Models of forest dynamics are pivotal for assessing future forest dynamics under the impacts of changes in climate and management practices. Such models must necessarily include a representation of tree growth, mortality, and regeneration. The empirical foundations for formulating tree growth are quite solid, but quantitative knowledge for developing robust models of tree mortality is scarce. In the context of the COST Action PROFOUND (“Towards robust projections of European forests under climate change”), 15 Dynamic Vegetation Models (DVMs) were evaluated in terms of their sensitivity to different, equally plausible formulations of tree mortality. The study included both „chronic“ („background“) mortality as well as mortality events induced by extreme conditions such as droughts.
The set of models included 8 DVMs at the stand scale, 4 at the landscape scale, and 3 at the global scale. While some models include empircally derived mortality models (e.g., based on inventory or tree-ring data), others are based on experimental data (e.g., drought experiments), whereas still others are based on theoretical reasoning alone (e.g., consideration of the simulated whole-plant carbon balance).
Each DVM was run with at least two alternative mortality algorithms. In a first step, model behavior was evaluated against past time series data. In a second step, the models were subjected to different scenarios of climate change for the 21st century.
Results/Conclusions
Most DVMs matched empirical data quite well, irrespective of the mortality formulation that was used. Based on the simulation results, the sensitivity of the models was classified, and the reasons for the different sensitivities were investigated. It became obvious that it is generally very difficult to assess the suitability of a mortality formulation based on past model behavior only. However, mortality algorithms that performed in a very similar manner when evaluated against past data were often found to lead to sharply different trajectories of future forest dynamics.
We conclude that it is indispensable to employ several alternative mortality formulations in DVMs when assessing future forest dynamics. Although this will increase the uncertainty of the simulation results because there are hardly any a priori reasons for favoring one alternative over the other, it precludes that decision makers draw erroneous conclusions based on seemingly clear simulated future trajectories.