Geography Reference
In-Depth Information
6.3.3 Integrationwithflood inundation
modelling
provide the basis for assimilation of remotely sensed data
(Alsdorf et al., 2005), for instance, in operational flood
forecasting systems.
In order to outline more clearly the techniques applied
and values associated with integrating SAR-derived flood
parameters with flood inundation models, the next
section describes by means of illustrative case studies
how SAR image data have recently been employed to
support flood modelling.
The integration of remotely sensed flood parameters with
flood inundation models requires a profound under-
standing of the many factors underlying both the remote
sensing and the flood modelling part. For this reason, rel-
atively few studies have looked at this complex interplay.
This chapter has so far illustrated that remote sensing can
provide information on both flood extent and area as well
as water level data which can be used in model calibration
and evaluation, and in building and understanding model
structures.
Obtaining accurate flood extents and centimetre-scale
accuracies in water level estimation from remote sensing
that are within expected accuracies of model predic-
tions enables the modeler to evaluate and also improve
uncertain flood inundation predictions (Schumann et al.,
2009b). Acknowledging and examining the extent of
uncertainty in both data and model should be generally
accepted as a key element in flood risk management
exercises (Pappenberger and Beven, 2006), especially
when integrating uncertain spatially distributed obser-
vation data such as those derived from SAR remote
sensing with uncertain model structures. This integration
can be performed indirectly, through employing more
traditional model-data comparison techniques (see e.g.
Pappenberger et al., 2007 and Hunter et al., 2005), or
directly via assimilation into flood models (see e.g. Neal
et al., 2009; Matgen et al., 2010 and Giustarini et al., 2011).
Combining satellite data with hydrodynamic modelling
has now become established as a powerful approach, the
robustness of which needs, however, to be examined fur-
ther, in particular for flood forecasting. Schumann et al.
(2009b), who provide a detailed review on recent progress
in integration of remote sensing-derived flood parame-
ters and hydraulic models, argue that what is certainly
to be gained from this development is that fundamen-
tal research issues in terms of both model evaluation
and remote sensing data processing techniques will be
addressed in one way or another. In a very similar sense,
Cazenave et al. (2004) argue that scientists have much to
gain from current and future satellite observations and
missions to provide (global) hydrological data sets that
could be used to evaluate process models. Moreover, it
is expected that satellite measurements combined with
models that allow direct integration of such data would
6.4 Case study examples
The following three case studies are taken from the recent
scientific literature and demonstrate the value of low and
high resolution SAR imagery to support flood inundation
modelling. The case studies are complementary to the
many elements discussed in this chapter and are intended
to give the reader an appreciation of recent research
activities in flood mapping and uncertainty, flood map-
ping specifically applied to a rural-urban interface, and
flood mapping integrated with hydraulic modelling. More
specifically, the first case describes a rural application
where uncertainty in image data was used to improve
flood model parameter identifiability and also to map
associated flood risk. The second case reports a first
application of a very high resolution TerraSAR-X image
inside an urban setting, and finally, the third example
outlines the use of multi-temporal SAR images to map
floodplain wetting and drying patterns to help understand
hydrodynamic model limitations.
6.4.1 Fuzziness inSARflooddetectionto increase
confidence infloodmodel simulations
As noted earlier in this chapter, SAR-derived inundation
maps are typically treated as deterministic (i.e. binary
wet/dry classification) maps. However, these flood extent
maps are unavoidably affected by many sources of inac-
curacies. This example refers to a rural case study, a river
reach of the Lower Dee (UK), and aims at discussing
the uncertainties in SAR-derived flood area maps and
how these might be used to improve flood model cal-
ibration. This test site is of particular interest because,
during a flood in December 2006, both coarse resolution
(ENVISAT ASAR WSM) and medium resolution (ERS-
2 SAR) satellite imagery were acquired simultaneously
Search WWH ::
Custom Search