Environmental Engineering Reference
In-Depth Information
the domain offers a great advantage over experimental
methods that often have a restricted number of mea-
suring points. The many forms of graphical techniques
such as shaded contours, velocity vectors, iso-surfaces and
particle trajectories allow CFD users to gain tremendous
and unprecedented insights into the calculated solutions.
However, these powerful visualization tools can give false
confidence in the results. Incorrect results are still incor-
rect, however well they are visualized.
should be kept within a physically realistic range and
the calibrated model should not be applied in situations
that differ too much from the original situation used for
calibration. Secondly, users should avoid the situation
that is sometimes seen where measured data are used
to calibrate a model and then used again to validate the
calibrated model. As is to be expected, the model that has
been calibrated to fit the data turns out the fit the data
well, which leads to the false conclusion that the model
is correct in this and other situations.
When testing models, it is necessary to give careful
thought as to which variables to use in comparison:
for example, in flood-inundation modelling is depth or
maximum flood-inundation extent most relevant? The
answer will depend on the geometry of the study area and
on the decisions to be made based on the model results.
Further, if amodel is validated against streamwise velocity
in an open channel, it is not guaranteed to give accurate
results for secondary circulations in the vertical, cross-
stream plane as these features are far smaller and more
sensitive to other aspects of the model such as boundary
conditions and turbulence representation. Often models
are used to analyse variables that are not directly predicted,
such as when a hydraulic model is used to predict habitat
suitability. For example Clifford
et al
. (2010) consider in
some detail which flow structures are most relevant in
modelling for meso-habitat concepts.
In the case of uncertain input parameters it may be
very difficult to test a model and in this case it may be
best to focus more on a sensitivity analysis which will
at least give an indication of which input parameters
are most important and also indicate what conclusions
can be drawn from the results. Note should be taken
of Beven's (2006) observation that many sets of input
parameters may give equally correct results - the concept
of 'equifinality' (see also Chapters 2 and 4).
6.2.8 Validationandverification
The rapid development of CFD has not brought it to the
level of a 'black-box' analysis tool. Experience and under-
standing are still needed on the part of the operator both in
setting up a problem and analysing its results. A number
of efforts have been made to assist and the ERCOF-
TAC Guidelines (Casey and Wintergerste, 2000) are an
example. Guidance is also available from reports such as
that produced by the ASME (1993) and AIAA (1998).
However, as observed by some (Lane and Richards, 2001)
these guidelines have been developed very much with
mechanical and process engineering in mind and so are
not as relevant to flows in the natural environment. That
is not to say that they are completely irrelevant and
some of the underlying philosophy is worthy of note. In
particular, the ASME guidelines have nine required tests
and one optional one. It may be surprising that the lat-
ter is comparison with experimental measurements. This
requirement is motivated by two reasons: the first is that
tests based on assessing whether the computational solu-
tion of the model is mathematically sound are necessary
before deciding whether that model is a good represen-
tation of physical reality and secondly that agreement
between the model results and one set of experimental
data may be fortuitous and should not indicate that the
model will work elsewhere. Lane
et al
. (2005) take this
argument further and discuss whether validation against
measurements should not be a requirement at all due to
its limitations, but this is perhaps an extreme view (see
also the discussion in Chapter 2).
Models in the natural environment cannot contain all
the physical processes and inevitably some are included
using empirical terms that require parameters that cannot
be measured or established directly (Beven, 2002), which
leads to the process of calibration where these parameters
are adjusted to produce model results that are close to
the measured data. This calibration process must be
carried out with care to ensure that results are useful
(Beven and Binley, 1992). First of all, the parameters
6.3 Applications of CFD in
environmental modelling
6.3.1 Free-surfaceflows
As mentioned above, the term CFD is often taken solely
to cover solutions to the Navier-Stokes equations in three
dimensions. However, in many environmental applica-
tions, such as rivers, estuaries and coastal areas, this
approach is not the most appropriate one. Flows in these
situations usually have one or two dimensions consid-
erably larger than the other. Fully 3D solutions are thus
neither feasible nor necessary. For example, in studying
Search WWH ::
Custom Search