1 Spatial Modelling of the Terrestrial Environment - Georeferencial

82 Spatial Modelling of the Terrestrial Environment
designed with hydraulic model validation in mind and the low density means that very few
data points are available internal to a reach scale model domain. Whilst such data, particularly
stage, are accurate, they are only one-dimensional in time and zero-dimensional
in space and do not directly test a model’s ability to predict distributed hydraulics. Remote
sensing has the potential to supplement these existing point data sources with
two other, broad-area, types of data: vector data of inundation extent and flow velocity
fields.
Inundation Extent. In the late 1990s inundation data were seen as a potential solution to
the problem of hydraulic model validation as, whilst these were 0D in time, their 2D spatial
format could provide ‘whole reach’ data for both distributed calibration and validation of
distributed predictions. Inundation extent was also seen as a sensitive test of a hydraulic
model, as small errors in predicted water surface elevation would lead to large errors in
shoreline position over flat floodplain topography. The sensors available for this task are
reviewed by Pearson et al. (2001) so only a brief synopsis is provided here. The available
sensors are:
optical imagers on airborne and satellite platforms;
synthetic aperture radar (SAR) imagers on airborne and satellite platforms;
digital video cameras mounted on surveillance aircraft.
These are compared in Table 5.1.
Table 5.1 demonstrates that the use of satellite optical platforms for inundation extent
monitoring is problematic, apart from for large rivers and non-storm conditions. For example,
Bates et al. (1997) used a Landsat Thematic Mapper (TM) image of the Missouri
River to validate a two-dimensional finite element model of a 60-km reach under normal
flow conditions by comparing the extent of islands and mid-channel bars in both the model
predictions and satellite data. In parallel work, Smith et al. (1997) used suspended sediment
concentration to map flow patterns where a turbid tributary entered the relatively clear
water of the Missouri main stem and hence were able to qualitatively validate the velocity
patterns predicted by the model. Such studies would not have been possible during storm
conditions and could not be considered a rigorous test.
Rather better data can be acquired from optical airborne platforms as occasionally such
systems can be flown below the cloud base during flood conditions, particularly if the reach
in question is some way down the drainage basin and away from major flood generation,
and hence rainfall, areas. A limited amount of such data has thus been collected on an
ad hoc basis. For example, Biggin and Blyth (1996) acquired air photo data co-incident
with an overpass of the ERS-1 SAR satellite for a 1-in-5-year flood on the upper Thames,
UK. Other regulatory agencies, such as the Environment Agency in the UK, also possess a
limited number of air photo datasets of flooding. However, conversion of oblique air photos
to synoptic maps of inundation extent is by no means straightforward, and the quality of
such data is highly dependent on illumination conditions. As a consequence of these errors,
Horritt (2000a) estimated that the flood shoreline that could be derived from the Biggin
and Blyth (1996) data was only accurate to ±12.5 m, equivalent to the pixel size of ERS
SAR. Nor can acquisition of visual imagery under the storm conditions prevalent during
flooding episodes be guaranteed.