Geography Reference
In-Depth Information
a riverscape appears to be suitable for bathing, strolling,
and children's play.
Following this first step, an analysis of the structure
of the landscape was performed to improve the inter-
pretation of these findings (Figure 18.6). The Roubion's
riverscape patches were statistically analysed in order to
distinguish different fluvial landscape types (Cossin and
Piegay, 2001). A series of variables describe 34 waterscapes
(Figure 18.6b). Each label identified the surface area occu-
pied by a landscape features (e.g. B
aesthetics of 34 photographs of floodplain lakes. These
photographs were sampled so that the diversity of the
floodplain lakes of the Ain River was represented as
much as possible. The visual variables characterising the
physical aspects of the waterbodies were selected based
on a previous study that focused on the perception of
floodplain lakes (Cottet et al. 2010), as well on bibliogra-
phy, enabling to select the most discriminating variables
to distinguish positive and negative judgments. Finally,
the modelling relies on six visual physical variables: (1)
green dominance, (2) grey or brown dominance, (3)
presence of warm and bright colours, (4) presence of
a badly structured aquatic vegetation, (5) presence of
sediments, and (6) muddy water. Each photograph was
characterised according to these variables and a multiple
correspondence analysis (MCA) was realised.
Strong correspondences are observed between vari-
ables (60% of the variance explained) (Figure 18.7). Axis
F1 shows above all the information on the colour of
the waterbodies; whereas axis F2 is structured by the
surrounding objects (sediments, aquatic vegetation). The
regression analysis was then based on the coordinates on
the 2 axes of the 6 variables selected: y
=
block, LW
=
lotic
water, and G
) in one portion of the photo-
graph (1 for the foreground, 2 for the mid-ground
=
gravel
...
). A
normalised Principal Component Analysis (nPCA) was
performed to determine a smaller number of variables.
The correlation circle of the first factorial plan provides
a graphic synthesis of the results (Figure 18.6a). On the
first component (F1), the open landscapes showing plains
(Pl) tend to be opposed to the closed landscapes of gorge
(Sl). The second component (F2) distinguishes on the
one hand lentic sections (1LE) with meadows (2M) and
semi-closed riverine forest (1TW and 2TW) and on the
other hand lotic sections with a background of forest
(3TU). Preferences for the nine waterscapes can be com-
pared with the coordinates of axis F1 (Figure 18.6c). A
clear relationship is then established between the aesthetic
score and the first factorial coordinate. There is a strong
preference for the closed waterscapes of gorges, contrary
to the open waterscapes of plains. This conclusion sup-
ports the analysis of responses to the open questions of
the questionnaire.
...
=
4.7
+
2 (F1 axis
coord.)
0.9 (F2 axis coord.). Several conclusions about
perception mechanisms can be drawn from the resulting
linear regression. The more the colour green dominates
and the more the warm and bright colours are present,
the more the waterbody is judged to be aesthetic. On the
contrary, the muddier the water, the more the grey or
brown colours dominate, and the more the badly struc-
tured aquatic vegetation are present, the less aesthetic the
waterbody is judged. The influence of sediments on the
perception is more uncertain. These results are rather
encouraging: 2/3 of the preferences are explained by the
model (r
2
+
18.4.4 Modellingandpredictingwater landscape
judgments
The last study is focused on modelling the public's aes-
thetic preferences of the waterbodies of floodplain lakes,
within the context of the ecological restoration of the
Ain River. It aims to predict the aesthetic assessments of
different waterbodies from a set of qualitative visual vari-
ables. The choice of using visual variables is due to several
operational stakes: it enables the model to be used by any
practitioner, whether they are ecological experts or not.
As in the previous study, the model consists in a factorial
regression analysis: the dependent variable is the mean
aesthetic grade given by the people surveyed; the inde-
pendent variables are the physical visual characteristics of
the waterbodies.
A photo-questionnaire survey was conducted in order
to obtain data concerning the aesthetic preferences of the
public: 100 students in geography were asked to assess the
66). Moreover the validation step, using
the leave-one-out method, showed the robustness of the
model (Figure 18.8): its power of generalisation can be
considered further. Such a model may be an efficient tool
in order to favour dialogue between stakeholders.
=
0
.
18.4.5 Photographsand landscapeperception, a
longhistoryof knowledgeproduction
The examples developed above illustrate a long collective
effort of knowledge production in this domain. Different
experiments have been published showing that humans
react differently to the amount of water and its character-
istics, but also to the type of waterscapes, the openness of
the landscape or the riparian characteristics and the level
of naturalness.
Search WWH ::
Custom Search