Environmental Engineering Reference
In-Depth Information
Estimation System (CES: see www.river-
conveyance.net) has advanced the selection of the
necessary parameters.
Above we introduced difficulties that inherent-
ly hinder the process of calibration of floodplain
inundation models, caused either by the non-ex-
istence of appropriate calibration data, or by the
problem of equifinality, which inevitably prevails
in over-parameterized models. These limitations
have in recent years motivated the development
of approaches to parameterize spatially varying
floodplain friction that do not involve calibration.
These approaches make use of the wealth of in-
formation provided by remote-sensing data such
as LiDAR (see above), from which spatially dis-
tributed details on vegetation thickness and den-
sity can be extracted (Asselman et al. 2002; Cobby
et al. 2003; Mason et al. 2003, 2007; Davenport
et al. 2004). However, it should be kept in mind
that output variables such as inundation extent
and point water levels may not be sensitive to
distributed friction values on river floodplains to
any discernible extent, as demonstrated byWerner
et al. (2005). This suggests that a methodology for
floodplain friction parameterization at a coarse
scale may be more appropriate than the use of
such technologies if output parameters that usu-
ally have low spatial variability such as water
levels are of interest. In applications where de-
tailed predictions of flow patterns (including ve-
locities) are sought, then spatially distributed
friction values have more relevance. Other types
of datasets such as high-resolution land-use maps
(Mason et al. 2007) may help in setting friction
parameter values. It should be added that even if
friction phenomena on different surfaces (roads,
etc.) and throughdifferent types of vegetationwere
adequately understood and modelled, there could
remain the issue of modelling very localized pro-
cesses involving head losses such as those caused
by hedges and fences. Again, these may be better
taken into account in a model parameterized at a
coarse scale where they would effectively be trea-
ted as subgrid processes in the sameway as bottom
friction. Considerable research aiming to design
such parameterization techniques for coarse grid
models is needed.
Fig. 12.7
1D-2D flood modelling, as can be used for
example in SOBEK. Water exchange occurs in the verti-
cal direction at bankfull level between the rivermodelled
in 1D (discharge Q) and the flood (modelled in 2D). From
Stelling and Verwey (2005).
linking nodes within a 2D model, and combined
1D/2D modelling approaches where a 1D sewer
system model can be linked to a 2D floodplain
model are also commercially available (see, e.g.,
Rungø and Olesen 2003; Danish Hydraulic
Institute 2007a).
Finally the approach consisting in coupling a
1D river model and a 2D floodplain model using
a vertical link (Fig. 12.7) should be mentioned
(Verwey 2001; Bates et al. 2005; Stelling and
Verwey 2005). This consists in representing the
floodplain using an uninterrupted 2D grid over-
laying the 1D river model. The 1Dmodel operates
on its own until the river reaches bank-full level,
at which point the water above this level is trans-
ferred to the 2D model.
Representation of surface roughness
and energy loss
In 1D modelling representation of surface rough-
ness and other energy losses is normally achieved
through a calibration and validation cycle, often
with reference to standard hydraulics texts to
informfirst estimates of the necessary parameters.
In the UK, the publication of the Conveyance
