Forest gap models have a long history in the study of forest dynamics, incl
uding predicting long-term succession patterns and assessing the potential
impacts of climate change and air pollution on forest structure and composi
tion. In most applications, existing models are adapted for the specific qu
estion at hand and little effort is devoted to evaluating alternative formu
lations for key processes, although this has the potential to significantly
influence model behavior. In the present study, we explore the implication
s of alternative formulations for selected ecological processes via the com
parison of several gap models. Baseline predictions of forest biomass, comp
osition and size structure generated by several gap models are compared to
each other and to measured data at boreal and temperate sites in North Amer
ica. The models ForClim and LINKAGES v2.0 were compared based on simulation
s of a temperate forest site in Tennessee, whereas FORSKA-2V, BOREALIS and
ForClim were compared at four boreal forest sites in central and eastern Ca
nada. Results for present-day conditions were evaluated on their success in
predicting forest cover, species composition, total biomass and stand dens
ity, and allocation of biomass among species. In addition, the sensitivity
of each model to climatic changes was investigated using a suite of six cli
mate change scenarios involving temperature and precipitation. In the tempe
rate forest simulations, both ForClim and LINKAGES v2.0 predicted mixed mes
ophytic forests dominated by oak species, which is expected for this region
of Tennessee. The models differed in their predictions of species composit
ion as well as with respect to the simulated rates of succession. Simulated
forest dynamics under the changed climates were qualitatively similar betw
een the two models, although aboveground biomass and species composition in
ForClim was more sensitive to drought than in LINKAGES v2.0. Under a warme
r climate, the modeled effects of temperature on tree growth in LINKAGES v2
.0 led to the unrealistic loss of several key species. In the boreal forest
simulations, ForClim predicted significant forest growth at only the most
mesic site, and failed to predict a realistic species composition. In contr
ast, FORSKA-2V and BOREALIS were successful in simulating forest cover, gen
eral species composition, and biomass at most sites. In the climate change
scenarios, ForClim was highly sensitive, whereas the other two models exhib
ited sensitivity only at the drier central Canadian sites. Although the stu
died sites differ strongly with respect to both the climatic regime and the
set of dominating species, a unifying feature emerged from these simulatio
n exercises. The major differences in model behavior were brought about by
differences in the internal representations of the seasonal water balance,
and they point to an important limitation in some gap model formulations fo
r assessing climate change impacts.