Operational European Flood Forecasting - Environmental Modelling: Finding Simplicity in Complexity - page 400

(Demeritt et al ., 2007, 2010; Ramos et al ., 2007; Nobert
et al ., 2009).
For medium-range flood forecasts, observations of
rainfall and river discharge can be used to initialize
the forecasts, and future information on rainfall can be
obtained fromweather forecasts. Such numerical weather
predictions (NWPs) are based on the extrapolation of cur-
rent weather conditions using a mathematical model to
predict the future state of the atmosphere (Coiffier, 2011).
Operationally, these predictions are usually produced
with a coarser spatial grid for global scale studies (cur-
rently on the order of 15-50 km resolution) and on a finer
grid for the local scale work (currently

(i) the principles of calculating flood-warning thresholds
(ii) the techniques used to model river discharge spatially;
(iii) the NWP meteorological forecasts used as input
forcing for the hydrological model and (iv) a description
of a Hydrological Ensemble Prediction System.

25.2.1 Flood-warningthresholds
An important issue when dealing with flood forecasting
on a regional or continental scale is the determination
of critical alarm levels. Typically, flood warnings are
issued for locations where river discharge [L 3 T − 1 ]or
river stage (water level) [L] is expected to be greater than
a defined warning threshold. At any particular location,
actual flooding will occur when the river stage is greater
than bankfull level (the height of the top of the river
bank). However, the height of the river banks and the
geomorphological form of the river channel will vary
along the river (as well as over time) and so it is difficult
to define one bankfull level for a particular river reach
as is commonly required with a modelling approach.
For this reason, bankfull level is usually defined for
representative points along a river, often at river gauging
stations where stage or discharge are monitored. Bankfull
is thus one of the critical flood thresholds to identify
and to forecast (Figure 25.1). By the above definition
predicting discharge greater than bankfull discharge is to
predict flooding. However, in practice bankfull along an
entire reach may not be the most useful threshold for
flood warning and instead thresholds that are defined for
river discharge or river stage at important locations (such
as power stations) or a specific number of properties being
inundated by flood waters may be preferred. A range of
warning thresholds is usually computed depending on
the needs of the end-users of a flood-forecasting system.
Flood-warning thresholds can be derived by com-
puting flood return periods from observed hydrological
river discharge or river-stage data, which is achieved by
the fitting of extreme value distributions and deriving
the probability of a flood magnitude (Hamed and Rao,
1999). However, on a European scale, the determination
of thresholds from observed data is currently impossible
due to two major limitations: (i) the paucity of available
operation rules for lakes, reservoirs or any other structural
measure along a river, which can lead to large discrep-
ancies between modelled and observed river discharge,
making a quantitative assessment based on a uniform
threshold (which is valid for simulated and observed)
problematic; (ii) the scarcity of river-discharge-gauging
stations across Europe (and also the accessibility of these

1 km). Quantita-
tive flood forecasting in the medium range thus depends
on a system of mathematical models used together with
high-resolution, spatial data sets. It is only possible with
powerful computing systems, data-storage capacities and
remote sensing technology for observational data collec-
tion.
There is some evidence from climate-impact mod-
elling (see Chapter 9) that climatic change may lead to
an increased probability of extreme precipitation and
river flooding in Europe (Milly et al ., 2002; Lehner et al .,
2006; Goubanova and Li, 2007; Arnell, 2011), thus high-
lighting the need for novel initiatives to improve the
current status of flood-forecasting capabilities worldwide
but also in Europe. In addition, as a result of economic
development and increasing infrastructural complexity
and dependency, the risk of economic damage and other
negative consequences of flooding may increase in the
future, even where the 'natural magnitude' of the flooding
does not. Over the past decade, the European Com-
mission has given support to improve medium-range
flood-forecasting capacity for large transnational river
catchments in Europe through the development of the
European Flood Alert System (EFAS), (full details pro-
vided in Section 25.3). In this chapter we will look at
the modelling approaches used in the EFAS and how
they address the challenges of operational early flood
forecasting at the European scale.

∼

25.2 Flood forecasting at the European
scale: the approaches

The EFAS uses several sets of spatially distributed
NWP forecasts to drive a rainfall-runoff-routing model.
The river discharge predictions from the model are
transformed into early flood warnings for national
hydrological services across Europe. This section covers