Environmental Engineering Reference
In-Depth Information
The need to understand the impacts of human behaviour.
With global human populations continuing to increase
and per capita resource use high and increasing in the
developed world and low but increasing in much of
the developing world, the need to achieve renewable
and nonrenewable resource use that can be sustained
into the distant future becomes more pressing. Better
understanding the impacts of human resource use
(fishing, foresting, hunting, agriculture, mining) on the
environment and its ability to sustain these resources
is thus an increasing thrust of environmental research.
Models, for many of the reasons outlined above, are
often employed to investigate the enhancement and
degradation of resources through human impact.
The need to understand the impacts on human behaviour.
With the human population so high and concen-
trated and with per capita resource needs so high and
sites of production so disparate from sites of con-
sumption, human society is increasingly sensitive to
environmental change. Where environmental change
affects resource supply, resource demand or the ease
and cost of resource transportation, the impact on
human populations is likely to be high. Therefore
understanding the nature of variation and change in
environmental systems and the feedbacks of human
impacts on the environment to human populations are
both increasingly important. Environmental science
increasingly needs to be a supplier of reliable forecasts
and understanding to the world of human health and
welfare, food and water security, development, politics,
peacekeeping and warmongering.
However, actual model applications may not be so
simple. We may be interested in trying to reconstruct past
environments, or the conditions that led to catastrophic
slope collapse or major flooding. In such cases, it is not
possible to measure all of the parameters of a model that
has a reasonable process basis, as the conditions we are
interested in no longer exist. In such cases, we may have
to make reasonable guesses (or estimates, if you prefer)
based on indirect evidence. The modelling procedure may
be carried out iteratively to investigate which of a number
of reconstructions may be most feasible.
Our optimal model structure may also produce param-
eters that it is not possible to measure in the field setting,
especially at the scales in which they are represented in
the model. The limitations may be due to cost, or the lack
of appropriate techniques. It may be necessary to derive
transfer functions from (surrogate) parameters that are
simpler to measure. For example, in the case of infiltra-
tion into hillslopes, the most realistic results are generally
obtained using rainfall simulation, as this approach best
represents the process we are trying to parameterize
(although simulated rain is never exactly the same as
real rain - see Wainwright et al ., 2000, for implications).
However, rainfall simulation is relatively difficult and
expensive to carry out, and generally requires large vol-
umes of water. It may not be feasible to obtain or transport
such quantities, particularly in remote locations - and
most catchments contain some remote locations. Thus, it
may be better to parameterize using an alternative mea-
surement such as cylinder infiltration, or pedo-transfer
functions that only require information about soil tex-
ture. Such measurements may not give exactly the same
values as would occur under real rainfall, so it may be nec-
essary to use some form of calibration or tuning for such
parameters to ensure agreement between model output
and observations. In extreme cases, it may be necessary
to attempt to calibrate the model parameter relative to
a known output if information is not available. We will
return to the problems with this approach later.
Parameterization is also costly. Work in the field
requires considerable investment of time and gener-
ally also money. Indeed, some sceptics suggest that the
research focus on modelling is driven by the need to
keep costs down and PhDs finished within three years
(Klemes, 1997). Equipment may also be expensive and
if it is providing a continuous monitored record, will
need periodic attention to download data and carry out
repairs. Therefore, it will generally never be possible to
obtain as many measurements as might be desirable in
any particular application. As a general rule of thumb,
2.2 Approaches to model building:
chickens, eggs, models and
parameters?
Should a model be designed around available measure-
ments or should data collection be carried out only once
the model structure has been fully developed? Many hard-
ened modellers would specify the latter choice as the most
appropriate. The parameters that are required to carry
out specific model applications are clearly best defined
by the model structure that best represents the processes
at work. Indeed, modelling can be used in this way to
design the larger research programme. Only by taking the
measurements that can demonstrate that the operation
of the model conforms to the 'real world' is it possible to
decide whether we have truly understood the processes
and their interactions.
Search WWH ::




Custom Search