Image Utilisation for the Study and Management of Riparian Vegetation: Overview and Applications - Fluvial Remote Sensing for Science and Management - page 219

Table 10.4 Examples of recommendation in image choice for riparian vegetation studies.

Parameter Advice Reference

Wavelength Bands Three bands, centred within the wavelength regions of 0.53-0 . 54 μm,

0.66-0

Davis et al. (2002)

m. Where four-band sensors are
available, a preference for the fourth wavelength band centre should be
either at 0

. 67 μ m, and 0.79-0 . 815 μ

m wavelength region, where
different riparian species also display distinctly different amounts of
reflectance.

. 70 μ mornearthe2.02-2 . 07 μ

Spatial Resolution Image data should be acquired at a resolution of about 20 cm per pixel to

retain the textural information for typical riparian vegetation

Davis et al. (2002)

Scale of images Adapt the scale to the target, for example for aerial photos 1/60 000 to

1/25 000 to map stand vs. 1/16 000 to 1/12 000 to map species.

Brohman and Bryant
(2005)*

Time acquisition Near summer solstice in order to minimise shading within vegetation Davis et al. (2002)

A winter mission can give a good access to water areas hidden by the canopy
in summer, a distinction of the sempervirens vegetation (coniferous,
Juniperus), and information about cultivated areas

Girel (1986)

Differentiating between the targeted species and others can be easier in some
specific conditions such as the season in the year when the species
becomes senescent

Dipietro et al. (2002);
Hamada et al. (2007)

Multidate images to study phenological stages Muller (1995)

∗ not specific to riparian areas

(Bendix and Stella, in press). Remotely-sensed imagery is
commonly used at this scale by managers for some appli-
cations, including historical mapping, characterisation of
inaccessible sites, or repetitive evaluation in some cover
type areas. But is it always the best approach?
Firstly, the possibility to characterise something by
image analysis does not necessarily mean that it is the
best way to do it, especially for managers. Even if the
data from other approaches (e.g. field survey) are not
error-free, they can be less time-consuming to collect
and process later, less expensive to acquire and/or easier
to implement. Managers considering the use of remote
imagery must balance the highest level of accuracy pos-
sible (Table 10.2), with the effort needed to acquire and
process the information, particularly for small reaches
or for specific information such as vegetation struc-
ture. In such cases, it could be more efficient to choose
a field mapping approach. For example, Coroi et al.
(2006) used a GPS field mapping technique for small
streams and patches and considered it more efficient than
image analysis since current image resolution was not
high enough for them, and new image types were too
expensive. Concerning vegetation structure, basic remote
sensing approach usually provides coarse vegetation type
categories, typically 5 to 10 categories (Muller, 1997).
Obviously, it is possible to do a finer analysis on a local

scale. For example Johnson et al. (1995) along the Snake
River (USA) mapped 14 cover types from 1/7920 colour-
infrared photographs and Dieck et al. (2004) provided a
floodplain vegetation classification system for the Upper
Midwest region (USA) that distinguished 31 vegetation
types. Moreover, new tools such as object-oriented clas-
sification or combining images and LIDAR data now
enhance the mapping process and provide very useful
data (Boyd and Danson, 2005; Geerling et al., 2009).
However, to reach such detail, specific problems emerge
that can make it inappropriate for managers, including
intensive field verification, the high level of technicality
needed, the amount of time spent validating the results,
or low transferability from one site to another.
Secondly, some data that are frequently needed by man-
agers are difficult or impossible to collect through remote
sensing (Rheinhardt et al., 2007), including the precise
distribution of all tree species, tree age, composition of
herbaceous layer (notably woody species regeneration)
flood marks, overbank sediment thickness and grain size
and evidence of grazing by herbivores. For example, the
accuracy in distinguishing species is species dependent:
Nagler et al. (2005b) proposed an efficient framework
for cottonwood and willow characterisation but some
other species such as saltcedar are more difficult to iden-
tify. Hopefully, the new images available (higher spatial,