Over the past fifteen years or so a consensus has begun to form that
the physics of catchment behavior can be captured in a meaningful way at
an appropriate scale. The basic argument is that a knowledge of
the spatial distribution of key variables suffices to capture
important physical responses at the catchment scale. Traditional
modelling approaches have handled spatial variability by dividing
a catchment into smaller geographical units on which hydrological
model computations are made, and by aggregating the results to provide
a simulation for the basin as a whole. Commonly used geographic units
can be sub-basins, terrain-based units, land cover classes, or
elevation zones. In all cases, modelling is simplified because areas
of the catchment within these units are assumed to behave similarly
in terms of their hydrological response. Of course, the question of when
to stop the process of division of a basin into ever smaller units
is vexed. As one possible answer to this question, the concept of
a representative elementary area (REA) was introduced by Wood et
al. (1988, 1990). The REA is the hypothesized smallest area, on the
order of 1 km
for catchments studied by Wood and his co-workers,
for which the pattern of local heterogeneity is relatively unimportant
in the sense that heterogeneities can be treated statistically,
without regard to the exact spatial pattern of the heterogeneity (
Wood et al. 1990). The ultimate utility of the REA concept to the
science of catchment hydrology remains to be determined. A number of
papers presented at a conference on ``Scale Issues in Hydrological
and Environmental Modelling, held in New South Wales, Australia in
December 1993, highlighted some difficulties with the REA and presented
some refinements to the concept.
Regardless of whether subdivision of a large catchment is made on the basis of an REA or on some other grounds, characterization of what is happening at the scale of the smallest subdivision (often referred to as the grid scale) on the basis of measurements made at a much finer scale---essentially at points within the subdivisions---is an unresolved problem for catchment modelling. Solving the problem of defining effective parameters at one scale and linking these values across scales is an area of active research ( Beven 1993a, 1994).
Much of the recent progress in catchment modelling is linked to advances in spatial data measurements, improvements in data quality, and new methods of extrapolating point data to areal values. The growing availability of digital spatial data and imagery along with efficient, automated methods of extracting information from them has provided a stimulus to use them in catchment models.