Modelling the change in aspen species composition in boreal mixedwoods

Abstract

The dynamics of aspen (Populus tremuloides Michx.) species composition (SC, defined as the ratio of aspen basal area to total stand basal area) in boreal mixedwoods were modelled in this study using the difference equation method, based on datasets collected from repeatedly measured permanent sample plots (PSPs) across Alberta. The aspen SC measured at time 1 was taken as a key predictor variable to project SC values at future ages. Since site quality was found to impact the aspen SC, site index (SI) was incorporated into the model. To correct the autocorrelation and heteroskedasticity problems associated with PSPs data, the non-linear mixed-model technique was applied to accommodate the variance-covariance structure of the error terms and to estimate model parameters. The final model was evaluated on an independent dataset collected from a different region in Alberta, based on a number of statistical measures including a goodness of prediction index (GOPI) from forward and backward projections, and examinations of residual and studentized residual plots. The low prediction bias and high GOPI value obtained from forward and backward projections on the validation data suggest that the model fitted the data well and can be reliably applied to predict changes in aspen SC in boreal mixedwoods across Alberta. The model showed a decline in the SC after year 30, but the decline was steeper in sites of low SI. The model can be used for modelling the transitions of forest compositions in boreal mixedwood forests. © Institute of Chartered Foresters, 2008. All rights reserved.