1 Spatial Modelling of the Terrestrial Environment - Georeferencial

88 Spatial Modelling of the Terrestrial Environment
there are acknowledged mis-classification problems (Fuller et al., 1990, 1994a, 1994b)
with such data, they are consistent, high resolution and have complete national coverage
that is likely to be repeated on an eight-year cycle. As yet no research has been conducted
to associate a hydraulic resistance, or range of resistance, with each vegetation class. However,
such a development would be relatively straightforward and rapid to accomplish and
could provide a first-order estimate of floodplain frictional resistance for the whole of the
UK at 25 m resolution.
Vegetation Biophysical Attributes. Whilst the association of vegetation classes with typical
hydraulic resistance values could be seen as an extension and standardization of the
Chow (1959) typography of river channels, other remote sensing techniques offer the
potential to directly measure vegetation biophysical parameters such as vegetation height,
density and stiffness. These measures may correlate with frictional resistance (Temple,
1987) and hence potentially provide a physically based method of determining spatial
variable resistance coefficients that overcomes the subjective nature of typographic approaches.
However, such approaches require further work over the medium term before
they can be used operationally.
To date, most work in this area has concentrated on the determination of vegetation height
as part of the standard processing chain for LiDAR data. The problem in processing LiDAR
data is how to separate ground hits from surface object hits on vegetation or buildings.
Ground hits can be used to construct a digital elevation model (DEM) of the underlying
ground surface, while surface object hits taken in conjunction with nearby ground hits
allow object heights to be determined. Papers by Mason et al. (1999), Cobby et al. (2000)
and Cobby et al. (2001) describe the development of a LiDAR range image segmentation
system to perform this separation (see Figure 5.1), and this is reviewed briefly here.
The system converts the input height image into two output raster images of surface
topography and vegetation height at each point. The river channel and the model domain
extent are also determined, as an aid to constructing a model discretization. The segmenter is
semi-automatic and requires minimal user intervention. Flood modelling does not require a
perfect segmentation as it involves large-area studies over many regions, so a straightforward
segmentation technique able to cope with large (∼Gb) datasets and designed to segment
rural scenes has been implemented.
The action of the segmenter can be illustrated using a 6 × 6 km sub-image from a large
140 km 2 LiDAR image of the Severn basin. The data were acquired in June 1999 by the
UK Environment Agency using an Optech ALTM1020 LiDAR system measuring time of
last return only. At this time of year leaf canopies were dense, so that relatively few LiDAR
pulses penetrated to the ground. The aircraft was flown at approximately 65 m s −1 and at
800 m height whilst scanning to a maximum of 19 ◦ off-nadir at a rate of 13 Hz. The laser
was pulsed at 5 kHz, which resulted in a mean cross-track point spacing of 3 m. Because
current one- and two-dimensional flood models are capable of predicting inundation extent
for river reaches up to 20 km long, with floodplains spanning associated large areas, the
segmenter has been designed to use LiDAR data with lower resolution than that used
for other applications (Gomes Pereira and Wicherson, 1999). Figure 5.2a shows (on its
left-hand side) the raw 3-m gridded sub-image used as input to the segmenter. The subimage
contains a 17-km reach of the Severn near Bicton west of Shrewsbury. The bright