Environmental Engineering Reference
In-Depth Information
city). Such a method may fail on its own if prediction
beyond this seven-hour time horizon is attempted, or if
an additional inflow between these points occurs.
groundwater-gravity remote sensing make this measure-
ment a future possibility Gruhier et al ., 2008; Strassberg
et al ., 2009). The state of the land phase is therefore often
derived by using the hydrological model with meteoro-
logical observations as input data input (Pappenberger
et al ., 2011).
System 2 - land-phase-to-river
The river discharge at any point along a river is
mainly dominated by water that is already in the land
phase - water that has fallen as rainfall and is now stored
or being transported through the catchment. In this case
we need to use a hydrological model to establish how
much water is going from the land into the river. The
type of hydrological model needed will depend largely
on the catchment characteristics (see Chapter 11) but it
should be noted that the selection of a suitable model
is not straightforward, and all modelling approaches are
necessarily based on a series of approximations and result
in uncertain results (Beven, 2001, 2008; Hughes, 2010).
The hydrological model could be a simple statistical
or conceptual model, which relates the soil moisture
and ground-water storage to river water by simple (not
necessarily physics-based) equations (Wohling et al .,
2006). Such a model needs to be calibrated explicitly - for
example, it needs a time series of river discharge and water
in land measurements. One could also use a 'physically
based' model with physics-based equations, which are
often more distributed in nature (Vieux et al ., 2004).
To build such a hydrological model, one needs to know
catchment characteristics such as topography, channel
dimensions, land use, soil properties and geology. Many
models are hybrids and use partly conceptual and partly
physically based equations (Beven and Freer, 2001; van
der Knijff et al ., 2010) or alternatively are derived from
land-surface schemes fromatmosphericmodels (Balsamo
et al ., 2011). The resolution andquantity of input data that
are needed will largely depend on the complexity of the
hydrological model involved. It is important to note that a
more complex model does not necessarily produce better
results than a simple design (Beven, 2008; Hughes, 2010).
The larger the river, the longer the lead times that
can be achieved with a land-phase-to-river system, for
example for border stations between Germany and the
Netherlands on the River Rhine a lead time of up to
seven days can be achieved, as the runoff from the land
(catchment response) and the time taken to route the
water down the river channel (flood wave lag time)
combined is approximately seven days.
It is often impossible to measure the exact amount
and distribution of water in the land phase, particu-
larly for ground water and the distribution of the water
table, (although recent developments in soil-moisture and
System 3 - atmosphere-to-river
An atmosphere-to-river system is used to forecast lead
times greater than those that can be obtained by just
using observations of river and catchment. The system
relies on the initial conditions provided by systems 1 and
2 together with predictions of the future meteorological
'forcings' (for example, precipitation, temperature,
evaporation). The type of meteorological forcing will
depend on the events that are to be predicted. For
example, flash flood events develop within six hours or
less and require good quality, high-resolution inputs on
precipitation. These are often provided by a combination
of rain-gauge observations and radar-derived rainfall
fields, which can also be extrapolated for a few hours
into the future by making assumptions about growth and
decay of rainfall cells (Collier, 1989; Bowler et al ., 2006;
Velasco-Forero et al ., 2009,). Recently, promising results
have been obtained by blending radar information with
high-resolution, short-range meteorological modelling in
order to increase the lead times for which such forecasts
can be reliable (Atencia et al ., 2010). For larger scale
flood events, longer lead times are achieved by relying
solely on numerical weather forecasts, such as NWP, for
their forcing. The type of the meteorological forcing can
be lumped (for example, only one catchment average) or
distributed (for instance, a pattern of rainfall) depending
on the hydrological model involved. The quality of the
meteorological forcing will largely influence the quality
of the flood forecast (Arnaud et al ., 2002, Pappenberger
et al ., 2011; Parkes et al ., in press).
The EFAS relies on a distributed hydro-meteorological
modelling system, which provides flood forecasts at every
point on a large spatial grid covering the European con-
tinent using distributed NWP forcing and a distributed
hydrological model (see Section 25.3 for more details).
Model equations, parameters and input data must be
specified for every single grid cell. For each grid cell
that contains a station with observations, these equations,
parameters and forcings are derived directly from the
observations; for all other cells they have to be derived
indirectly through techniques such as interpolation. Such
parameterization introduces important uncertainties into
the predictions. It is essential to have a quantitative
Search WWH ::




Custom Search