Towards non-destructive estimation of tree age

Abstract

Accurate tree age information is required in many contexts ranging from nature conservation to forest science and management. Currently available methods for tree age estimation are either destructive or often inaccurate, the latter mostly because they do not tap the full potential of available data and knowledge on tree growth. We compared two new approaches for tree age estimation based on nonlinear age–diameter relationships to a traditional polynomial approach. The nonlinear approaches were based on repeated diameter measurements. One of them included environmental covariates (slope, elevation, aspect, water holding capacity and a drought index) based on the fixed effects of a mixed-effects model. The accuracy of the approaches was evaluated for 237 oaks (Quercus spp.) growing along an environmental gradient in Switzerland and comprising ages from 23 to 284years. The potential of the nonlinear approach with covariates was assessed by additionally including the random effects of the mixed-effects model.The nonlinear approach with covariates and the polynomial approach were of similar accuracy except for extreme sites, where the polynomial approach performed better. The nonlinear approach without covariates was least accurate. Additionally including the random effects in the nonlinear approach with covariates strongly improved the age estimates and reduced the relative errors below 40% for 98% of the trees.Including repeated diameter measurements and environmental covariates led to similarly accurate age estimates as the traditional polynomial approach. However, the accuracy of the nonlinear approach with covariates has a high potential for further improvements. Additionally, the nonlinearity and the site information that is explicitly included allow for applications beyond currently represented ages and sites. This transferability and the potential for extrapolation obviate the need for model fitting in further applications, making it entirely non-destructive, which is a large advantage over the polynomial approach, which requires new fitting for new sites. Thus, applying the nonlinear approach with covariates is highly suitable e.g. in protected forests, where destructive age determination is not allowed.