Professional Documents
Culture Documents
Ecological Indicators
journal homepage: www.elsevier.com/locate/ecolind
Original Articles
A R T I C LE I N FO A B S T R A C T
Keywords: Enjoyable landscapes are important resources for recreational activities and the socio-economic development of
Spatial indicators tourism destinations. A profound understanding of landscape preferences can support landscape management
Perception survey and planning. Despite the increasing integration of the socio-cultural perspective in landscape preferences re-
Landscape metrics search, little is known about the links between landscape characteristics and individual landscape preferences. In
Scenic beauty
this study, we aimed to estimate landscape preferences at the individual level based on a set of landscape
Aesthetic value
indicators, allowing us to measure the preferences of each person. We thereby evaluated the suitability of
conjoint analysis to identify the relative importance of selected landscape indicators and the corresponding part-
worth utilities of their characteristics. We further examined whether the preferences are homogeneous or if we
can identify groups with largely different preferences. We related the picture ratings from a photo-based survey
of landscapes in the Central Alps to a set of 11 landscape indicators, measuring the landscape pattern and
features of each picture. Each indicator was divided into two or three levels and used to calculate importance
scores and part-worth utilities by hierarchical Bayes analysis for individuals. In our study area, 11 indicators
were sufficient to predict the individual choice between two landscapes for ∼90% of the respondents. Our
results indicate non-linear relationships between some landscape indicators and landscape preferences and re-
vealed considerable heterogeneity for the vectors of part-worth utilities, suggesting some methodological pro-
blems when applying aggregated linear prediction models. Our findings may therefore enhance predictive
models and support landscape planning and management, but further research is necessary to understand the
driving forces behind the observed differences.
1. Introduction et al., 2007). Despite lower rates of land-use changes, forest regrowth
will remain an issue due to slow succession (Tasser et al., 2017) but is
Mountain environments provide many opportunities for enjoying expected to transform patchy landscape mosaics into more homo-
nature and practising outdoor recreational activities (Schirpke et al., geneous patterns through the closing of open pasture patches
2016, 2018). In addition to the positive effects on human physical and (Garbarino et al., 2014). The result is fewer viewpoints and reduced
mental health (Romagosa et al., 2015; Triguero-Mas et al., 2015), an landscape diversity, which leads to lower aesthetic landscape values
appealing landscape contributes to socio-economic development by (Schirpke et al., 2013b; Weinstoerffer and Girardin, 2000). In the light
attracting tourists (Bonzanigo et al., 2016; Chhetri et al., 2004). How- of these current and expected landscape changes in mountain regions, it
ever, the landscape in the European Alps is changing due to past and is important to understand human preferences for landscape char-
current land-use changes, which include intensification of agricultural acteristics to be able to evaluate influencing factors and adapt land-
use and increasing urbanisation in favourable areas in the valley bot- scape management and planning.
toms, as well as the abandonment of steep meadows and subalpine and Landscape preferences involve both the biophysical characteristics
alpine pastures (Egarter Vigl et al., 2016; Price et al., 2015). Subsequent of the natural environment and human perceptions and have been as-
reforestation processes on abandoned grassland have already increased sessed through a variety of approaches based on different underlying
forest cover in many parts of the Alps (Schneeberger et al., 2007; Tasser concepts. An important conceptual distinction can be made between
⁎
Corresponding author at: Sternwartestraße 15, 6020 Innsbruck, Austria.
E-mail address: uta.schirpke@uibk.ac.at (U. Schirpke).
1
Joint first authors.
https://doi.org/10.1016/j.ecolind.2018.09.001
Received 10 January 2018; Received in revised form 20 August 2018; Accepted 1 September 2018
1470-160X/ © 2018 Elsevier Ltd. All rights reserved.
U. Schirpke et al. Ecological Indicators 96 (2019) 202–212
203
U. Schirpke et al. Ecological Indicators 96 (2019) 202–212
Fig. 1. Location of the study sites, land cover distribution (year 2000) and surface area of the four study sites. Lech Valley (1) includes the municipalities of Gramais,
Hinterhornbach, Pfafflar and Stanzach. Stubai Valley (2) contains the municipalities of Neustift im Stubai and Fulpmes. Puster Valley (3) comprises the municipalities
of Gsies, Rasen-Antholz, Sand in Taufers and Prettau. Vinschgau (4) includes the municipalities of Glurns, Graun im Vinschgau, Mals and Schluderns.
204
U. Schirpke et al. Ecological Indicators 96 (2019) 202–212
2.4. Landscape indicators estimates of individual values by combining the individual’s informa-
tion with the complete sample. Thereby, initial estimates of the prob-
Spatial landscape indicators can be used to measure the visual abilities can be revised using information provided by the data to obtain
properties of landscapes (Ode et al., 2009). To describe the landscape a posterior probability estimation that utilizes both the initial in-
pattern and features of the 24 pictures from the survey, we used a set of formation and information from the data (Andrews et al., 2002); for
indicators (Table S2 in the Supplementary material) which were iden- technical details see Lenk et al. (1996) or Sawtooth Software (2009).
tified in previous studies by Schirpke et al. (2013b, 2016) to best ex- Individual part-worth utilities and importance scores were estimated in
plain landscape preferences. Pictures 9, 23 and 24 were excluded as Light House Studio 9.0. First, 10,000 preliminary iterations were run,
they represented inner forest patches which could not be described followed by another 10,000 random draws per respondent. Relative
through landscape indicators. As most of the indicators could not be importance scores reflect the relative effect each indicator has on the
derived directly from the pictures, we applied the GIS-based approach landscape perceptions in per cent, but it does not provide information
developed by Schirpke et al. (2013b) to represent the visible landscape on the effect of each level of the indicator on the overall index. In
in the pictures by maps. First, a visibility analysis was performed in contrast, the part-worth utility of an indicator level represents the
ArcGIS 9.3TM (ESRI, Redlands, CA, USA) to identify the visible area preference for that level, with a higher value indicating a greater pre-
from the position where each picture was taken using a digital surface ference. For ease of interpretation, part-worth utilities were zero-
model. To account for the effect of distance on the perceived size and centred and the sum of the differences between the minimum and
colour of landscape elements in mountain regions (Germino et al., maximum part-worth utilities of each indicator equals 100.
2001), we distinguished three distance zones (near zone 0–1.5 km, The fit of the conjoint model was interpreted using the root like-
middle zone 1.5–10 km, far zone 10–50 km). Long vistas from mountain lihood (RLH) values. RLH values range from 0 to 1 for each person.
peaks were limited to 50 km, which corresponds to the range of visi- Considering only two landscapes in each choice, a RLH value of 0.5
bility outside population centres in Europe (Horvath, 1995). The visible means the choice is only by chance, whereas a value of 1 means that the
areas for each zone were overlaid with land cover datasets that differed choice is completely determined by the landscape indicators of the two
in number of land cover classes and spatial resolution (Table S2 in the landscapes. The RLH is therefore a measure of the consistency of the
Supplementary material). The three resulting maps were recomposed choice process at the individual level.
into a single map with a maximum number of 49 land cover classes and
a spatial resolution of 20 × 20 m. This map was used to calculate all 2.6. Testing the homogeneity of preferences
indicators using FRAGSTATS version 3.3. (McGarigal et al., 2002). The
indicators included landscape metrics, which measure spatial char- Several studies (Beza, 2010; Dramstad et al., 2006; Hunziker et al.,
acteristics related to landscape composition and configuration 2008; Schirpke et al., 2013a; Tveit, 2009) found differences between
(McGarigal et al., 2002) as well as indicators describing the presence or social groups using socio-demographic variables such as gender, age,
absence of specific features such as water, forest, settlements and road cultural background (language group), and level of education among
infrastructure. Further details can be found in Schirpke et al. (2013b, others. We therefore tested both the importance scores and the part-
2016). worth utilities for significant differences based on available information
For the conjoint analysis, the values of each indicator were first related to gender, type of visitor, nationality, age and size of town. As
divided into two (present or absent) or three levels (high, medium or this resulted in a number of significant differences, which, however,
low level) to guarantee robustness related to uncertainties in the cal- were difficult to interpret due to the high number of tests performed,
culation or perception thresholds of respondents (Sayadi et al., 2009; we combined a cluster analysis with a discriminant analysis.
Van Cauwenberg et al., 2016; Veitch et al., 2017). To obtain three levels In a first step, to analyze the homogeneity of the preferences, part-
of the continuous landscape metrics, three groups of equal size (terciles) worth utilities were used to group the respondents according to their
were identified based on the value ranges of each indicator based on all preferences by hierarchical cluster analysis with the SPSS Statistics
photos, whereas the landscape features already had two classes (present software. We used the squared Euclidean distance to measure the dis-
or absent). All photos were then assigned the corresponding levels of similarity of the preferences of two persons and applying Ward’s
the indicators. linkage method to aggregate the clusters. The agglomeration coeffi-
To select only indicators that contain enough independent in- cient, measuring heterogeneity as the distance at which clusters are
formation, we based our selection on the coefficient of determination formed, was examined to identify the number of clusters that are too
R2, which measures the common linear information content. An different to form a homogeneous group (Hair et al., 2010). We decided
R2 < 0.5 means that two indicators have less than 50% of the in- to use the four-cluster solution, as the highest leap of heterogeneity
formation in common and corresponds approximately to an R < 0.7. change occurred between four and five clusters.
Therefore, from each pair of indicators with a Pearson’s R < 0.7 In a second step, we calculated average importance scores and part-
(Virgili et al., 2017), one was excluded from further analysis. In this worth utilities as well as 95% confidence intervals for each group. We
way, the problem of multicollinearity was avoided and the information applied a discriminant analysis based on the available socio-demo-
content of the set of indicators was preserved to a large extent. We graphic variables to examine whether membership in one of the four
finally obtained a set of 11 indicators out of 20 (Table 1), which were groups can be explained by socio-demographic differences (gender,
used for further analysis. type of visitor, nationality, age and size of town).
We first converted the ratings of the 967 respondents into rankings 3.1. Influence of landscape pattern and features on landscape preferences
for pairwise comparison of landscapes to assess preferences between
two distinct alternatives, resulting in 210 pairwise comparisons per The mean values of the pictures varied between 4.29 and 8.26
respondent. All pairs with identical ratings (about 37,000 ties), and (Schirpke et al., 2016; Table S4 in the Supplementary material). Re-
therefore no clear preference, were excluded, resulting in 163,773 spondents attributed very high values to pictures representing sub-
pairwise comparisons. alpine and alpine landscapes with a diversity of landscape elements
Based on this dataset, a conjoint analysis, using hierarchical Bayes (pictures 6 and 14), whereas pictures including settlements, infra-
analyses, was performed to assess the relative importance scores and structure or intensive agricultural use were rated lower (pictures 7, 13
part-worth utilities (Rao, 2014). This method can be used to derive and 22).
205
U. Schirpke et al. Ecological Indicators 96 (2019) 202–212
Table 1
Spatial landscape indicators and corresponding levels used for the conjoint analysis.
Type Indicators Description Levels
Landscape metrics Number of patches (NP) NP measures the total number of patches in the landscape. Larger values indicate that the Low, medium or
landscape is more fragmented. high
Patch richness (PR) PR indicates the number of different patch types across the landscape. Larger values mean
higher landscape diversity.
Area-weighted mean shape index SHAPE_AM measures the complexity of patch shape compared to a standard shape (square)
distribution (SHAPE_AM) of the same size. A larger value of SHAPE_AM indicates that the patches in the landscape are
more complex and irregular in shape.
Median shape index distribution SHAPE_MD represents the midpoint of the rank order distribution of patch metric values
(SHAPE_MD) based on all patches in the landscape. A larger value of SHAPE_MD means a higher level of
complexity.
Coefficient of variation shape index SHAPE_CV describes the variation of shape complexity across the landscape. Larger values
distribution (SHAPE_CV) mean a higher level of variation of shape complexity.
Total area (TA) TA represents the total visible area. Larger values indicate a greater view depth.
Near zone (TA_N) TA_N measures the visible landscape area within a distance of 1.5 km from the observer.
Larger values indicate that more area is visible in proximity to the observer.
Landscape features Urban Presence of urban settlements in the picture. Absent or present
Road Presence of road infrastructure in the picture.
Forest Presence of forest in the picture.
Water Presence of water in the picture.
Fig. 3. Distribution of RLH values. shaped (Shape_CV and TA) and inverted u-shaped (Shape_AM and
Shape_MD). In monotonic relationships, there is no change of direction,
The mean RLH of the model is 0.93 with a standard deviation of i.e. the function is always increasing or decreasing, whereas the re-
0.065 and a minimum of 0.61. For 80% of the sample, the RLH is lationship is non-monotonic if the direction switches from increasing to
greater than 0.90, and 60% show a RLH > 0.95. Only 12% of the re- decreasing (Rao, 2014). In addition to the visual interpretation of the
spondents have a RLH < 0.85 (Fig. 3). This demonstrates that for part-worth utilities, this can be tested by calculating the differences
∼10% of the sample, the fit of the model is modest, but for 90%, the between the different levels and applying a T-test for significance
choices are very well predicted for a model based on the 11 indicators. (Netzer and Srinivasan, 2011). In case the difference between low and
With the exception of Water and Forest, the distribution of the medium levels has a different sign than the difference between the
importance scores is nearly balanced between the landscape indicators. medium and high levels and both differences are significant, the re-
Forest had the greatest influence on landscape preferences, followed by lationship is significantly non-monotonic. For TA, Shape_AM, Sha-
SHAPE_MD, TA, PR and SHAPE_AM, for which the importance scores pe_MD, Shape_CV and TA1, the results show highly significant non-
did not significantly differ from each other (Fig. 4). The importance of monotonic relationships for TA, Shape_AM, Shape_MD, Shape_CV and
SHAPE_CV was significantly lower than SHAPE_AM, but not much TA1 (Table S5 in the Supplementary material). As linear variables in-
lower than NP. Importance scores decreased further significantly for crease (or decrease) in the same direction at the same rate, this im-
Road and TA_N, but not for Urban from TA_N. Water had the lowest plicitly implies nonlinearity (Yitzhaki and Schechtman, 2012).
importance of all indicators. Because of the large sample size, small A less fragmented landscape (low NP) had significantly higher part-
differences were statistically significant, but it is questionable whether worth utilities than a medium or highly fragmented landscape. In
a difference of two per cent points in important scores should be in- comparison, a low landscape diversity (PR) had significantly lower
terpreted as substantial. part-worth utility than high landscape diversity. Medium complexity of
Fig. 5 illustrates the mean part-worth utilities for each landscape patch shape (SHAPE_AM) was preferred over simple or very complex
indicator. The levels of the landscape indicators were related in dif- shapes. Accordingly, a medium level of SHAPE_MD had the highest
ferent ways to landscape preferences, as the relationships include part-worth utility compared to a low or high level of patch shape
monotonic increasing (PR), monotonic decreasing (NP and TA_N), u- complexity. Higher variations of patch shape complexity (SHAPE_CV)
were more positively perceived than low variations. Medium patch
206
U. Schirpke et al. Ecological Indicators 96 (2019) 202–212
Fig. 5. Box-plots of part-worth utilities of the different levels of landscape indicators. The line in the middle of the boxes indicates the median. The bottom of the box
represents the 25th percentile, the top of the box indicates the 75th percentile. Whiskers extent to minimum and maximum values. Part-worth utilities should be
compared within one indicator (not across indicators).
207
U. Schirpke et al. Ecological Indicators 96 (2019) 202–212
Water
Urban
TA_N
Road
SHAPE_CV
NP
SHAPE_AM
PR
TA
SHAPE_MD
Forest
0 2 4 6 8 10 12 14 16
Importance (%)
shape complexity was negatively related to landscape preferences. Low are not able to explain the differences in the landscape preferences of
and high TA had significantly higher part-worth utilities than medium different social groups.
TA, whereas low and medium TA_N were preferred with respect to high
TA_N. The presence of urban structures, roads and forest had a negative 4. Discussion
influence, whereas water in the landscape increased part-worth utility.
4.1. Insights into landscape preferences provided by conjoint analysis
3.2. Differences between groups
This study confirms that conjoint analysis is suitable to predict
The respondents were grouped based on their preferences (part- landscape preferences of individuals with a limited number of land-
worth utilities) by hierarchical cluster analysis into four groups (Group scape indicators. Our results demonstrate that the choice of one land-
1: n = 384, Group 2: n = 270, Group 3: n = 213 and Group 4: scape over another is rather consistent among individuals (RLH > 0.9)
n = 100). The differences between the importance scores of the four and the estimation of the choice is limited for only 10% of the re-
groups ranged from small (Water: 1.4%) to medium (Ta_1, NP: ∼3%) to spondents. These choices can be reproduced with only 11 indicators,
important (PR, Road, Forest: > 5%), indicating that the preferences which were not derived from the pictures (e.g., Arriaza et al., 2004;
were not homogeneous between the four groups (Fig. 6). Junge et al., 2015) but calculated using a digital surface model and land
The importance scores capture only a part of the information re- cover maps, allowing us to calculate the aesthetic value of a landscape
garding preferences. Thus, part-worth utilities offer a deeper insight based only on secondary available data.
into the preferences and their homogeneity between the four clusters. By applying conjoint analysis a posteriori, several limitations of
The part-worth utilities of the four groups followed similar trends for conjoint analysis could be reduced (Alriksson and Öberg, 2008). We
SHAPE_MD, SHAPE_AM, TA (except for Group 1) and NP (except for relied on real instead of hypothetical landscapes created by researchers.
Group 4) (Fig. 7). The greatest differences occurred for Road, PR and We used a sound study design thanks to a clear and easy to understand
SHAPE_CV, demonstrating partially opposed preferences. For example, photo-based questionnaire and obtained a large sample. Further, we
the presence of roads was perceived positively by Groups 2 and 3, included landscape indicators that might be difficult to explain to re-
whereas Groups 1 and 4 preferred pictures without roads. Group 1 spondents. Studies applying conjoint analysis usually include only a few
differed from the other groups mainly in the perception of SHAPE_MD, different attributes to not burden the respondents with too many
TA and Urban, whereas the differences between the different levels of choices. This study examined 11 different indicators and therefore
SHAPE_AM and PR were more strongly pronounced in Group 2. Group 3 comprised more landscape indicators compared to other studies
was the only group that attributed high positive values to the presence (Álvarez-Farizo and Hanley, 2002; Sayadi et al., 2009; Tempesta et al.,
of Forest, whereas Group 4 demonstrated the greatest variations mainly 2010).
for PR, NP and TA_N. In the following, we compare our results with those of other studies
For Shape_AM, Shape_MD and TA, the relationship between land- to discuss the reliability of the approach. Forest was an important in-
scape indicators and preferences differed only a little among the four dicator and most people demonstrated a preference for open land-
groups, whereas important differences in the relationship occurred for scapes. This result is in line with other studies (Tveit et al., 2006) which
PR, NP, TA_N, Road and Forest (Fig. 7). The relationship monotonically related a higher degree of openness to higher landscape preferences,
decreased for some groups, but monotonically increased for other although forest is an appreciated landscape feature (Pastorella et al.,
groups, revealing strong heterogeneity of landscape preferences. 2017). Substantial landscape changes due to natural reforestation
We tested, by linear discriminant analysis, whether the membership processes may therefore result in reduced aesthetic appreciation of the
of the four groups can be explained by socio-demographic variables. landscape (Hunziker and Kienast, 1999), particularly at higher eleva-
Although the F-values were significant for gender and Italy, the results tions (Soliva et al., 2010). In comparison to forest, the other landscape
did not show a satisfactory degree of discrimination power. The dis- features had a minor influence on landscape preferences. Although
criminant analysis could not correctly classify more cases as could be differences between the two levels are low, most people preferred
expected with random assignment, indicating that the variables used pictures without urban settlements and road infrastructures, which
208
U. Schirpke et al. Ecological Indicators 96 (2019) 202–212
NP PR
high high
medium medium
low low
-10 -5 0 5 10 -10 -5 0 5 10
SHAPE_AM SHAPE_MD
high high
medium medium
low low
-10 -5 0 5 10 -10 -5 0 5 10
SHAPE_CV TA
high high
medium medium
low low
-10 -5 0 5 10 -10 -5 0 5 10
TA_N Urban
high
present
medium
absent
low
-10 -5 0 5 10 -10 -5 0 5 10
Road Forest
present present
absent absent
-10 -5 0 5 10 -10 -5 0 5 10
Water
present
absent
-10 -5 0 5 10
confirms previous findings (Grêt-Regamey et al., 2007; Jessel, 2006; a general preference for medium complexity of patch shape compared
Ode et al., 2009; Schirpke et al., 2013a). Hence, unrestrained con- to simple or very complex shapes. Similar results were reported by Ode
struction of settlements and roads may have rather negative impacts on et al. (2009) who found higher preference scores for landscapes with a
landscape preferences (Grêt-Regamey et al., 2007). low to medium level of shape complexity than for those showing a high
Landscape pattern had a medium influence on landscape pre- level of shape complexity. The degree of landscape diversity was di-
ferences, and the importance scores of the various indicators did not rectly linked to higher preferences, which is consistent with other stu-
differ significantly from each other in most cases. Our analysis revealed dies (Dramstad et al., 2006; Fry et al., 2009; Palmer, 2004; Sang et al.,
209
U. Schirpke et al. Ecological Indicators 96 (2019) 202–212
2008), although a limited number of land cover types is preferred (de la analyses, could not reliably reproduce group differences based on the
Fuente de Val et al., 2006). Landscape diversity is also influenced by variables used. This result does not mean that the variables are not
coherence, i.e., the arrangement of landscape elements, and inter- significant, but it indicates that important characteristics influencing
mediate levels of coherence are liked best (Kuper, 2017; Van der Jagt landscape preferences of individuals were not included. In addition to
et al., 2014). Finally, long vistas are considered to increase the aesthetic the socio-demographic variables that were available from the used
landscape appreciation (Germino et al., 2001; Schirpke et al., 2013a). questionnaire and which were also included in several studies (e.g.
Accordingly, we found higher preferences for a high level of visible area Bauer et al., 2009; Hunziker et al., 2008; Schirpke et al., 2013a; Tveit,
than for the intermediate level. A narrower viewing field, but still a 2009), further variables may influence individual landscape preferences
great view depth, may explain the high scores for the low level. such as the educational level (Dramstad et al., 2006), the environmental
Moreover, pictures with a large viewshed but less liked features in the value orientation (Howley, 2011; Kaltenborn and Bjerke, 2002) as well
foreground, which has the greatest influence on landscape preferences as the type of interaction or connection with the landscape (e.g. farmer,
(Schirpke et al., 2013b), may have decreased the part-worth utilities for tourist) (Sayadi et al., 2009; Van den Berg et al., 1998). This implies
the medium and high levels. In summary, our results indicate that the that future research on the variables impacting the individual weighting
method proposed is reliable with respect to quantitative measures function of landscape indicators is needed and that these variables have
(RLH) and stylized facts compared to other studies. to be included in future assessments to evaluate group differences. In
general, the cluster analysis revealed differences in landscape pre-
4.2. Relationships between of landscape indicators and landscape ferences, but the information from the survey did not allow us to suf-
preferences ficiently explain these differences and their causes.
210
U. Schirpke et al. Ecological Indicators 96 (2019) 202–212
may explain the unexpected low importance of water in our study Environment’ at the University of Innsbruck.
compared to the literature. Many studies indicated that water is one of
the most preferred landscape features (Jessel, 2006; Kaplan and Kaplan, Appendix A. Supplementary data
1989; Pastorella et al., 2017). Indeed, the highest rated picture of the
used questionnaire depicts a lake as a central feature. This methodo- Supplementary data associated with this article can be found, in the
logical weakness should be overcome in future studies by visually online version, at https://doi.org/10.1016/j.ecolind.2018.09.001.
analyzing the pictures or improving the GIS-based model. Secondly, we
measured only presence/absence of the landscape features. However, References
for example the percentage of forest has considerable effects on human
preferences (Hunziker and Kienast, 1999; Tveit et al., 2006), and it is Alriksson, S., Öberg, T., 2008. Conjoint analysis for environmental evaluation. Environ.
important to consider whether features are located in proximity to the Sci. Pollut. Res. 15, 244–257.
Álvarez-Farizo, B., Hanley, N., 2002. Using conjoint analysis to quantify public pre-
observer or in the background (Schirpke et al., 2013a). Therefore, ferences over the environmental impacts of wind farms. An example from Spain.
further studies should account for the spatial distribution of landscape Energy Policy 30, 107–116.
features in quantitative terms and possibly assign weights for different Andrews, R.L., Ansari, A., Currim, I.S., 2002. Hierarchical Bayes versus finite mixture
conjoint analysis models: a comparison of fit, prediction, and partworth recovery. J.
distance zones. These limitations, however, do not reduce the principal Market. Res. 39, 87–98.
suitability of this approach for analyzing landscape preferences. Arriaza, M., Cañas-Ortega, J.F., Cañas-Madueño, J.A., Ruiz-Aviles, P., 2004. Assessing the
visual quality of rural landscapes. Landscape Urban Plann. 69, 115–125. https://doi.
org/10.1016/j.landurbplan.2003.10.029.
5. Conclusions
Bauer, N., Wallner, A., Hunziker, M., 2009. The change of European landscapes: human-
nature relationships, public attitudes towards rewilding, and the implications for
From our results, we can draw the following main conclusions: landscape management in Switzerland. J. Environ. Manage. 90, 2910–2920. https://
doi.org/10.1016/j.jenvman.2008.01.021.
• This study found that 11 indicators are sufficient to predict the in-
Beza, B.B., 2010. The aesthetic value of a mountain landscape: A study of the Mt. Everest
Trek. Landscape Urban Plann. 97, 306–317. https://doi.org/10.1016/j.landurbplan.
dividual choice between two landscapes for ∼90% of the re- 2010.07.003.
spondents, but the study provides no evidence to determine if this is Bonzanigo, L., Giupponi, C., Balbi, S., 2016. Sustainable tourism planning and climate
change adaptation in the Alps: a case study of winter tourism in mountain commu-
true also for other types of landscapes (e.g., flat or collinear) and if nities in the Dolomites. J. Sustainable Tourism 24, 637–652.
so, whether the landscape indicators are the same or whether they Cañas, I., Ayuga, E., Ayuga, F., 2009. A contribution to the assessment of scenic quality of
depend on the type of landscape. To investigate these questions, landscapes based on preferences expressed by the public. Land Use Policy 26,
1173–1181. https://doi.org/10.1016/j.landusepol.2009.02.007.
further research is needed. Chhetri, P., Arrowsmith, C., Jackson, M., 2004. Determining hiking experiences in nature-
• The results obtained indicate that the relationships between the based tourist destinations. Tourism Manage. 25, 31–43.
Daniel, T.C., 2001. Whither scenic beauty? Visual landscape quality assessment in the
levels of certain landscape indicators and the landscape preferences
21st century. Landscape Urban Plann. 54, 267–281. https://doi.org/10.1016/S0169-
are not always linear. We limited our study to a maximum of three 2046(01)00141-4.
levels to achieve maximum robustness in the estimates and obtain a de la Fuente de Val, Gonzalo, Atauri, J.A., de Lucio, J.V., 2006. Relationship between
clear result. With more detailed scales for the indicators used, in- landscape visual attributes and spatial pattern indices: a test study in Mediterranean-
climate landscapes. Landscape Urban Plann. 77, 393–407. https://doi.org/10.1016/j.
teresting results regarding the complex relationships between in- landurbplan.2005.05.003.
dicators and preferences including nonlinearities in aesthetic per- Dramstad, W.E., Tveit, M.S., Fjellstad, W.J., Fry, G.L.A., 2006. Relationships between
ception may be obtainable. visual landscape preferences and map-based indicators of landscape structure.
• Considerable heterogeneity occurred for the vectors of part-worth Landscape Urban Plann. 78, 465–474. https://doi.org/10.1016/j.landurbplan.2005.
12.006.
utilities. Even though we were not able to explain the causes and Egarter Vigl, L., Schirpke, U., Tasser, E., Tappeiner, U., 2016. Linking long-term landscape
nature of these differences, these results suggest that using ag- dynamics to the multiple interactions among ecosystem services in the European
gregated means to estimate the aesthetic landscape preferences Alps. Landscape Ecol. 31 (9), 1903–1918. https://doi.org/10.1007/s10980-016-
0389-3.
implies some methodical problems when applying aggregated linear Fliri, F., 1984. Synoptische Klimatographie der Alpen zwischen Mont Blanc und Hohen
prediction models. Tauern (Synoptic climatology of the Alps between Mont Blanc and Hohen Tauern).
Wagner, Innsbruck.
Fry, G., Tveit, M.S., Ode, Å., Velarde, M.D., 2009. The ecology of visual landscapes: ex-
In summary, conjoint analysis applied a posteriori can be used for ploring the conceptual common ground of visual and ecological landscape indicators.
obtaining additional information about landscape preferences from Ecol. Ind. 9, 933–947. https://doi.org/10.1016/j.ecolind.2008.11.008.
perception surveys. By evaluating picture ratings on the individual Gao, J., Barbieri, C., Valdivia, C., 2014. Agricultural landscape preferences: implications
for agritourism development. J. Travel Res. 53, 366–379.
level, the usefulness of a great number of landscape indicators can be Garbarino, M., Sibona, E., Lingua, E., Motta, R., 2014. Decline of traditional landscape in
identified and the links between preferences and landscape pattern or a protected area of the southwestern Alps: the fate of enclosed pasture patches in the
features can be analyzed in detail. Our findings can therefore enhance land mosaic shift. J. Mountain Sci. 11, 544.
Germino, M.J., Reiners, W.A., Blasko, B.J., McLeod, D., Bastian, C.T., 2001. Estimating
spatial modelling approaches predicting aesthetic landscape values and visual properties of Rocky Mountain landscapes using GIS. Landscape Urban Plann.
provide valuable insights into landscape preferences that can support 53, 71–83. https://doi.org/10.1016/S0169-2046(00)00141-9.
landscape planning and management. Further research, however, is Goossen, M., Langers, F., 2000. Assessing quality of rural areas in the Netherlands: finding
the most important indicators for recreation. Landscape Urban Plann. 46, 241–251.
necessary to explain the differences between social groups and to
Grêt-Regamey, A., Bebi, P., Bishop, I.D., Schmid, W.A., 2008. Linking GIS-based models to
combine theoretical concepts with information from spatial and sta- value ecosystem services in an Alpine region. J. Environ. Manage. 89, 197–208.
tistical analyses. https://doi.org/10.1016/j.jenvman.2007.05.019.
Grêt-Regamey, A., Bishop, I.D., Bebi, P., 2007. Predicting the scenic beauty value of
mapped landscape changes in a mountainous region through the use of GIS. Environ.
Acknowledgements Plann. B 34, 50–67.
Hair, J., Black, W., Babin, B., Anderson, R., 2010. In: Multivariate Data Analysis: A Global
This study was co-financed by the Austrian Federal Ministry of Perspective. Pearson Education, pp. 800.
Horvath, H., 1995. Estimation of the average visibility in central Europe. Atmos. Environ.
Science, Research and Economy with the HRSM– cooperation project 29 (2), 241–246.
KLIMAGRO, the European Regional Development Fund through the Howley, P., 2011. Landscape aesthetics: assessing the general publics' preferences to-
Interreg Alpine Space programme (“AlpES” project, CUP: wards rural landscapes. Ecol. Econ. 72, 161–169.
Hunziker, M., Felber, P., Gehring, K., Buchecker, M., Bauer, N., Kienast, F., 2008.
D52I16000220007), and supported by the KuLaWi project (INTERREG Evaluation of landscape change by different social groups. Mt. Res. Dev. 28, 140–147.
IV–EU project (Agri)cultural landscape – Strategies for the cultural https://doi.org/10.1659/mrd.0952.
landscape of the future, project no. 4684, CUP: B26D09000010007). UT Hunziker, M., Kienast, F., 1999. Potential impacts of changing agricultural activities on
scenic beauty - a prototypical technique for automated rapid assessment. Landscape
and GT are members of the research focus ‘Alpine Space—Man and
211
U. Schirpke et al. Ecological Indicators 96 (2019) 202–212
Ecol. 14, 161–176. temporal patterns of outdoor recreation in the European Alps and their surroundings.
Jessel, B., 2006. Elements, characteristics and character–Information functions of land- Ecosyst. Serv. 31C, 336–350. https://doi.org/10.1016/j.ecoser.2017.11.017.
scapes in terms of indicators. Ecol. Ind. 6, 153–167. Schirpke, U., Tasser, E., Tappeiner, U., 2013b. Predicting scenic beauty of mountain re-
Junge, X., Schüpbach, B., Walter, T., Schmid, B., Lindemann-Matthies, P., 2015. Aesthetic gions. Landscape Urban Plann. 111, 1–12.
quality of agricultural landscape elements in different seasonal stages in Switzerland. Schirpke, U., Timmermann, F., Tappeiner, U., Tasser, E., 2016. Cultural ecosystem ser-
Landscape Urban Plann. 133, 67–77. vices of mountain regions: modelling the aesthetic value. Ecol. Ind. 69, 78–90.
Kaltenborn, B.P., Bjerke, T., 2002. Associations between environmental value orientations https://doi.org/10.1016/j.ecolind.2016.04.001.
and landscape preferences. Landscape Urban Plann. 59 (1), 1–11. Schneeberger, N., Bürgi, M., Kienast, F., 2007. Rates of landscape change at the northern
Kaplan, R., Kaplan, S., 1989. The Experience of Nature: A Psychological Perspective. CUP fringe of the Swiss Alps: Historical and recent tendencies. Landscape Urban Plann. 80,
Archive. 127–136. https://doi.org/10.1016/j.landurbplan.2006.06.006.
Kuper, R., 2017. Evaluations of landscape preference, complexity, and coherence for Soliva, R., Bolliger, J., Hunziker, M., 2010. Differences in preferences towards potential
designed digital landscape models. Landscape Urban Plann. 157, 407–421. future landscapes in the Swiss Alps. Landscape Res. 35, 671–696.
Lenk, P.J., DeSarbo, W.S., Green, P.E., Young, M.R., 1996. Hierarchical Bayes conjoint Soliva, R., Hunziker, M., 2009. Beyond the visual dimension: Using ideal type narratives
analysis: recovery of partworth heterogeneity from reduced experimental designs. to analyse people's assessments of landscape scenarios. Land Use Policy 26, 284–294.
Market. Sci. 15, 173–191. https://doi.org/10.1016/j.landusepol.2008.03.007.
Lindemann-Matthies, P., Briegel, R., Schüpbach, B., Junge, X., 2010. Aesthetic preference Tasser, E., Leitinger, G., Tappeiner, U., 2017. Climate change versus land-use chan-
for a Swiss alpine landscape: the impact of different agricultural land-use with dif- ge—what affects the mountain landscapes more? Land Use Policy 60, 60–72.
ferent biodiversity. Landscape Urban Plann. 98, 99–109. https://doi.org/10.1016/j. Tasser, E., Walde, J., Tappeiner, U., Teutsch, A., Noggler, W., 2007. Land-use changes and
landurbplan.2010.07.015. natural reforestation in the Eastern Central Alps. Agric. Ecosyst. Environ. 118,
Lothian, A., 1999. Landscape and the philosophy of aesthetics: is landscape quality in- 115–129. https://doi.org/10.1016/j.agee.2006.05.004.
herent in the landscape or in the eye of the beholder? Landscape Urban Plann. 44, Tempesta, T., Giancristofaro, R.A., Corain, L., Salmaso, L., Tomasi, D., Boatto, V., 2010.
177–198. https://doi.org/10.1016/S0169-2046(99)00019-5. The importance of landscape in wine quality perception: an integrated approach
McGarigal, K., Cushman, S.A., Neel, M.C., Ene, E., 2002. FRAGSTATS: Spatial Pattern using choice-based conjoint analysis and combination-based permutation tests. Food
Analysis Program for Categorical Maps. Computer software program produced by the Qual. Prefer. 21, 827–836.
authors at the University of Massachusetts, Amherst. http://www.umass.edu/ Triguero-Mas, M., Dadvand, P., Cirach, M., Martínez, D., Medina, A., Mompart, A.,
landeco/research/fragstats/fragstats.html (accessed: 01/08/2017). Basagaña, X., Gražulevičienė, R., Nieuwenhuijsen, M.J., 2015. Natural outdoor en-
Netzer, O., Srinivasan, V., 2011. Adaptive self-explication of multiattribute preferences. J. vironments and mental and physical health: relationships and mechanisms. Environ.
Mark. Res. 48 (1), 140–156. Int. 77, 35–41.
Ode, Å., Tveit, M.S., Fry, G., 2008. Capturing landscape visual character using indicators: Tveit, M., Ode, Å., Fry, G., 2006. Key concepts in a framework for analysing visual
touching base with landscape aesthetic theory. Landscape Res. 33 (1), 89–117. landscape character. Landscape Res. 31, 229–255.
Ode, Å., Fry, G., Tveit, M.S., Messager, P., Miller, D., 2009. Indicators of perceived nat- Tveit, M.S., 2009. Indicators of visual scale as predictors of landscape preference; a
uralness as drivers of landscape preference. J. Environ. Manage. 90, 375–383. comparison between groups. J. Environ. Manage. 90, 2882–2888. https://doi.org/10.
https://doi.org/10.1016/j.jenvman.2007.10.013. 1016/j.jenvman.2007.12.021.
Palmer, J.F., 2004. Using spatial metrics to predict scenic perception in a changing Yitzhaki, S., Schechtman, E., 2012. Identifying monotonic and non-monotonic relation-
landscape: Dennis, Massachusetts. Landscape Urban Plann. 69, 201–218. https://doi. ships. Econ. Lett. 116 (1), 23–25.
org/10.1016/j.landurbplan.2003.08.010. van Berkel, D.B., Verburg, P.H., 2014. Spatial quantification and valuation of cultural
Pastorella, F., Giacovelli, G., De Meo, I., Paletto, A., 2017. People’s preferences for Alpine ecosystem services in an agricultural landscape. Ecol. Indic. 37 (Part A), 163–174.
forest landscapes: results of an internet-based survey. J. For. Res. 22, 36–43. https://doi.org/10.1016/j.ecolind.2012.06.025.
Price, B., Kienast, F., Seidl, I., Ginzler, C., Verburg, P.H., Bolliger, J., 2015. Future J. Van Cauwenberg I. De Bourdeaudhuij P. Clarys J. Nasar J. Salmon L. Goubert B.
landscapes of Switzerland: Risk areas for urbanisation and land abandonment. Appl. Deforche 2016. Street characteristics preferred for transportation walking among
Geogr. 57, 32–41. older adults: a choice-based conjoint analysis with manipulated photographs, Int. J.
Rao, V.R., 2014. Applied Conjoint Analysis. Springer Science & Business Media. Behav. Nutr. Phys. Act. 13, 6.
Romagosa, F., Eagles, P.F., Lemieux, C.J., 2015. From the inside out to the outside in: Van den Berg, A.E., Vlek, C.A., Coeterier, J.F., 1998. Group differences in the aesthetic
Exploring the role of parks and protected areas as providers of human health and evaluation of nature development plans: a multilevel approach. J. Environ. Psychol.
well-being. J. Outdoor Recreation Tourism 10, 70–77. 18 (2), 141–157.
Sang, N., Ode, Å., Miller, D., 2008. Landscape metrics and visual topology in the analysis der Jagt, Van, Alexander, P.N., Craig, T., Anable, J., Brewer, M.J., Pearson, D.G., 2014.
of landscape preference. Environ. Plann. B 35, 504–520. Unearthing the picturesque: the validity of the preference matrix as a measure of
Sawtooth Software, 2009. The CBC/HB Module for Hierarchical Bayes Estimation. landscape aesthetics. Landscape Urban Plann. 124, 1–13.
Technical Paper. https://www.sawtoothsoftware.com/support/technical-papers/ Veitch, J., Salmon, J., Deforche, B., Ghekiere, A., Van Cauwenberg, J., Bangay, S.,
hierarchical-bayes-estimation/cbc-hb-technical-paper-2009 (last accessed 06 May Timperio, A., 2017. Park attributes that encourage park visitation among adolescents:
2018). a conjoint analysis. Landscape Urban Plann. 161, 52–58.
Sayadi, S., González-Roa, M.C., Calatrava-Requena, J., 2009. Public preferences for Virgili, A., Racine, M., Authier, M., Monestiez, P., Ridoux, V., 2017. Comparison of ha-
landscape features: the case of agricultural landscape in mountainous Mediterranean bitat models for scarcely detected species. Ecol. Model. 346, 88–98. https://doi.org/
areas. Land Use Policy 26, 334–344. https://doi.org/10.1016/j.landusepol.2008.04. 10.1016/j.ecolmodel.2016.12.013.
003. Weinstoerffer, J., Girardin, P., 2000. Assessment of the contribution of land use pattern
Sayadi, S., Roa, M.C.G., Requena, J.C., 2005. Ranking versus scale rating in conjoint and intensity to landscape quality: use of a landscape indicator. Ecol. Model. 130,
analysis: evaluating landscapes in mountainous regions in southeastern Spain. Ecol. 95–109. https://doi.org/10.1016/S0304-3800(00)00209-X.
Econ. 55, 539–550. Zaunbrecher, B.S., Linzenich, A., Ziefle, M., 2017. A mast is a mast is a mast...?
Schirpke, U., Hölzler, S., Leitinger, G., Bacher, M., Tappeiner, U., Tasser, E., 2013a. Can Comparison of preferences for location-scenarios of electricity pylons and wind
we model the scenic beauty of an alpine landscape? Sustainability 5, 1080–1094. power plants using conjoint analysis. Energy Policy 105, 429–439.
Schirpke, U., Meisch, C., Marsoner, T., Tappeiner, U., 2018. Revealing spatial and
212