Classification and regression tree based survival analysis in oak-dominated forests of Missouri's Ozark highlands

Abstract

Tree survival or mortality is a stochastic process and highly variable over time and space. Many factors contribute to this process, including tree age, tree size, competition, drought, insects, and diseases. Traditional parametric approaches to modeling tree survival or mortality are often unable to capture this variation, especially in natural, mixed-species forests. We analyzed tree survival in Missouri Ozark oak forests using a combination of classification and regression tree (CART) and survival analysis of more than 35 000 trees with DBH >11 cm measured four times between 1992 and 2002. We employed a log-rank test with CART to classify trees into seven disjoint survival groups and used a nonparametric Kaplan–Meier (product limit) method to estimate tree survival rate and construct confidence intervals for each survival group. We found that tree species, crown class, DBH, and basal area of larger trees were the variables most closely associated with differences in tree survival rates. In these mature oak forests, mortality for the red oak species group was three to six times greater than for the white oak, hickory, or shortleaf pine species group. The results provide practical information to guide development of silvicultural prescriptions to reduce losses to mortality.