Using a segmented logistic model to predict trees to be harvested in forest growth forecasts
Aim of the study: Predicting future harvested trees is a prerequisite to growth forecasts in managed forest stands. While harvest algorithms have been traditionally used, statistical harvest models could be an interesting alternative approach. The objective of this study was to fit statistical harvest models for different partial cutting treatments. Area of study: The study has been carried out in the province of Quebec, Canada. Material and methods: Data from provincial control survey were
... rol survey were used to fit harvest models for three different partial cutting treatments. A two-segment logistic modelling approach was used. The harvest models were designed to be compatible with the ARTEMIS growth simulator, which is currently in use in this province. Main results: The results showed that the probability of being harvested is different across the treatments and primarily depends on tree diameter at breast height and species. In selection cutting treatments in particular, trees close to the merchantable limit (e.g., 23 cm in this study) tended to be less frequently harvested than those with smaller or larger diameters, yielding a sinusoidal pattern that was well captured by the segmented approach. Research highlights: Although the models were of average accuracy as indicated by fit statistics, they made it possible to compare different scenarios in terms of productivity and rotation length when coupled with the ARTEMIS growth simulator. Moreover, compatibility requirements between the simulator and the harvest models appeared to be a major limitation in some cases.