Modeling Tree Mortality Using the Cox Proportional Hazards Model Based on Data Derived from Permanent Plots of Cryptomeria japonica Planted Forest.

Abstract

The objective of this study was to develop a tree mortality prediction model using a semi-parametric survival model based on longitudinal data derived from permanent plots. Here, tree mortality data derived from four permanent plots of Cryptomeria japonica planted forest and the Cox proportional hazards model was applied to this data using two time-independent covariates and seven time-dependent covariates. Relative spacing, which is a time-dependent covariate, was selected as the best single predictor for the tree mortality model, while, site index, which is a time-independent covariate, was better single predictors than other time-dependent variables. Through a model selection procedure, the tree mortality model that included site index (time-independent) and stand density and cumulative number of decreased trees (time-dependent) was selected as the best model.