Estimating the probability of survival of individual shortleaf pine (Pinus echinata Mill.) trees

An individual tree survival model was developed for shortleaf pine ( Pinus echinata Mill.) trees. Prediction of the probability of survival of an individual tree is essential when considering growth and yield of a stand. Data for this study were from more than 200 permanently established plots on even-aged natural shortleaf pine stands that were located in the Ozark and Ouachita National Forests. Plots were established during the period of 1985-1987. Plots have been remeasured every 4, 5 or 6 years, and individual tree survival or mortality was recorded at each measurement. These plots were selected to represent a range of ages, densities and site qualities. Logistic regression was used to find the best sets of significant predictor variables in which the response variable was a binary variable `1' for the survival tree and `0' for the mortality tree. Significant variables found in predicting the survival were mid-period basal area per acre (Mid-BA), inverse of ratio of quadratic mean diameter to diameter at breast height (DBH) (DRINV), their interaction and square of DBH. Parameters of the logistic equation were estimated using iteratively re-weighted nonlinear regression. A nonlinear mixed-effects approach was also applied to investigate the plot level effect on the model. Model performance was evaluated using Chi-square goodness-of-fit test, and it was found that the model worked better while estimating the parameters using iteratively reweighted non linear regression than with the nonlinear mixed model. This individual tree survival model can be used to predict the annual survival rate of individual trees of even-aged shortleaf pine forests located in Ozark and Ouachita National Forests and in the surrounding regions.