297
contrast between north and south facing
slopes. Slopes 1 and 2 are the first slopes
on the left and right, respectively, looking
upstream from the watershed outlet. Their
seasonal trajectories of ET (figure 4) are
driven by contrasts in their radiation
environments, SWC and LAI.
Validation of the distributed model
results can effectively be accomplished by
outlining a suite of model products that can
be periodically observed on different slopes
through the growing season. This amounts
to replicating standard sampling schemes for
model validation usually carried out for
point models, with the added benefit of
gaining spatial patterns of either agreement
or disagreement of model predictions with
field observations. Unfortunately, this also
increases the effort and difficulty of field
sampling in proportion to the number of
observation points taken to gain that pattern.
Another possibility for model validation
using the geographic pattern of model
performance would be to use satellite or
airbomed imagery taken at different key
parts of the growing season, using spectral
bands or combinations of bands that could
be correlated with model output. As an
example, LWP or ET may be correlated with
thermal bands on a very detailed and
observable landscape level (from hillslope to
hillslope or within hillslopes) as variations in
the spatial patterns of the latent heat flux
over the season pick up those areas
experiencing progressive moisture stress.
DISCUSSION
We have briefly outlined the basic
structure of RESSYS. The system automates
the raw geographic data processing and the
parameterization and execution of a forest
ecosystem model, FOREST-BGC. We are
currently expanding the capabilities of the
ecosystem model, and therefore the
corresponding functions of geographic
information processing components of
RESSYS. The hydrologic submodel of
FOREST-BGC is being supplanted with a
version of TOPMODEL (Beven and Kirkby,
1979) to simulate lateral flux of soil water
down the hillslopes. Other subcomponents
are also being modified or added to in order
to generalize the model to different
environments. This results in a situation in
which a given environment will require a
given set of submodels and geographic
information processors to be activated within
RESSYS.
Currently, RESSYS can flexibly
partition a landscape into different numbers
of elemental data and simulation units,
giving us the ability to choose the
appropriate scale of landscape representation
and simulation or allowing us to explore the
impact of surface representation detail on
simulation behaviour. Using naturally
occuring, functional hydrologic units
(hillslopes) as our basic landscape units
rather than arbitrarily located grid cells also
allows us to preserve greater detail in
landscape patterns that would otherwise be
averaged out by aggregating over distinctly
different local microenvironments (e.g.
averaging over north and south facing
slopes.) The range of model outputs that are
observable or measurable in the field,
coupled with the preservation of distinct
microenvironments through the analysis and
simulations should enhance our ability to
validate these distributed models.
So far, we have used the method on
watersheds up to 1600 square kilometers in
western Montana and are currently in the
beginning phase of assembling a detailed
representation of an area comparable in size
to a GCM grid cell, or a number of
mesoscale circulation model cells in order to
produce surface hydroecologic models that
can be run over regional to subcontinental
scales, but can also be locally validated by
redefinition of the landscape across this scale
range.
Acknowledgements: This research was
supported by funding from the Land Surface
Processes Branch of NASA, NAGW-1234.