<br />(II t, .
<br />0110068
<br />
<br />1790
<br />
<br />BROW,.. AND DANIEL: LANDSCAPE AESTHETICS OF RIPARIAN ENVIRONMENTS
<br />
<br />about 9% of the recreational trips, while less than 1% were
<br />for floating in a raft or kayak. The most popular activities
<br />were sightseeing, picnicking, and hiking, in that order. We
<br />have no information about the Tucson respondents' famil-
<br />iarity with rivers in the Rocky Mountains, but it is safe to
<br />assume that only the rare observer was familiar with the
<br />Poudre River.
<br />Group-to-group reliability [Ebel, 1951] is of primary inter-
<br />est in this study because it focuses on group mean prefer-
<br />ences, rather than individual observer differences. Group-to-
<br />group reliability indicates the expected correlation between
<br />panels of the same size as the sample panel, drawn from the
<br />same population. The reliability coefficients were 0.81 (n =
<br />27). 0.87 (n = 32), and 0.89 (n = 26) for the three Tucson
<br />pbOIO sets and 0.91 (n = 34),0.91 (n = 43), and 0.94 (n = 37)
<br />for the three Fort Collins sets. These reliability coefficients
<br />indicate substantial agreement among observer groups about
<br />the relative scenic beauty of the scenes but somewhat more
<br />agreement in Fort Collins than Tucson. The coefficients are
<br />similar to those we have obtained for scenic beauty ratings of
<br />forest scenes [Brown and Daniel, 1987].
<br />The correlations across photo sets of SBEs for the base-
<br />line scenes, which were common to all observer groups,
<br />provide another indication of the level to which the groups
<br />agreed about the relative scenic beauty of the scenes.
<br />Correlations among the three Tucson scene sets are 0.85,
<br />0.86, and 0.87, and the three correlations for the Fort Collins
<br />scene sets are 0,83, 0.87. and 0.87, These baseline correla-
<br />tions are similar to those for forest scenes obtained from
<br />groups viewing color slides [Daniel el aI., 1989]. .
<br />Paired-comparison experiment. Eight views for which
<br />relatively good photographic representations were available
<br />for four different flow categories were chosen for presenta-
<br />tion to observers in pairs, each pair showing the same view
<br />at two different flow levels. The flow categories and rheir
<br />nominal flow levels were low (152 cfs (4.30 ml/s)), medium
<br />low (479 or 715 cf, (13.6 or 20.: mJ/s)). medium high (1202
<br />cfs (34.04 mJ/s)), and high (2304 or 2642 cfs (65.24 or 74.81
<br />m lIs)). All six possible pairs of the four flow categories were
<br />shown for each of the eight views, requiring 48 paired
<br />comparisons. Further, to lessen the flow-oriented focus
<br />somewhat, 28 additional pairs of different views at the same
<br />flow level (medium low) were randomly interspersed among
<br />the same-view pairs. for a total of 76 comparisons.
<br />The pairs of scenes were shown to two Tucson student
<br />observer groups, of 15 and 19 participants, for 12 s each on
<br />a pair of monitors placed side by side. The general introduc-
<br />tion to observers about the purpose of the study was the
<br />same as that for the rating experiments, but the specific
<br />instructions were quite different. Observers were instructed
<br />to "decide which of the two scenes is more attractive."
<br />They were further informed that, for many of the compari-
<br />sons,
<br />
<br />you will see essentially the same river scene on both screens.
<br />This is because these areas were filmed al several different
<br />times during the past summer. One of our concerns in this
<br />research is whether the scenic quality of river areas changes
<br />with the sea~ons; and even during (he course of one summer
<br />there are subtle variations In such variables as the flow rate of
<br />the river. the amount of sun-ounding vegetation. and the color
<br />of the vegetation that might influence perceplions of scenic
<br />quality,
<br />
<br />Thus the instructions tended to soften the emphasis on flow
<br />level by mentioning vegetation differences as well.
<br />Each group of observers' preferences was scaled to Z
<br />scores according to Thurstone's law of comparative judg_
<br />ment, as outlined by Torgerson [1958] and reviewed by Hull
<br />et al. [1984]. This is a common scaling procedure for paired
<br />comparisons and has been used extensively in other scenic
<br />quality studies [e.g., Buhyoff and Leuschner, 1978; Buhyoff
<br />and Wellman, t980]. The resulting Z scores provide an
<br />interval scale of preference. The Z scores of the two ob-
<br />server groups were averaged, and the averages were multi-
<br />plied by 100 to eliminate decimal~, making them similar in
<br />magnitude to SBEs. These adjusted Z scores are called
<br />paired-comparison estimates (PCEs) herein.
<br />
<br />RELATIONSHIP OF FLOW TO SCENIC BEAUTY
<br />
<br />Plots of SHE versus flow for the two rating experiments
<br />show a wide range in scenic beauty among scenes at any
<br />given flow level, reflecring the large differences in vegeta-
<br />tion, topography. view perspective. depth of view. weather,
<br />and other features among the scenes. However. within all
<br />that diversity, a concave relationship is noticeable, as scenic
<br />beauty initially incr-eases with flow and then decreases as
<br />flow continues to increase. Similarly, a plot of PCE from the
<br />paired-comparison experiment verses flow for the eight
<br />views each shown at four flow levels shows a wide range in
<br />scenic beauty across the sites at a given flow level, but a
<br />concave relationship of scenic beauty to flow is evident.
<br />Furthermore, the concave relationship is somewhat more
<br />pronounced than it is for the rating experiments, which is
<br />expected given the greater emphasis on flow in the paired-
<br />comparison format and the omission of the scenes showing
<br />the most cloudy weather.
<br />As discussed above, flow rate was estimated from USGS
<br />records. Numerous other variables were estimated by visual
<br />inspection of the Video scenes. These variables. listed in
<br />Table 2, include characteristics such as percent of the scenes
<br />in rock, sky, and water, distance of the view, color of the
<br />water, distance ofthe camera from the stream channel, and
<br />color trueness of the photography. Four observers indepen.
<br />dently estimated each variable. and the mean of their judg-
<br />ments was computed. The reliability coefficients for these
<br />means ranged from 0.47 to 0.98, with a median of 0.91,
<br />indicating generally good relative agreement among the four
<br />observers. The lowest reliability coefficient was for the
<br />percent of the scene showing ex.posed river bed (BED, Table
<br />2), which tended to be a very small percentage of rhe scene
<br />and thus difficult to precisely estimate. The nex.t lowest
<br />coefficient was 0.72 for photographic lighting accuracy
<br />(PHOTOL), which is a relmively subjective variable.
<br />The relationship of flow to scenic beauty was assessed by
<br />regressing the scenic beauty measure on variables represent-
<br />ing scene features, using slepwise least squares. Because
<br />plots indicated a concave relationship of scenic beauty to
<br />flow, flow was represented by simple and squared terms in
<br />the regressions. Other independent variables were included,
<br />as described in Table 2. Separate models were produced for
<br />the three experiments.
<br />The regression models of each experiment included both
<br />flow terms and supported Litton's (1984) hypothesis that
<br />scenic beaulY of streams is diminished at both very low and
<br />abnormally high flow levels, SSE reached a maximum 3t
<br />
<br />
<br />N"
<br />
<br />Indepe:J
<br />FLO
<br />SOIt
<br />
<br />VEG
<br />
<br />ROC
<br />
<br />SKY
<br />WAl
<br />BED
<br />
<br />VIE'
<br />
<br />sou
<br />
<br />WIN
<br />WID
<br />
<br />WC(
<br />
<br />L1G'
<br />
<br />PHC
<br />
<br />PHC
<br />
<br />HDJ
<br />
<br />VDI
<br />
<br />-All
<br />sequer
<br />
<br />1092
<br />06,3\
<br />maxir
<br />value
<br />of flo
<br />rating
<br />exper
<br />about
<br />mJ/s)
<br />nic b~
<br />reach
<br />All
<br />error
<br />beaul
<br />thai t
<br />equal
<br />dUre.
<br />Th
<br />perio
<br />prefe
<br />betw
<br />klos
<br />ftow~
<br />reSer
<br />read'
<br />0,
<br />
<br />
|