Environmental Engineering Reference
In-Depth Information
Fig. 11.6
The reliability diagram. The uncertain prediction (top left) is classified into areas of similar Cj (top middle),
then the ratio of observed wet/dry cells (top right) calculated. These then make up the reliability plot (bottom), with
a reliable model being one with points close to the 1:1 line. After Horritt (2006).
With the launch of new radar satellites and
missions (e.g. RADARSAT-2, ALOS, Cosmo-
SkyMed, TerraSAR-X and SWOT) with better spa-
tial and radiometric resolutions, the uncertainties
of water level estimates will presumably be
further reduced, getting closer to the desired
centimetre-scale accuracy.
in the upstream boundary inflow (Andreadis
et al. 2007; Matgen et al. 2007b) and channel rough-
ness have been considered the only sources of model
error, and other likely sources of model uncertainty
such as floodplain roughness, channel andfloodplain
topography, and model structure have been ne-
glected. There is a consensus that a simple reinitia-
lization of models with distributed water stages
obtained from remote sensing does not lead to sig-
nificant improvement because of the dominating
effect of the forcing terms on the modelling results
(Matgen et al. 2007b). By sequentially confronting
models with remote-sensing observations it be-
comes possible to 'diagnose' the latest model set-up
and to find out how modelling can be improved.
Water stage assimilation into
inundation models
Studies in the field of assimilation of remote-
sensingdatainfloodforecastingsystemsareat
present only very few in number, as there are a
number of considerable challenges to be faced. Errors
