Geography Reference
In-Depth Information
of features of specific interest, such as seabed morphology
(Finkl, 2005) or immersed objects (West and Lillycrop,
1999). These last authors reported a 100% probability for
detecting a 2-m immersed cube with a sampling density of
3 m x 3 m after Guenther et al. (1996). Full waveform sig-
nals would also allow the detection of mid-water objects
as long as their energy return is not mixed with volume
scattering or weakened by too large of a beam dispersion
at great depths.
Temporal studies based on a DEM comparison showed
the possible detection of moving sandbars (Irish and
White, 1998) or artificial objects such as a disposal mound
under three meters of water (Irish and Lillycrop, 1999).
Full waveform information can be used to derive not
only DEMs but also bottom roughness maps, focusing on
signal variability (Brock et al., 2004).
Finally, both amplitude and full waveform information
are processed to classify the seabed into different bottom
types (Wang and Philpot, 2007; Kuus et al., 2008; Collin
et al., 2010, Collin et al., 2011b). In these last studies, the
unmixing between contributions of depth, local bottom
shape, and acquisition configuration was solved through
statistical analysis of the full waveforms datasets, which
allows the classification of bottom types into either maps
describing vegetation cover or maps of seabed habitats.
The surveyed Gardon River section contained two
riffle-pool sequences and two different types of river
bottom. Upstream, there was a rough bedrock made
of marls, and downstream there was a smooth gravel
bed composed of silica gravels. Both sub-sections had
the same bottom albedo because periphyton covered the
entire reach. At the time of the survey, the water had a
Secchi depth of about 0.8 m, and the water velocity had a
maximum of 1.5 m.s
−
1
in riffles. The water depths ranged
up to 4.45 m in pools with a mean depth of 0.64 m, and
43% of its area had a depth less than 0.4 m. The mean
longitudinal slope of the reach was 0.12%. There was little
riparian vegetation on the right riverbank.
Concurrently with the LiDAR flight, tacheometry was
used to survey the polylines marking the riverbanks, and
dual frequency GPS was used to survey the river bottom. A
total of 5,443 topographic points of about 2-cm accuracy
for Z position inside the wetted riverbed were collected
and referenced. In depths beyond those that can be waded,
a tagline mounted on a prism onboard a boat, combined
to a tacheometer to accurately shoot the water surface
position, were used. Higher positioning errors can occur
in these higher depths (dispersion error of
5cmfor
one standard deviation on Z). A depth was attributed to
each of the reference points. The spatial sampling scheme
of the topographic points consisted of profiles across the
river (Figure 7.6a-A).
Approximately 50,000 Lidar X,Y,Z points on the river
bottom were obtained (depicted in colour as the Z val-
ues in Figure 7.6a-B). This set of points gave a mean
density of one point per 0.9 m
2
. Computing the accuracy
of data having a spatial location (interpolation problem)
and resolution (scaling problem) different from the ref-
erence data is not direct. For that reason, a method was
developed to compute LiDAR riverbed elevation data
bias and accuracy (random error) in comparison with
the reference and data limitations, i.e., the minimum
and maximum detectable depth (Bailly et al., 2010). This
method was based on geostatistics using anisotropic,
ordinary block-kriging (Atkinson and Tate, 2000) within
a curvilinear channel-fitted coordinate reference system
similar to the one proposed by Legleiter and Kyriakidis
(2006). This technique permits upscaling as well as the
interpolation of reference data, and it takes into account
the uncertainties in the results (the so-called weighted
errors).
Regarding the LiDAR accuracy, the computations for
river bottom elevation exhibited a bias depending on
water depth (Figure 7.6b-A). Added to that bias, a ran-
dom error with a 0.32-m standard deviation was found
∼
7.6 LiDAR experiments on rivers:
accuracies, limitations
7.6.1 LiDARfor rivermorphologydescription:
theGardonRiver casestudy
The Gardon River is a Rh one River tributary located in
the South of France. The Gardon flows from an elevation
of 1,450 m to 30 m over a 120-km length. It is a gravel-
bed river within a sub-humid Mediterranean hydrological
context, i.e., with strong rain events in autumn that alter
the river morphology.
A 1.5-km long reach located in the mid-reaches of
the Gardon River was surveyed in March 2007 with the
HawkEye II system mounted on an Aerocommander
690A airplane flying at a height of 250 m. It took 30
minutes to fly over this reach using three parallel and
overlapping strips, each having a width of about 110 m.
The LiDAR waveform data were processed using the
coastal survey studio suite, which provided X,Y,Z LiDAR
points on the wetted riverbed, the water surface, and
the riverbanks or vegetation canopy (Bailly et al., 2010).
Below, only the wetted riverbed points are considered.
Search WWH ::
Custom Search