning). One of them is named self-thinned model which uses a specific self-thinning parameter α and adapted to self-managed forests and an other model is named thinned model adapted to managed forests. The latter is compared to a Mitscherlich model. The self-thinned model takes into account the light competition between trees relying on easily observable parameters (e.g. stand density). A Bayesian inference was carried out to determine parameters values according to a large database collected. Results: In managed forest, Bayesian inference results showed obviously a lack of identifiability of Mitscherlich model parameters and a strong evidence for the thinned model in comparison to Mitscherlich model. In self-thinning forest, the results of Bayesian inference are in accordance with the self-thinning 3/2 rule (α=1.4). Structural dependence between stand density and stand yield in self-thinned model allows to qualifying the expression of biological time as a function of physical time and better qualify growth and mortality rate. Relative mortality rate is 2.5 times more important than relative growth rate after about 40 years old. Stand density and stand yield can be expressed as function of biological time, showing that yield is independent of initial density. Conclusions: This paper addressed stand biomass dynamic models of evergreen forests in order to improve biomass growth dynamic assessment at regional scale relying on easily observable parameters. These models can be used to dynamically estimate forest biomass and more generally estimate the carbon balance and could contribute to a better understanding of climate change factors.