Modelling mixed forest growth:: a review of models for forest management

Most forests today are multi-specific and heterogeneous forests ('mixed forests'). However, forest modelling has been focusing on mono-specific stands for a long time, only recently have models been developed for mixed forests. Previous reviews of mixed forest modelling were restricted to certain categories of models only and were generally not considering application and suitability. The purpose of this paper is to give an overview of the models designed for or applied to modelling mixed forest growth and dynamics and to review the suitability of the different model types according to their intended purposes. The first part of the paper gives an overview of previous classifications, after which a new and overall classification scheme is presented. Next, the characteristics of the six modelling approaches that were distinguished are described: distance-dependent stand models, distribution models, average tree models, distance-dependent tree models, distance-independent tree models and gap models. All, except gap models, are close to mono-specific stands modelling approaches. The second part of the paper describes the main applications of these modelling approaches and presents a critical analysis of their suitability. Applications can be separated between growth and yield studies and forest dynamics simulation studies. Attention must be paid to recruitment sub-models, which appear to be inadequate in many models, but which highly influence the simulation outcome. All types of model were used as management tools. Stand level simulations fit the yield data better than tree level simulations, as a result of cumulated model errors from tree to stand level. However, tree level approaches seem most appropriate to understand stand growth as affected by competition between individuals of different species. Forest dynamics were mostly modelled using distribution models, gap models and distance-dependent tree models. The latter appeared to be less suitable because of the difficulties in modellin 3D stand structure over large periods and areas. Gap models could be applied to larger areas and time periods than distribution models, especially when they included detailed descriptions of the ecological functioning of the ecosystem. In sum the empirical models appeared more accurate in their predictions than mechanistic models, but they are highly dependent on the data used for parameterisation. That makes them unsuitable for extrapolation to other systems or conditions. Although mechanistic models can also be misused, adding mechanistic approaches to empirical observations is necessary to model the growth and dynamics of complex forest systems. (C) 2002 Elsevier Science B.V. All rights reserved.