Improving the formulation of tree growth and succession in a ...

S. Schumacher et al. / Ecological Modelling 180 (2004) 175–194 189
structure well, but less so species composition (Hefti
et al., 1986); therefore, the estimated species distribution
patterns may not be very accurate. Second, the
simulated distributions (Fig. 4b,c) varied between simulation
runs due to the stochastic processes embedded
in the model. Still, it is evident that the model yields
spatial vegetation patterns under the harvesting regime
(Fig. 4b) that are similar to those of the current landscape
(Fig. 4a). Below, we discuss model behavior as
aggregated over the elevational gradient, and also at the
stand-scale (cf. Section 5.1.2).
The simulated landscape for the year 2000 a.d.
showed total biomass values of about 200 t/ha at lower
elevations and gradually decreasing values with altitude
(Fig. 5b). These values are similar to those
derived from forest inventory data (Hefti et al.,
1986; cf. Fig. 5a), and to values of 200–250 t/ha
(300–400 m 3 /ha) for managed spruce forests reported
by Leibundgut (1986). The simulation results match the
measured values at lower elevations quite accurately,
whereas at higher altitudes biomass seems to be somewhat
underestimated by the model. Also, the model
projects a rather linear decrease of total biomass between
1650 and 2100 m a.s.l., whereas the measured
data may point at a slightly non-linear decrease, but
these differences are not large.
The comparison between simulation results and inventory
data also suggests that the model correctly
represents the dominance of Picea abies below 2000
m a.s.l. However, the model seems to underestimate
the amount of Larix decidua biomass, whereas Pinus
cembra is over-represented, particularly at higher altitudes.
It is quite likely that these differences are due
to the fact that our management regime is not only relatively
coarse, but did not include wooded pastures,
where L. decidua is favored quite strongly over more
shade-casting trees such as P. abies or P. cembra (cf.
Walder, 1983). Larix decidua is an early successional
species that can co-exist or dominate only after disturbances
(Ott et al., 1997). Overall, we can conclude that
the simulated biomass and species distribution are in
good agreement with the available measured data.
The simulated forests under natural, unmanaged
conditions (Figs.4c and 5c) cannot be compared
directly with the inventory data (Figs. 4a and 5a);
however, literature data is available for this model test.
The simulation yielded dominance by three species:
Picea abies, Larix decidua and Pinus cembra (Fig. 4c,
Fig. 5c). These are the three main tree species of the
current forest cover (Walder, 1983) (Fig. 4a), and are
also the three main species of the potential natural
forest vegetation in this area (Landolt et al., 1986;
Zumbühl and Burnand, 1986; Ellenberg, 1996; Ott et
al., 1997). Compared to current conditions, L. decidua
and P. cembra forests are thought to have covered
larger areas before forest management began (Landolt
et al., 1986), and this is reflected in the simulation
results, where the share of these two species is higher
in Fig. 5c than in Fig. 5a, particularly at elevations between
1650 and 2100 m a.s.l. The simulated fraction of
biomass of L. decidua (Fig. 5c) is considerably higher
than under the management scenario (Fig. 5b), but still
somewhat smaller than currently observed in the study
area (Fig. 5a). Measurements of aboveground standing
volume in a P. abies virgin forest that is similar to
potential natural forests in the lower elevations of
our study area resulted in biomass values of about
300 t/ha (550 m 3 /ha) (Hillgarter, 1971). Simulated
values at lower elevations (Fig. 5c) are considerably
higher than under the management scenario (Fig. 5b),
but still somewhat smaller than this measured data.
The presence of a belt of L. decidua/P. cembra
forests above 2000 m a.s.l. agrees with literature
data (e.g., Landolt et al., 1986, Ellenberg, 1996). Its
biomass has been estimated to amount to 50–100 t/ha
(100–200 m 3 /ha) (Leibundgut, 1986); again, simulated
values are very similar, although somewhat lower
above 2150 m a.s.l. Finally, the simulated presence
of a strip of Pinus-Alnus ‘krummholz’ just below
the tree line is quite realistic (Landolt et al., 1986).
Similar experiments were conducted with a forest gap
model (Bugmann, 1999; Bugmann and Pfister, 2000).
They showed that the sequence of tree species and
forest types simulated along altitudinal gradients is
largely the result of competitive interactions, i.e. the
realized niche of the species (Fig. 5c) is only partly
dictated by the species’ autecological properties (their
fundamental niche; cf. Table 3). Therefore, we can
conclude that the simulated vegetation characteristics
are an emergent property of the modeled processes.
The potential timberline (the upper elevation limit
of closed forest; cf. Körner, 1998) in the study area
is estimated to be located at about 2200 m a.s.l., i.e.
about 100 m higher than the current timberline (Walder,
1983). This upper potential limit corresponds well to
the simulated timberline elevation, where biomass of