Environmental Engineering Reference
In-Depth Information
urban flood modelling made possible by the avail-
ability of high-resolution LIDAR data is but one
example. Whilst the progress made to date is
significant, much further work remains to exploit
fully the data from current sources in the model-
ling process. In addition, anticipated develop-
ments in data sources in the near future mean
that the ongoing revolution in the production of
data for inundation models is likely to proceed
for some time yet. These developments suggest
a number of topics for future research to better
meet the data requirements of inundation mod-
ellers, and these are set out below.
Flood inundationmodels are only as good as the
data used to validate them. There is a need for
better model validation datasets, particularly in
urban areas. In rural areas, 2-D flood models have
been successfully validated using flood extents
determined from SAR data, typically ERS and
ASAR. However, these have too low a resolution
for use in urban areas. This situation should
improve in the near future as the number of
operational SARs and their spatial resolutions
increase. The high-resolution TerraSar-X,
RADARSAT-2, ALOS PALSAR and the first two
of the COSMO-SkyMed satellites have recently
been launched. When the four satellites in the
COSMO-SkyMed constellation become opera-
tional, a flood revisit time of a few hours should
be possible. However, even with resolutions of
only a fewmetres, the side-looking nature of SAR
means that substantial areas of ground surfacewill
not be visible due to shadowing and layover caused
by buildings, and it will be necessary to correct for
these in estimating flood extent (see Fig. 11.4).
The number of operational SARs, coupled with
the rise in importance of the Disasters Charter
Agency and the advent of the European Space
Agency's Heterogeneous Missions Accessibility
(HMA) project, also bodes well for the production
of SAR image sequences for future flooding
events. The Disasters Charter Agency (www.
disasterscharter.org) aims to provide a unified sys-
temof space data acquisition and delivery to those
affected by disasters such as flooding via its mem-
ber space agencies. The HMA project is establish-
ing harmonized access to Earth Observation data
In this respect, the use of error forecasting
models (Andreadis et al. 2007; Neal et al. 2007)
in the context of spatially distributed gauge mea-
surements seems to indicate a promising way
forward. A persistent improvement can only be
obtained by looking for, and identifying, the
reasons that cause disagreement between model
results and observations. The objective must be to
identify and correct components that are respon-
sible for the discrepancy between modelled and
observed variables.
It may be argued that such a 'diagnostic
approach' (Gupta et al. 2008) ensures efficiency
when assimilating remote sensing into floodmod-
els. It is reasonable to assume that the difference
between the observed and simulated water sur-
faces might indicate that the model set-up is ques-
tionable. An adequate understanding of all the
different error sources and interactions is needed
to conduct the model development in a meaning-
ful way. Since flood inundation models are cali-
bratedwith data of a past flood event, the potential
reasons for a mismatch can indeed be numerous:
the rating curve used to describe the boundary
condition might become erroneous after a given
flow magnitude; model parameters (i.e. channel
and floodplain roughness values) may vary with
time (e.g. due to vegetation growth); important
intermediate inflows may have been neglected; or
the model structure may become inappropriate
with increased inflow (Schumann et al. 2009).
Put more bluntly, if there is considerable disagree-
ment between observations and model simula-
tions, then both need to be questioned and
improvement of
the modelling
should be
envisaged.
Conclusion and Future Research
The foregoing has hopefully illustrated the wide
variety of ways in which data are currently being
utilized in flood inundation models. Substantial
progress can be seen to have been made over the
past decade in the development of new data
sources, and on the integration of the data they
produce into the models. The improvement in
