On the Ecology of Mountainous Forests in a Changing Climate: A ...

174 Chapter 7
7.4 Model validation
The three models FORECE, FORCLIM-E/P and FORCLIM-E/P/S all produced plausible
species composition when applied at sites along a climatological gradient in the European
Alps (cf. Kienast & Kuhn 1989a,b). Based on these results alone, it would not be possible
to favour one of the models over the others, although the formulation of FORCLIM is
mathematically more rigorous, it depends to a larger extent on causal relationships, and it
is simpler. Only the systematic simulation studies performed in a climatological parameter
space spanned by the annual mean temperature and the annual precipitation sum revealed
that FORECE contains several unrealistic thresholds and that it produces unrealistic
species compositions in a larger fraction of this (T,P) space than FORCLIM.
There are two areas in the (T,P) space where both FORECE and FORCLIM encounter
major difficulties: (1) The warm-dry zone in central Alpine valleys as well as outside the
Alps, such as in large areas of Germany and France, where the models fail to simulate realistic
species compositions along drought gradients; (2) The insubrian and mediterranean
zones, where the models fail to simulate the occurrence of drought. These shortcomings
may be especially important for sites on the Swiss Plateau, where climatic change could
lead to such conditions (cf. Gyalistras et al. 1994). Hence further research on soil water
balance and the ecological effects of drought should be conducted.
The study in the (T,P) space was confronted with a serious methodological problem: The
hypothesis on the dominating species in this space (Rehder 1965, Ellenberg 1986) lacks
an exact quantification. Thus, the comparison of the simulated species compositions with
phytosociological data was possible on a qualitative basis only, and many aspects of the
simulation results had to be ignored although they could give important indications on
deficiencies of FORECE and FORCLIM. The comparisons of model output with phytosociological
data performed along an altitudinal and a latitudinal gradient in the European
Alps and in eastern North America, respectively, were faced with similar problems: There
is a mismatch between the qualitative nature of phytosociological descriptions of nearnatural
forests (e.g. Ellenberg & Klötzli 1972, Küchler 1975) and the quantitative data
obtained from forest gap models.
The simulation results obtained along a latitudinal gradient in eastern North America,
which differs from European conditions both climatologically (i.e. larger continentality)
and ecologically (i.e. different species), are encouraging for two reasons: First, the performance
of the unmodified FORCLIM model was realistic at many locations. Second,