Geoscience Reference
In-Depth Information
outstations. Thus, in any flood forecasting situation it is useful to have data assimilation methods that
both allow for
real-time updating
of the forecasts as the event proceeds and constrain the uncertainty in
the forecasts. This is particularly the case where the sources of uncertainty are not simply random but
can be the result of lack of knowledge. Krzysztofowicz (2002), for example, suggests that
operational uncertainty is caused by erroneous or missing data, human processing errors,
unpredictable interventions (e.g. changes in reservoir releases not communicated by a dam
operator to the forecaster), unpredictable obstacles within a river channel (e.g. ice jams),
and the like.
To these we might add the potential for unexpected breaches of flood defences during an event.
In forecasting (rather than simulation), these sources of epistemic uncertainty can be at least partially
compensated by the use of adaptive modelling methods.
Forecasting methods based on transfer function modelling are ideally suited to these requirements and
have been quite widely implemented (see, for example, Sempere Torres
et al.
, 1992; Cluckie, 1993; Moore
et al.
, 1990; Lees
et al.
, 1994; Young, 2002; Romanowicz
et al.
, 2006, 2008). An example application
using the DBM methodology is provided by Leedal
et al.
(2008), who describe the implementation of a
flood forecasting system for the town of Carlisle on the River Eden in Cumbria (see Section 8.5).
8.3 Data Requirements for Flood Forecasting
One of the most important ways of mitigating the costs of flood damage is the provision of adequate
warnings, allowing people to act to protect their property and themselves. The responsibility for flood
warnings varies between countries. In most cases, the system is based on local flood warning offices that
become operational as soon as a potential flood-forming rainfall is forecast. The offices use rainfall-
runoff modelling in real time to predict the likely flood discharges and stage in different areas as a basis
for decisions about whether to issue flood warnings.
This process is very much easier in large catchments where the build up of a flood and the transmission
of the flood wave downstream can take days or even weeks. Recent extreme cases include the Tennessee
floods in 2010; the Pakistan floods in 2010; and the Toowoomba and Brisbane, Australia floods of 2011.
In small catchments, with short times to peak, real-time flood forecasting, is much more difficult. The
most extreme discharges in such catchments tend to occur as a result of localised convective rainfalls or
high intensity cells within larger synoptic weather systems. Even if there are telemetering raingauges or
a rainfall radar monitoring the area, the response times may be too short to issue warnings in real time.
Events where the required lead time for issuing a warning is less than the catchment response time are
called “flash floods”. Recent examples include the 2009 floods in Cumbria, England, and in Arkansas,
Singapore, Ladakh and the Var, France, in 2010. The best option for such flash floods is generally to
issue warnings on the basis of forecasts of extreme rainfalls (e.g. Hapuarachchi
et al.
, 2011), but such
warnings tend to be very general since the precipitation forecasts of weather models are not sufficiently
accurate in either time or space (Pappenberger
et al.
, 2008b). The potential for a flood-producing rainfall
might be recognised but it may be very difficult to specify exactly where. A good example of this occurs
in southern France and northern Italy where, particularly in September, weather systems moving inland
from the Mediterranean and being subjected to orographic rise can give rise to very intense rainfalls
and floods. There is at least a minor flood somewhere in this broad region virtually every year. Well-
known examples are the 1988 flood in Nice and the 1992 flood on the 560 km
2
Ouveze catchment at
Vaison-la-Romaine in France.
In the case of Vaison-la-Romaine, the French Bureau de Meteorologie (now Meteo-France) issued
a warning of the potential for flash flooding for the region. It is difficult in such a case for individual
