Geography Reference
In-Depth Information
is an excellent example of data mining used in fluvial
remote sensing. The toolkit allows users of bathymetric
LiDAR (see Chapter 7) acquired by the USGSto organise
and querry the large data volumes which are typical of
LiDAR datasets. Carbonneau et al. (2011) present results
obtained with a newly designed system: the Fluvial Infor-
mation System (FIS), which integrates a suite of cutting
edge procedures designed to analyse hyperspatial imagery
and manage the results in a spatially explicit framework
which is tailored to river environments. The aim of the
FIS is to apply the riverscape concept discussed by several
authors such as Fausch et al. (2002), Ward et al. (2002) and
Wiens (2002) in order to provide a continuous, spatially
explicit analysis of rivers which includes key variables
such as width, depth, particle size, and elevation. This
data therefore allowed Carbonneau et al. (2011) to derive
hydraulic information (velocity, stream power, Froude
number and shear stress) in order to provide parameters
for aquatic habitat characterisation and modelling over
the entire length of a small Scottish river.
channel and is mainly covered by willow and cottonwood
patches. Field reconnaissance showed significant crown
dieback of cottonwoods which, it was hypothesised, was
a link to water table depletion following a period of active
mining during the 1980s (Lejot et al., 2011).
Hyperspatial imagery was first acquired for the entire
reach with a 'Pixy' UAV (see Figure 8.3c). The Pixy
UAV is a flying paramotor developed by the French IRD
(Institut de Recherche pour le Developpement). It has a
paraglider-type nylon tubular wing from which the alu-
minium chassis is suspended. It has a light gas-powered
engine (7 kg) which allows for a maximum payload 4 kg.
It is launched like a plane and lands by gliding. Take-offs
require a flat open surface (asphalt or short grass) of at
least 40 m in length and 20 m in width. Since this very
rarely occurs in the vicinity of rivers, a takeoff track usu-
ally needs to be built. The pixy operates at speeds between
15 and 35 km/h. This low velocity allows high quality
image acquisition with minimal motion blur. Maximum
flight altitude is 800 m and maximum flight duration is
approximately 1 hour. The pixy is also equipped with
an onboard GPS (Garmin II
) which logs flights tracks.
Altitude, speed, and trajectory are all controlled by
radio-remote with live feedback sent by the aircraft to a
field laptop PC. Flight parameters (altitude, position, and
velocity) are recorded and used in order to generate meta-
data. The pixy UAV has been shown capable of collecting
data over wide spatial scales with both Lejot et al. (2007)
and Hervouet et al. (2011) collecting high resolution
imagery for river corridors of 5 km in length. However,
this range was not surveyed from a single launch point
and several such points were distributed along the reach.
This UAV dataset was supplemented with ULAV data
acquired from a manned paraglider (see Figure 8.3a).
Compared to the pixy, the paraglider has the advantages
of flying higher, up to 1500-1800 m thanks to a gas
powered engine and a 30 m
2
soft fabric wing. It requires
from 50 m to 400 m for take-off and can land in even
shorter spaces as small as 25m. Furthermore, this landing
area can be a relatively rough surface such as a gravel bar.
With a 20 litre tank of fuel, the paraglider can remain
airborne for 2-3 hours. From this position, the pilot uses
a camera equipped with a spirit level and manually takes
photographs. This imagery is usually at a higher altitude to
the UAV imagery and therefore provides a set of context
images of slightly lower resolution but of greater ground
footprints and total areas without the need to land and
re-launch.
In the data analysis phase, a continuous processing of
the hyperspatial imagery was found to be overly labour
+
8.4 From data acquisition to fluvial form
and process understanding
The last part of this chapter is focused on different
examples which illustrate the use of hyperspatial imagery
in practical applications. The examples are grouped under
two separate headings: (i) Feature detection with hyper-
spatial imagery, (ii) Repeat monitoring with hyperspatial
imagery.
8.4.1 Featuredetectionwithhyperspatial imagery
In this section we present two contrasting examples which
illustrate the opportunities presented by hyperspatial
imagery. The first example focuses on understanding
the dieback of cottonwood tree crowns observed along
the river Dr ome, France. The second example uses hyper-
spatial imagery in order to provide a baseline geometry
of secondary channels in a restoration project along the
Rh one River, France.
In the Dr ome case study, a specific procedure was
developed in order to assess the health of cottonwood
units based on UAV, ULAV, airborne and satellite images.
The study area was a 5 km reach downstream of Luc-en-
Diois. This reach drains a sub-catchment of 225 km
2
.The
reach has both single thread channels with average widths
of c.10m and larger braided reaches with widths in excess
of 200 m. The riparian corridor has similar width to the
Search WWH ::
Custom Search