Predicting tree mortality from diameter growth: a comparison of maximum likelihood and Bayesian approaches

Ecologists and foresters have long noted a link between tree growth rate and mortality, and recent work suggests that interspecific differences in low growth tolerance is a key force shaping forest structure. Little information is available, however, on the growth-mortality relationship for most species. We present three methods for estimating growth-mortality functions from readily obtainable field data. All use annual mortality rates and the recent growth rates of living and dead individuals. Annual mortality rates are estimated using both survival analysis and a Bayesian approach. Growth rates are obtained from increment cores. Growth-mortality functions are fitted using two parametric approaches and a nonparametric approach. The three methods are compared using bootstrapped confidence intervals and likelihood ratio tests. For two example species, Acer rubrum L. and Cornus florida L., growth-mortality functions indicate a substantial difference in the two species' abilities to withstand slow growth. Both survival analysis and Bayesian estimates of mortality rates lead to similar growth-mortality functions, with the Bayesian approach providing a means to overcome the absence of long-term census data. In fitting growth-mortality functions, the nonparametric approach reveals that inflexibility in parametric methods can lead to errors in estimating mortality risk at low growth. We thus suggest that nonparametric fits be used as a tool for assessing parametric models.