Prediction of floods in ungauged basins - Runoff Prediction in Ungauged Basins: Synthesis across Processes, Places and Scales - page 217

Figure 9.21. Cross-section of the
Alcantara bridge (Tagus River,
Spain) showing different peak water
stages during historic floods, which
have been related to the rating curve
(right) obtained from one-
dimensional hydraulic modelling of
a 1 km long reach. In the centre,
the date and estimated runoff from
the rating curve are indicated for the
reported historical floods. From
Benito ( 2003 ).

amount of information on the catchment processes that is
used for the selection of the parameters and the model struc-
ture. It is likely that the relative performance depends more
on the available information and the skill of the modeller than
on generalisable characteristics of the model or the approach
(see Section 3.7.2). Other information on the behaviour of the
catchment during floods would therefore be extremely useful
in order to assist in flood frequency estimation in ungauged
basins. One such piece of information is proxy data.

of chronicles and documents that do not usually give many
details on the particular flood, but do provide information
on its impacts (proxy data; e.g., Benito et al., 2004 ). In
particular, historical archives may provide the date of
occurrence of past floods, information on the meteoro-
logical situation and some indication about the peak water
levels (e.g., flood marks on buildings, bridges, trees etc.;
see Figure 9.21 ).
Historical water levels can be converted into runoff by
means of hydraulic modelling (e.g., Calenda et al., 2005 ).
In this context, interesting examples of historical flood data
are the Nile flood stages, from AD 640 to 1921 (Hassan,
1981 ), and the reconstruction of Tiber flood peaks per-
formed by Calenda et al.( 2005 ). The Hydrological Sci-
ences Journal recently published a Special Issue on this
topic (Brázdil and Kundzewicz, 2006 ). Hydraulic proced-
ures unavoidably introduce additional sources of uncer-
tainty as several assumptions (e.g., slope, roughness and
cross-section geometry) need to be made to complete the

9.4.3 Proxy data on flood processes
Historical flood information
In ungauged basins, a continuous runoff record will be
unavailable, but there may be observations of a few extra-
ordinary floods. Indeed, non-systematic historical flood
data (i.e., human records of flood peaks; Stedinger and
Baker, 1987 ) can be very useful in predicting floods.
Historical flood data are usually derived by the analysis