Comparative review of models for sustainable yield in indigenous forests

Abstract

A review of the forest models developed and applied by Timberlands West Coast Ltd (TWCL) and Landcare Research Ltd (LRL) has been carried out. The models were reviewed on the basis of default settings for red beech in the Maruia Working Circle. After identifying the similarities and differences between the two models, a sensitivity analysis was carried out to quantify the impact of any differences on model outputs, in particular stand structure and harvest yield. A sequence of model variants was developed and run, starting with the TWCL default model and ending with the LRL default model. Each variant in the sequence differed in only one factor, thereby allowing quantification of the relative sensitivity of model outputs to that factor. The review focuses on the impact of differences between the models in terms of mathematical formulation, input data and assumptions. However, it excludes any analysis of the appropriateness and relative merit of the different mathematical formulations, input data and underlying assumptions. Although these are important considerations they are beyond the scope of this review. Both models can be categorised as Stand Class Models and use the Stand Table Projection Method to project the growth of a stand by simulating the growth of classes of trees. This is a commonly used approach for modelling, particularly for uneven-aged forests. The differences in mathematical formulation between the LRL model and the TWC model are: 1. Mortality is included in the transition coefficients in the LRL model whereas it is treated as an absolute reduction in the TWCL model. 2. The transition coefficients have a different structure because of different assumptions about the distribution of trees within a size class and the residence time of trees in each class. Incorporating mortality within the transition coefficients rather than as an absolute reduction has a minimal impact on model outputs. The use of LRL transition coefficients, without any other model changes, has a major impact on model outputs. However, once mortality is adjusted to reflect the different coefficients, model outputs for the LRL approach are similar to model outputs for the TWCL approach. Another difference between the models is that the LRL model allows for compensatory growth (Version 1.1) and mortality (Version 2). The model includes functions which allow tree growth rates and mortality to vary in response to changes in stand basal area. Invoking these functions can have a major impact on model outputs. Both models have the same initial tree size distribution. There are minor differences between the tree growth rates and the recruitment rates specified in the two models. These differences have a negligible impact on model outputs. The models (in terms of default settings) differ in the relationship between harvest and mortality. A fundamental assumption of the TWCL model "is that mortality is subsumed into harvest through the careful selection for harvest of trees already prone to direct mortality or mortality by association with dying or falling trees". In contrast, an underlying assumption of the LRL model is that "logging imposes mortality that is largely additional to natural mortality in any one year". These differences have a major impact on model output.