Modelling dominant height growth including a rainfall effect using the algebraic difference approach

ABSTRACT

Background:

Estimating forest productivity is critical for effective management and site assessment. The dominant height is used to calculate the Site Index (SI), which is commonly used to assess forest productivity. In this study, an algebraic difference approach was used to develop a dominant height model incorporating the rainfall effect for Eucalyptus grandis x Eucalyptus urophylla ( E. Grandis x E. Urophylla). The dataset consists of 75 permanent sample plots ranging in age from 0.5 to 11 years, as well as 7 rainfall stations spread across plantations in Coastal Zululand, South Africa. Using fixed and mixed-effects in the predictor function, twelve candidate models were derived from the Bertalanffy-Richards, Lundqvist-Korf, McDill-Amateis, and Hossfeld growth functions. A continuous-time autoregressive error structure was used to account for serial autocorrelation in the longitudinal unbalanced data. Model fit statistics and graphical methods were used to evaluate the candidate models.

Results:

The addition of the rainfall effect increased model precision by 37%. The mixed-effects formulation produced 18% more precision when compared to similar models with all parameters fixed. Due to their compatibility with expected biological behaviour and good performance on validation data, mixed-effects models based on Lundqvist-Korf and McDill-Amateis functions were chosen as the final models.

Conclusion:

Unlike similar models that do not take rainfall into account, these models can capture the effects of severe rainfall conditions such as drought and can thus be used in short-rotation pulp forests with fluctuating rainfall.

Keywords:
Site index; algebraic difference approach; polymorphism; fixed-effects; mixed-effects; climatic effects

HIGHLIGHTS

The nonlinear mixed-effects technique was used to forecast dominant height over time for Eucalyptus grandis x Eucalyptus urophylla in Coastal Zululand, South Africa, taking into account changes in rainfall.

We looked into and contrasted nonlinear fixed and mixed-effects modeling methodologies. In comparison to the strictly fixed effects model, the final nonlinear mixed-effects models were 18% more precise.

The addition of the rainfall effect increased model precision by 37%.

The models developed in this study can be utilized in short-rotation pulp forests with varying rainfall to capture the effects of extreme rainfall conditions such as drought.