A generalized mixed logistic model for predicting individual tree urvival probability with unequal measurement intervals

Abstract

Based on a generalized linear mixed-model approach, three survival probability models were developed for lodgepole pine (Pinus contorta var. latifolia Engelm.) trees in boreal mixedwood stands in Alberta, Canada. Besides a traditional logistic model and a revised form of that model, a new logistic model was proposed to handle unequal measurement intervals typical of repeatedly measured growth-and-yield data. Various stand and tree characteristics were examined for their contribution to model improvement. Annual diameter increment, the ratio of basal area of larger trees to total stand basal area, and the basal area of larger deciduous trees were found to be significant predictors and were included in the final mixed models for predicting the probability of a tree surviving the coming growth period. Two random parameters were incorporated in the traditional and new models, one added to the intercept of the linear predictors and the other added to the parameter for annual diameter increment. One random parameter was added to the parameter for annual diameter increment in the revised traditional model. The new model provided improved stand density predictions at the population and plot levels over the two forms of the traditional model for a new data set. © 2013 by the Society of American Foresters.