How Random Are Predictions Of Forest Growth? The Importance Of Weather Variability

Quantifying the output uncertainty and tracking down its origins is key to interpreting the results of modelling studies. We performed such an uncertainty analysis on the predictions of forest growth and yield under climate change. We specifically focused on the effect of the interannual climate variability. For that, the climate years in the model input (daily resolution) were randomly shuffled within each 5-year period. In total, 540 simulations (10 parameter sets, nine climate shuffles, three global climate models, and two mitigation scenarios) were made for one growing cycle (80 years) of a Scots pine (Pinus sylvestris L.) forest growing in Peitz, Germany. Our results show that, besides the important effect of the parameter set, the random order of climate years can significantly change results such as basal area and produced volume, as well as the response of these to climate change. We stress that the effect of weather variability should be included in the design of impact model ensembles and in the accompanying uncertainty analysis. We further suggest presenting model results as likelihoods to allow risk assessment. For example, in our study, the likelihood of a decrease in basal area of >10% with no mitigation was 20.4%, whereas the likelihood of an increase >10% was 34.4%.