Gap models have a rich history of being used to simulate individual tree in
teractions that impact species diversity and patterns of forest succession.
Questions arise, however, as to whether these same models can be used to s
tudy the response of forest structure and composition under a changing clim
ate. In contrast to many process-based models, gap models have traditionall
y been based on rather descriptive representations of species-specific grow
th processes. Opportunities now exist to expand upon these simple empirical
relationships with more mechanistic descriptions of growth, the response o
f growth to environmental variables, and competition among species for avai
lable light, water, and nutrient resources. In this paper, we focus on seve
ral areas of below-ground research with the potential to improve the utilit
y of gap models for predicting forest composition in response to a changing
climate. Specific areas for model improvement include (1) improved descrip
tions of the soil environment for seed germination and subsequent seedling
establishment, (2) multi-layer representations of soil water and nutrient a
vailability, (3) more accurate information on biomass allocation to roots a
nd root distribution within the soil profile, (4) improved treatment of int
er- and intra-specific competition for available soil resources, (5) increa
sed consideration of spatial processes as related to land-surface hydrology
, and (6) improved attention to above- and below-ground interactions. This
list is meant to stimulate discussion and provide guidance for future field
research and model development. As an example of how increased attention t
o below-ground processes could help address intra-specific competition for
water among trees of differing size classes, the gap model LINKAGES was mod
ified to include a sub-model of multi-layered soil hydrology. It was then u
sed to examine the impact of root distribution within soils on the simulate
d drought response of seedlings, saplings, and mature trees. An annual simu
lation of soil water content for a deciduous forest in eastern Tennessee sh
owed that seedlings whose roots were restricted to the upper 20-cm of the s
oil experienced far more 'drought days' than did saplings and larger trees
that otherwise had access to deeper soil water reserves. We recognize that
models of forest succession cannot include mechanistic detail on all potent
ial below-ground processes and that there are obvious tradeoffs between mod
el simplicity and more sophisticated parameterizations. We conclude, howeve
r, that feedbacks among global environmental change, seed germination and s
eedling establishment, above- and below-ground carbon allocation, root dist
ribution within the soil profile, and soil water and nutrient dynamics will
be critically important for predicting forest dynamics and ecosystem funct
ion in the 21st century. As a result, steps should now be taken to ensure t
hat these processes are represented in future gap models.