Environmental Engineering Reference
In-Depth Information
classes (assuming there are far fewer classes
than areas).
As one might expect, this approach is fraught
with difficulties, not least that: (i) the small-scale
model must encapsulate information derived
from small-scale measurements - but small-scale
hydrology is never simple, particularly when nat-
ural hydrological functioning has been altered
through intensive agriculture; and (ii) by its very
nature, the classification/regionalization method
must use some type of large-scale (approximate)
information, such as maps of land cover. These
difficulties, though, are not specific to the problem
of predicting change impacts, but are present in
some formwhenever a bottom-up approach is used
in catchment rainfall-runoff modelling.
To get right to the heart of the problem of
predicting change effects, Question 3 can be
broadened out into a question about how detailed
knowledge and understanding gained at a small
scale (e.g. in field plots where land use/manage-
ment or rainfall is manipulated) finally ends up
affecting predictions made for impacts at large
scales (e.g. flooding of a floodplain, downstream).
The reason why it broadens out in this way is
that, as shown in the previous section, we are far
more likely to gain information about change
effects at small scales than at large scales. Perhaps
if progress can be made with this broader ques-
tion, then the other questions will be easier to
answer.
Part of the solution must lie in creating a direct
link between small-scale parameters and large-
scale impacts, and making sure that nothing is
done in the modelling process that corrupts or
breaks this link. There are a few ways that the
link can be made, including multiscale modelling
wheremodels at different scales are combined (e.g.
Bronstert et al. 2007), and using direct transfers via
explicit functions that define how large-scale para-
meters are related to small-scale parameters (e.g.
Hundecha and Bardossy 2004). Another approach
is metamodelling (Ewen 1997; Ewen et al. 1999;
Kilsby et al. 1999; Audsley et al. 2008; Jackson
et al. 2008), as described in Chapter 3. The main
steps in a metamodelling approach are:
1
Use models at the small scale with parameters
that are, as far as practical, based directly on
measurements made at the small scale (e.g. phys-
ically based runoff generation models).
2
Build a simple model that reproduces the re-
sponses produced by the small-scale model. That
is, build an emulation model (this is called the
metamodel).
3
Build the necessary large-scale model:
a
break the catchment into many areas (e.g. by
grid or subcatchment);
b
apply the emulation model to each area;
c
add a routing scheme to link the areas; and
d
create a classification/regionalization meth-
od, so that the effort spent on emulation is
reduced from finding parameters for each of
the areas to finding parameters for each of the
Information tracking
and vulnerability mapping
Despite the problems outlined above, and given
the pressing needs of flood risk policymakers
and managers, it is inevitable that impact
models will be used in developing catchment
flood risk management plans. As noted above,
a portfolio of interventions will need to be
considered, including land use management mea-
sures. 'Vulnerability mapping' represents a useful
way of assembling the available knowledge and
understanding into a form suitable to support
decision-making. For example, it would be useful
to have amap showing locations in the catchment
where changes of various types should be restrict-
ed because they increase the flood hazard down-
stream, and conversely it would be useful to have
maps that show the optimum locations to imple-
ment mitigation measures that decrease the flood
hazard downstream. It would be simple to create
vulnerability maps of this type if any given inter-
vention (i.e. implemented change) is always
detrimental or beneficial in the same way and by
the same amount at any scale. If this was the case,
then vulnerability maps could be created by
agricultural scientists, soil scientists and field
hydrologists, using only the small-scale informa-
tion they routinely work with and map. Reality,
however, is much more complicated, because
