Modeling sensitivity to climate change and estimating the uncertainty of its impact: A probabilistic concept for risk assessment in forestry.

Many uncertainties emerge dealing with future climate conditions and their possible impacts on forests. In this paper we suggest a probabilistic approach extending existing ensemble species distribution modeling concepts by addressing important sources of uncertainty. We exemplify our approach using European beech as the target tree species and bioclimatic predictors derived from WorldClim data. Model parameter uncertainty is represented by 1000 parameter samples from a Bayesian generalized linear model. Climate change impact ( CCI ) is based on 63 different climate model outputs using four RCP-scenarios ( R epresentative C oncentration P athways) in addition to the parameter uncertainty. The proposed difference of the probability of occurrence p occ to a predefined threshold, allows for evaluation of parameter uncertainty as well as for the uncertainty of future climate and describes the changing niche position. Further, we suggest the probability that the probability of occurrence exceeds a predefined threshold ( p exc ) as a metric for the distance of a site to the niche edge. These metrics are unambiguously determinable, intuitive and evident. Stands at a central, a marginal and an intermediate niche position are taken to exemplify deduction and application of p occ and p exc . A regional exercise shows how a map of p exc may support forest management planning and decision making under severe uncertainty. A key advantage of our novel metrics is the reformulation of common species distribution model (SDM) outputs in terms of risk thereby accounting for two important sources of uncertainty. [ABSTRACT FROM AUTHOR]