Environmental Applications of Computational Fluid Dynamics - Environmental Modelling: Finding Simplicity in Complexity - page 96

a number of different researchers. Commercial codes
such as TELEMAC (Hervouet, 1994; SOGREAH, 2011),
MIKE31 (DHI) and Delft3D (Deltares) have been
developed and applied extensively.
In addition to the solution of the depth-averaged
equations, computational solutions in this field require
account to be taken of the effects of wind shear and
Coriolis. The former is accommodated as a momentum
source on the free surface and the latter as a source term
in the momentum equation.
The motion of waves is a significant factor in coastal
studies. The modelling of nearshore waves is not possi-
ble with the shallow-water equations because they do
not account for vertical acceleration and the conse-
quent assumption of no significant free-surface curvature.
Models have been developed based on the Boussinesq
equations for waves in shallow water (Borsboom et al .,
2001; Sorenson et al ., 2004). Li and Fleming (2001) and
Haas and Warner (2009) have developed fully 3D models
for coastal zones that predict wave motions directly.
In estuaries there can be significant effects due to
thermal and saline stratification. Adequate resolution of
these effects necessitates the use of software that solves
the shallow-water equations within a number of vertical
layers. This method allows for different temperatures and
salinities in the vertical and makes some allowance for
vertical velocities. For example, Falconer and Lin (1997)
have developed this technique and applied it to studies of
morphological changes and the transport of heavy metals
in the Humber estuary in the UK.

problems with accurate solution. Other techniques have
been proposed (Garcia-Navarro et al ., 1992; Alcrudo and
Garcia-Navarro, 1993; Garcia-Navarro et al ., 1995; Cross-
ley, 1999; Lee and Wright, 2009), but these have yet to be
put into common use.
Two-dimensional modelling of rivers is used com-
mercially, but to different extents in different countries.
Various research groups have developed and validated
these techniques: Bates et al . (1996); Sleigh et al . (1998);
Stelling et al . (1998); Bates et al . (1999); Pender and Neelz
(2010). A significant amount of work has been carried
out on the use of remotely sensed data in 2D modelling
(Horritt et al ., 2001; Schumann et al ., 2009), which offers
the advantages of increased accuracy and faster model
development.
In many cases 2D solutions are only required in parts
of a river model and so in the majority of the river system
a 1D model is adequate. Several authors (Dhondia and
Stelling, 2002; Hunter et al ., 2005; Lin et al ., 2006; Vil-
lanueva andWright, 2006) have implemented a combined
1D/2D approach that allows for a complete 1D model to
be augmented, rather than replaced, in key areas.
Since the early 2000s there has been renewed interest
in simplified models for inundation in 2D based on
simplified representations (Bates and de Roo, 2000; Yu
and Lane, 2006a; Yu and Lane, 2006b; Hunter et al .,
2007; Lamb et al ., 2008; Bates et al ., 2010). This interest is
motivated by their speed, but it has been found that this is
only the case at coarsermeshes (Hunter et al ., 2008) due to
timestep restrictions as the mesh is refined. Later versions
have overcome this limitation through inclusion of better
representations of momentum (Bates et al ., 2010).
The use of fully 3DCFD codes within rivers is still a pre-
dominantly research activity. However, this work can give
new understanding of the fundamentals of fluid flow in
channels. The first attempts at using CFD considered sim-
plified channels (Rastogi and Rodi, 1978; Leschziner and
Rodi, 1979; Rodi, 1980; Naot and Rodi, 1982; Demuren
and Rodi, 1986; Gibson and Rodi, 1989), but did demon-
strate the potential application. Lane et al . (1999) carried
out 3D simulations and compared the results with a
2D simulation for flow at a river confluence for which
high-quality field data was available. They found that
3D offered better predictions for secondary circulations.
The 3D model also gave better estimates of shear stress
and provided better information for calculating mixing
processes. More recent work has extended solutions with
Large Eddy Simulation for laboratory channels (Brad-
brook et al ., 2000) and rivers (Lane et al ., 2002; Hardy
et al ., 2005; Keylock et al ., 2005). Such work is also now

6.3.1.2 Rivers

One-dimensional models are in widespread use in com-
mercial consultancy and are the main tool for assessing
flood risk, water quality and construction impact in rivers
and channels. In addition to modelling the discharge and
depth at river cross-sections these codes have units that
model hydraulic structures such as sluice gates and weirs.
Through use of these units, a model of a complex network
of rivers and artificial channels can be built up. In this
way current computer power is beginning to allow for
models of whole catchments and real-time flood predic-
tion. Cunge et al . (1980) provide an excellent background
to the theory and application of river models. Predomi-
nantly, the 1D equations are solved by the Preissmann or
other 'box' methods. These methods allow for the simple
inclusion of units such as sluice gates and weirs. However,
they do not take account of the fact that the equations
are hyperbolic (see Smith, 1978, for definition) leading to