Models for prediction of mortality in even-aged forest

Models for predicting mortality in even-aged stands were developed. The models rely on data from the Norwegian National Forest Inventory, and were designed for use in large-scale forestry scenario models. A two-step modelling strategy was applied: (1) logistic regression models predicting the probability of complete survival occurring, and (2) multiplicative regression models for stem number reduction and diameter calibration. A joint model for all species predicting the probability of survival occurring on a plot was developed. Separate models for forests dominated by spruce, pine and broadleaved trees were developed for stem number reduction, while no appropriate models for diameter calibration were found. The phenomenon mortality is a stochastic, rare and irregular event, and this was reflected as low R-2 in the models. However, the model performance appeared logical and the results of validations based on independent data were reasonably good, i.e. the presented models may be applied to large-scale forestry scenario analyses. With new rotations of permanent sample plot measurements, the models should be evaluated and, if necessary, revised.