Introduction
Developers of greenfield plantations need accurate predictions of forest growth and yield but inevitably face a lack of data to develop empirical models as they wait for their forests to mature. Forest managers might hope to borrow models developed in similar plantations, but this may not be feasible when introducing a new species or forest systems to new regions. Furthermore, published models may not provide the detailed product level outputs required for operational and financial planning. As plantations mature intermediate static models are often developed to meet immediate needs and these can be replaced with more accurate operational models as plantations approach rotation age.
Developing Models
Form Ghana Ltd are a teak plantation grower located in Ghana’s Ashanti region and focussed on producing high-quality teak logs for export and poles for the domestic market. Margules Groome was engaged by Form Ghana to develop a teak growth and yield modelling framework to predict forest growth and product outturn. A permanent sample plot network including repeat measurements was used for growth model development. All growth model components used the projection form allowing forest inventory measurements to be incorporated in a dynamic manner. Stand-level models for basal area growth, height growth, and mortality were developed. A variety of modelling approaches were trialled for all variables of interest. For modelling forest height, the generalised algebraic difference approach (GADA) was adopted. GADA allows the derivation of dynamic height-age equations based on static height-age curves. Tested models included projection forms of the Chapman-Richards, Korf and Hossfeld IV equations as these have been widely used to model the relationship between height and age in plantation trees. Anamorphic, polymorphic, single asymptote, and polymorphic asymptote forms of the models were trialled. All models were developed in the R statistical computing environment.
Stand-level mortality was modelled using the Clutter Jones equation[1], stand height increment was modelled using a polymorphic, single-asymptote projection form of the Chapman-Richards equation[2], and basal area increment was modelled using a polymorphic projection of the Korf growth equation[3]. Extensive cross validation suggested that models were unbiassed and had good accuracy levels across the range of the fitting dataset.
To predict upper stem diameters and log volumes a taper function was also fitted. The trigonometric variable form taper equation originally developed for Eucalyptus trees in natural forests[4] was selected. The taper model is stable in specification yet flexible in its ability to fit to data from a wide range of tree species and sizes with different stem forms. The base function is constructed from trigonometric volume-ratio equations following the geometry of a tree stem and the specification for the exponent is based on the Fourier transformation. By allowing both the base function and the exponent to vary with tree size, the model has a high degree of flexibility for depicting stem shapes. The taper function has been parameterised for a wide range of tree species and is generally noted to be one of the most reliable taper models. To the best of our knowledge, this was the first parameterisation of this model to teak plantations. The model was parameterised using sectional measurement data collected from 1 478 teak trees and can be used to estimate over bark diameter for any point on a stem assuming that a measure of diameter at breast height over bark (DBHOB) and total tree height are available. Over bark volume can also be estimated for any stem or between any two points on the stem using integration of the taper function. The taper equation is highly flexible, stable, and fits well to the sectional measurement dataset used for model development. The model appears to extrapolate well to teak trees of most sizes within the tree size range that are likely to be encountered by Form Ghana.
Implementation
Failing to consider how the mathematical models developed will be implemented commonly limits the usefulness of model outputs to fulfill forest manager’s information needs. We opted to implement our models in the commercial yield prediction software YTGen (Silmetra, New Zealand). This was the first use of YTGen for growth and yield prediction in teak plantations. YTGen is a robust platform for forest yield prediction that includes a growth engine, thinning modelling framework, and an optimal log bucking algorithm designed to output yield tables as specified by the user.
To ease implementation all developed models were reversible and path-invariant. Reversible means the model can be run backwards as well as forwards within reasonable ranges. Path invariance means that growing a plot from age 5 to age 10 will give the same results as growing the same plot from age 5 to 6 and then on to age 10 in steps. These behaviours mean that prediction is robust to inventory timing and step size and that predictions for past states can be outputted if required.
Unlike many alternatives YTGen maintains an internal measured tree list that is adjusted using the growth model components. Stand-level growth predictions are scaled to update individual tree dimensions and survival probabilities. This negates the need to model DBH distributions as inventory measured tree lists are maintained and grown throughout the simulation.
YTGen users can include descriptions of features on standing trees that impact product recovery. This might include branch sizes, stem sweep, or defects. For teak trees merchantable height is an important concept that can be measured on plot trees. Merchantable height is typically measured at the point on the stem where the tree forks or where there is a significant defect that results in stem deformation or a reduction in diameter. Due to the teak log market requirements all volume above the merchantable height is typically not recovered. YTGen facilitates modelling of this constraint.
With our biometric models embedded in YTGen Margules Groome worked with Form Ghana to develop a new set of yield tables for their estate to support woodflow modelling and to gain a better understanding of vital future cashflows. The framework used by Margules Groome is robust and transferable to other species and regions with a reasonable PSP dataset.
If you are interested in taking the next step towards improving your growth and yield prediction, then please get in contact with our technical team.
[1] Clutter, J.L. and Jones, E.P. (1980) Prediction of growth after thinning in old-field slash pine plantations. Research Paper SE-217, USDA Forest Service
[2] Manso, R., Davidson, R., McLean, JP. (2022) Diameter, height and volume increment models for improved Sitka spruce in Great Britain. Forestry: An International Journal of Forest Research, 95 (3) pp 391-404 https://doi.org/10.1093/forestry/cpab049
[3] Zeide, B. (1993) Analysis of Growth Equations. Forest Science 39 (3). pp 594-616
[4] Bi,H. (2000) “Trigonometric Variable-Form Taper Equations for Australian Eucalypts.” Forest Science 46(3), pp 397-409