The agony of choice: different empirical mortality models lead to sharply different future forest dynamics

Abstract

Dynamic models are pivotal for projecting forest dynamics in a changing climate, from the local to the global scale. They encapsulate the processes of tree population dynamics with varying resolution. Yet, almost invariably, tree mortality is modeled based on simple, theoretical assumptions that lack a physiological and/or empirical basis. Although this has been widely criticized and a growing number of empirically derived alternatives are available, they have not been tested systematically in models of forest dynamics. We implemented an inventory-based and a tree-ring-based mortality routine in the forest gap model ForClim v3.0. We combined these routines with a stochastic and a deterministic approach for the determination of tree status (alive vs. dead). We tested the four new model versions for two Norway spruce forests in the Swiss Alps, one of which was managed (inventory time series spanning 72 years) and the other was unmanaged (41 years). Furthermore, we ran long-term simulations (-400 years) into the future under three climate scenarios to test model behavior under changing environmental conditions. The tests against inventory data showed an excellent match of simulated basal area and stem numbers at the managed site and a fair agreement at the unmanaged site for three of the four empirical mortality models, thus rendering the choice of one particular model difficult. However, long-term simulations under current climate revealed very different behavior of the mortality models in terms of simulated changes of basal area and stem numbers, both in timing and magnitude, thus indicating high sensitivity of simulated forest dynamics to assumptions on tree mortality. Our results underpin the potential of using empirical mortality routines in forest gap models. However, further tests are needed that span other climatic conditions and mixed forests. Short-term simulations to benchmark model behavior against empirical data are insufficient; long-term tests are needed that include both nonequilibrium and equilibrium conditions. Thus, there is the potential to greatly improve the robustness of future projections of forest dynamics via more reliable tree mortality submodels.

Climate change‐driven extinctions of tree species affect forest functioning more than random extinctions

“…Rasche, Fahse, Zingg, and Bugmann () tested the accuracy of the diameter–height relationship in the same locations and projected an increase in productivity in subalpine and upper montane forests and a decrease in the lower montane and colline zones, as a consequence of the direct effects of climate change. Bircher, Cailleret, and Bugmann () evaluated the mortality algorithm against growth‐and‐yield as well as forest reserves data from Switzerland, obtaining fair to excellent agreement of plot‐level data for three of four empirical mortality models. Mina, Martin‐Benito, Bugmann, and Cailleret () tested the drought response of the model extensively against empirical data from sixteen sites along a moisture gradient from Central Spain to the Swiss Alps.…”

Why do trees die? Characterizing the drivers of background tree mortality

“…, Bircher et al. ). Such models typically assume that probability of tree death can be understood in terms of two broad classes of death: ambient mortality and vigor mortality, with vigor mortality representing an inverse relationship between tree growth and mortality probability (Shugart , Botkin , Bugmann ).…”

Does one model fit all? Patterns of beech mortality in natural forests of three European regions

“…, Bircher et al. ). In particular, dynamic vegetation models (DVMs) at stand, landscape, and global scales, which are a key tool to quantify future changes of forest ecosystems, typically include theoretical mortality algorithms that lack mechanistic and/or empirical justification.…”