Ninety-five years of observed disturbance-based tree mortality modeled with climate-sensitive accelerated failure time models

Abstract

Modeling disturbance-based tree mortality is becoming increasingly important in the discussion of how to adapt forests to climate change and to preserve their ecosystem services and mitigate the risk of economic losses. In this study, we fitted species-specific interval-censored Accelerated Failure Time models for five major tree species to derive the influence of climate, soil, silvicultural measures, stand and tree characteristics on survival times. We coded all disturbance-based mortality causes as events and analyzed 473,501 individual trees distributed across 2248 long-term (1929–2014) forest growth and yield plots in southwestern Germany. We observed different survival probabilities among tree species with Douglas-fir having the lowest survival probability at age 100 years, followed by Norway spruce and Silver fir. Contrastingly, beech and oak had survival probabilities above 0.98 at age 100 years. Most important factor influencing these survival times was climate. Higher summer temperature shortens the survival time of beech, Silver fir and oak, while Norway spruce suffers more from warmer and wetter winters. Beside climatic factors, base saturation showed a significant positive relationship to survival time for all investigated tree species, except for Norway spruce, which had shorter survival times with increasing cation exchange capacity of the soil. Additionally, short-term effects of destabilization after thinning were found. In conclusion, favoring broadleaved tree species, avoiding heavy thinning in older stands and limiting tree age reduce the probability of disturbance-based tree mortality. However, some of the effects found that cause-unspecific mortality modeling has limited potential to describe the mortality–climate change relation.