Predicting scenic beauty of mountain regions R P U a b c h � � � � a A R R 2 A K G L L P 1 n 2 c a ( 0 h Landscape and Urban Planning 111 (2013) 1– 12 Contents lists available at SciVerse ScienceDirect Landscape and Urban Planning j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / l a n d u r b p l a n esearch paper redicting scenic beauty of mountain regions ta Schirpke a,∗, Erich Tasser b, Ulrike Tappeiner c Institute for Alpine Environment, European Academy Bolzano/Bozen, Viale Druso 1, I-39100 Bolzano, Italy Institute for Alpine Environment, European Academy Bolzano/Bozen, Italy Institute of Ecology, University of Innsbruck, Institute for Alpine Environment, European Academy Bolzano/Bozen, Italy i g h l i g h t s Developed method allows predicting scenic beauty of mountain regions. Good prediction of scenic beauty (R2 = 0.72). Near zone contributes to scenic beauty by 48%. Method can be used for decision making and landscape planning. g r a p h i c a l a b s t r a c t r t i c l e i n f o rticle history: eceived 24 November 2011 eceived in revised form 1 November 2012 ccepted 26 November 2012 eywords: IS andscape metrics a b s t r a c t Scenic beauty of mountain landscapes contributes to human well-being. Valuation of natural scenery and specific landscape properties by perception studies is complex and time-consuming. Sophisticated spatial analysis tools can support the assessment of scenic beauty by quantitative methods. We implemented an innovative GIS-based modeling approach for mountain regions which combines objective methods with perception-based methods. Based on viewpoints, spatial patterns of visible landscape were analyzed by means of landscape metrics. A set of 60 landscape metrics were reduced by principal component analysis (PCA) to 11 components explaining 93% of the variance. The components were related to perceived scenic beauty values found through a perception study via stepwise regression analysis. We found that two com- 2 and use erception study ponents, shape complexity and landscape diversity, are positively related to visual quality (R = 0.72). In the Central Alps, especially areas above the tree line are characterized by high scenic beauty. Abandon- ment of agriculturally used areas implies a loss of scenic beauty, mainly in the valley bottom and in the subalpine forest belt, as a result of urban sprawl and natural reforestation. The GIS-based model offers a valid instrument for scenic beauty assessments of mountain regions as a basis for policy making and landscape planning. . Introduction Humans find great opportunities for recreation and leisure in atural ecosystems (de Groot, Alkemade, Braat, Hein, & Willemen, 010). The demand for outdoor recreation has been growing ontinuously, and especially mountain environments are highly ppreciated by tourists (Raitz & Dakhil, 1988) because of scenic ∗ Corresponding author. Tel.: +39 0471 055 333; fax: +39 0471 055 399. E-mail addresses: uta.schirpke@eurac.edu (U. Schirpke), erich.tasser@eurac.edu E. Tasser), Ulrike.Tappeiner@uibk.ac.at (U. Tappeiner). 169-2046/$ – see front matter © 2012 Elsevier B.V. All rights reserved. ttp://dx.doi.org/10.1016/j.landurbplan.2012.11.010 © 2012 Elsevier B.V. All rights reserved. beauty, fresh air, varied topography, and forests (Beza, 2010; Scarpa, Chilton, Hutchinson, & Buongiorno, 2000). The cultural landscape of mountain regions has been shaped by hundreds of years of agricultural activities (Fischer, Rudmann-Maurer, Weyand, & Stöcklin, 2008) leading to a mosaic of agricultural land, natu- ral grassland and forests. During the last decade, many European mountain regions have become affected by land abandonment (Rutherford, Bebi, Edwards, & Zimmermann, 2008; Schneeberger, Bürgi, & Kienast, 2007), and non-agricultural sources of income, in particular tourism, have become more important for the local population. Particularly the abandonment of alpine pastures and meadows results in natural forest re-growth (Sitzia, Semenzato, dx.doi.org/10.1016/j.landurbplan.2012.11.010 http://www.sciencedirect.com/science/journal/01692046 http://www.elsevier.com/locate/landurbplan mailto:uta.schirpke@eurac.edu mailto:erich.tasser@eurac.edu mailto:Ulrike.Tappeiner@uibk.ac.at dx.doi.org/10.1016/j.landurbplan.2012.11.010 2 and U & w & e i d t 2 b h b a o C B b 2 s C M h s 1 t ( e t R e v t a b s a c e p o e a f t s F P p e b H o d t l 2 a p m p t v r l U. Schirpke et al. / Landscape Trentanovi, 2010; Tasser, Schermer, Siegl, & Tappeiner, 2012) hich, however, humans perceive a loss in scenic beauty (Hunziker Kienast, 1999). As the recreational quality of a region is to a great xtent linked to its scenic beauty (Chhetri & Arrowsmith, 2008), t constitutes a competitive advantage in respect to other tourist estinations. Therefore, scenic beauty assessments are an impor- ant aid for planners and stakeholders (Ribe, 2009; Tasser et al., 012). The scenic beauty of a landscape comes from the interaction etween its biophysical features and the human observer which as led to perception-based and expert-based methods for scenic eauty assessments (Daniel, 2001). Perception-based methods ssess community perceptions and analyze perceived scenic beauty n-site or by presenting photographs (Arriaza, Cañas-Ortega, añas-Madueño, & Ruiz-Aviles, 2004; Grêt-Regamey, Bishop, & ebi, 2007). Although some perception studies found differences etween groups by age, gender, social stratum (Hunziker et al., 008; Tveit, 2009) or cultural background (Zube & Pitt, 1981), many tudies suggest substantial agreement across different groups (e.g. añas, Ayuga, & Ayuga, 2009; Kearney et al., 2008; Ode, Fry, Tveit, essager, & Miller, 2009). Perception-based assessments have a igh level of reliability (Daniel, 2001), but they are relatively expen- ive, time-consuming and difficult to organize on site (Lothian, 999). Assessments of large complex landscapes are often limited o locations along linear features such as roads, trails, and rivers Meitner, 2004; Beza, 2010). In contrast, expert-based approaches xamine defined visual properties and biophysical features of he landscape by quantitative methods (Daniel, 2001). Germino, einers, Blasko, McLeod, and Bastian (2001) estimated visual prop- rties of Rocky Mountains landscapes quantifying dimensions of iews, e.g. areal extent, depth, relief, and composition of views in erms of diversity and edge of land cover. de la Fuente de Val, Atauri, nd de Lucio (2006) correlated different variables describing scenic eauty, e.g. coherence, legibility, complexity, mystery and diver- ity, to landscape metrics. The major advantage of an expert-based ssessment is its efficiency (Lothian, 1999), which allows the appli- ation for whole regions by using automated procedures. However, xpert-based assessments have not reached the high reliability of erception-based methods because they are extremely dependent n the professional knowledge of the assessor (Daniel, 2001). To benefit from the advantages of each assessment method, sev- ral authors linked perception-based approaches to expert-based pproaches by examining the relationship between landscape pre- erences and landscape patterns by landscape metrics. They found hat diversity indices in particular are positively correlated to land- cape preferences (Dramstad, Tveit, Fjellstad, & Fry, 2006; Franco, ranco, Mannino, & Zanetto, 2003; Hunziker & Kienast, 1999). almer (2004) identified a better relation of scenic beauty to com- osition metrics than to configuration metrics. Only few studies xamined the scenic beauty of mountain landscapes using a com- ination of perception-based and expert-based methods. Whereas unziker and Kienast (1999) examined landscape metrics based n photographs, Grêt-Regamey et al. (2007) included a three- imensional view analysis but concentrated on only three land-use ypes. In contrast to flat landscapes, where only artificial features ike wind turbines are visible at greater distances (Shang & Bishop, 000), in mountain areas topographic characteristics like slope and spect have to be considered in addition to distance. Especially laces of higher elevation than their surrounding area, such as ountain peaks, have long vistas (Germino et al., 2001) and visual roperties as size and perceived landscape color change with dis- ance (Bishop, 2003). While in flat landscapes artificial elements or egetation can block the view, in mountain regions vistas in lower egions such as valley bottoms can be limited by mountains. To estimate scenic beauty for any viewpoint within mountain andscapes, an efficient spatially explicit assessment method that rban Planning 111 (2013) 1– 12 accounts for the implications of topography on view properties, and, at the same time, ensures the high reliability of perception- based assessments is still lacking. To fill this gap, we aimed at developing a modeling approach to predict scenic beauty of moun- tain regions and divided our research into the following steps: (1) explore relief-dependent visual properties by using a geographical information system (GIS), (2) examine composition and configura- tion of landscape by landscape metrics, (3) test if perceived scenic beauty can be related to landscape pattern, and (4) estimate scenic beauty for the Central Alps, aiming to investigate relationships of land use and scenic beauty. 2. Methods The methodology followed in this paper can be divided into six distinct parts. First, we selected representative study sites for the Central Alps. Second, we introduced distance zones to explore visual properties of mountain regions. Third, we determined the necessary input data for the different distance zones. Fourth, we conducted a visibility analysis to determine the visible area seen from an observer point by using GIS. The visible area was then intersected with the land-cover maps of each distance zone and land cover mosaics were created. Fifth, we calculated landscape metrics based on the land cover mosaics. Finally, to assess human perceptions, we carried out a perception survey to obtain perceived scenic beauty values. We related the perceived scenic beauty val- ues to landscape metrics by a regression analysis to predict scenic beauty for any viewpoint. 2.1. Study sites We developed our model for the greater region of the Central Alps. To cover geographical variations of relief and land cover, we selected four minor study sites (Fig. 1): (1) Lech Valley, Austria (municipalities of Gramais, Hinterhornbach, Pfafflar and Stan- zach); (2) Stubai Valley, Austria (municipalities of Neustift im Stubai and Fulpmes); (3) Pustertal, Italy (municipalities of Gsies, Rasen-Antholz, Sand in Taufers, Prettau) and (4) Vinschgau, Italy (municipalities of Glurns, Graun im Vinschgau, Mals, Schluderns). Their landscapes are mainly composed of forest and grassland with different management intensities, from intensively used grassland in lower regions to alpine pastures and abandoned land, mostly in regions above the tree line, with higher areas covered by rocks and glaciers. The study sites belong to the Northern Central European climate zone except for the Vinschgau, which is part of the Central Alpine arid climate zone (Fliri, 1984). They are characterized by dif- ferent relief properties and have diverse land cover distributions (Table 1). 2.2. Distance zones In mountain landscapes with long vistas, object appearance, color difference, and lightness contrast of an object and its surroundings decrease with increasing distance, leading to less dis- cernible detail (Bishop, 2003). To account for the effect of distance on the perception of size and color, several authors introduced distance zones and divided the landscape into foreground, mid- dle ground and background (Bishop & Hulse, 1994; de la Fuente de Val et al., 2006; Germino et al., 2001). While Bishop and Hulse (1994) limited the viewshed analysis to just 2 km from the obser- vation point, Germino et al. (2001) defined background as up to 150 km. In contrast to flat landscapes, where only elements rising from the landscape are visible, the landscape in mountain regions can be seen in top view from viewpoints at higher positions than the surrounding area. We adapted the distance zones to the high U. Schirpke et al. / Landscape and Urban Planning 111 (2013) 1– 12 3 Fig. 1. Location of study sites. Table 1 Areal extent, relief-dependent properties, and land cover distribution of the four study sites. Study site Area [km2 ] Elevation [m a.s.l.] Slope [◦ ] Land cover distribution [%] Min Max Mean Min Max Mean Agricultural area Forest Settlement 1 Lech Valley 150 905 2727 1724 0 79 31 5 51 <1 v t d a • F v ( 2 Stubai Valley 265 890 3488 2167 3 Pustertal 482 833 3456 2026 4 Vinschgau 491 882 3723 2196 ariability of landscape pattern and relief properties of the Cen- ral Alps. Based on the distinguishability of landscape elements, we efined three distance zones from a viewpoint within the study rea (Fig. 2): near zone, up to 1.5 km. Details of single features such as trees or buildings are clearly identifiable. ig. 2. Visible area intersected with land use of the three distance zones from a iewpoint (black dot). (A) Near zone (0–1.5 km), (B) middle zone (1.5–10 km) and C) far zone (10–50 km). 0 86 31 10 34 <1 0 75 28 14 41 <1 0 69 26 34 29 <1 • middle zone, from 1.5 to 10 km. Single elements merge, e.g. single trees form a forest or buildings make up a village. • far zone, up to 50 km. Although views of up to 150 km are pos- sible from mountain peaks (Germino et al., 2001), good visibility outside population centers in Europe is considered as 40–50 km, and longer vistas occur only under rare occasions (Horvath, 1995). The number of perceivable land-cover classes decreases, whereas edge and outline of the landform still play a major role for the perception of space. The different distance zones are used to select input data with diverse spatial and thematic resolution for the GIS-based model. The visibility analysis within the model was performed taking into account scale and perceived color dependencies from a distance. The distance zones were also applied for attributing weights to the pictures of the perception survey. 2.3. Data collection For each distance zone, spatial information was selected and/or aggregated with regard to content and spatial resolution. Digital elevation models (DEM) were applied to determine visible area and to derive relief-dependent variables. For the near zone, we used DEM with a resolution of 20 m × 20 m, provided by the Tyrolean Information System (tiris, ©Land Tirol) of the Province of Tyrol and the Autonomous Province of Bolzano-South Tyrol. For the mid- dle and far zone, elevation was obtained from a DEM consisting of processed data from the Shuttle Radar Topography Mission (SRTM) with a resolution of 3 arc-seconds (Jarvis et al., 2008). The resolution was adapted to 100 m × 100 m for the middle zone and resampled to 1 km × 1 km for the far zone. 4 and U m t d a n i m f s g t l s g b p t m t s d o 2 l i z f g i g s P & t K o a n b t w l H C s t 2 t c o g v s v e & e v t o b U. Schirpke et al. / Landscape Habitat and land cover maps were used to calculate landscape etrics. For the near zone, the analysis was performed using habi- ats which are essential in the analysis of species and landscape iversity (Dudley, Baldock, Nasi, & Stolton, 2005). The habitat map s applied by Tasser, Ruffini, and Tappeiner (2009) is a register of atural, near-natural, and artificial habitats, e.g. grassland habitats n valleys are distinguished from those on the subalpine belt, or anaged coniferous forests are different from mixed or deciduous orests. The variety of habitats helps to capture landscape diver- ity. Structural elements like point or linear landscape features (e.g. roves, hedges, single trees, banks, debris areas) explain landscape exture (Michel, Burel, Legendre, & Butet, 2007) and help to express andscape quality better (Weinstoerffer & Girardin, 2000). A land- cape structure map was intersected with the habitat map, both enerated for the study areas from orthophotos (scale 1:10,000) y on-screen digitizing in a GIS. Additionally, land cover was sup- lemented by three spatial datasets: major streams selected from he river network; plus roads, both provided by the Tyrolean Infor- ation System (tiris, ©Land Tirol) of the Province of Tyrol and he Autonomous Province of Bolzano-South Tyrol, and mapped ingle settlement points. All datasets were converted to raster atasets with a spatial resolution of 20 m × 20 m and merged into ne dataset. For the middle and the far zone, we used CORINE land cover 000 (CLC2000) seamless vector database (EEA, 2009). Based on the ow number of land cover classes in the Central Alps and accord- ng to the distinguishability of elements as defined for the distance ones, the 44 CLC-level-3 classes were aggregated into six classes or the middle zone: forest, grassland, settlement, rock, water and lacier. For the far zone, rock was included in the grassland because t is often covered by sparse vegetation and therefore less distin- uishable from alpine grassland with increasing distance. Water, ettlements and glaciers constitute important landscape elements. resence of water has a positive influence on scenic beauty (Bishop Hulse, 1994) and offers a wide range of recreational activi- ies. Glaciers are also important tourist attractions (Scott, Jones, & onopek, 2007). In contrast, large settlements have negative effects n scenic beauty in mountain regions (Grêt-Regamey et al., 2007) nd perceived scenic beauty is strongly correlated with natural- ess (Lamb & Purcell, 1990). Color differences, which are greater etween the bright color of settlements and glaciers with vegeta- ion than between different vegetation types, and the reflection of ater surfaces also support visibility and distinguishability of these andscape elements from greater distances (Litton, 1977; García, ernández, & Ayuga, 2003). To account for the large scale of the ORINE land cover map and to include all areas of the classes water, ettlements and glaciers, these classes were treated as a priority in he conversion from polygon to raster datasets. .4. Visibility analysis Visual properties of the landscape are determined by the loca- ion of a viewpoint. Rather than any specific restricted view as aptured by photographs, the surroundings affect the perception f the visual environment in their entirety (Meitner, 2004). A geo- raphic information system is a suitable tool for analyzing 360◦ iews from a viewpoint. Due to the topography of mountain land- capes, some areas of the landscape may not be visible from the iewpoint. By using an algorithm for estimating whether or not ach target cell is within the observer’s line-of-sight (Kim, Rana, Wise, 2004), viewsheds can be calculated and non-visible areas xcluded. A DEM does not take into account feature height from egetation or buildings which can narrow or completely block he view. Heights of mapped surface features were superimposed nto the DEM and a digital surface model (DSM) was generated y adding the feature heights to the ground elevation (DEM). An rban Planning 111 (2013) 1– 12 average height of 20 m was assigned to forest (Wallentin, Tappeiner, Strobl, & Tasser, 2008) and 2 m to shrubs (Dullinger, Dirnböck, & Grabherr, 2003), while the average height of buildings was estimated as 10 m. We created a set of 5565 viewpoints for all study sites (Lech Valley 602; Stubai Valley 1068; Pustertal 1928; Vinschgau 1967), regularly distributed over the whole study area, by placing a viewpoint every 500 m to account for the landscape variability but to maintain feasible computing time. Each view- point was assigned a unique ID in order to relate all non-spatial information to the specific viewpoint. Viewpoints within forest and settlement areas were excluded from viewshed analysis because of viewing restrictions. For all other viewpoints, three viewsheds, one for each distance zone, were computed, based on the DSM using an eye level of 1.6 m. To obtain the visible land cover for each dis- tance zone, viewsheds were intersected with the corresponding land cover datasets (Fig. 2). A mosaic of the three resulting datasets was created for further analysis because the different zones are seen from the viewpoint as one scene belonging together. To repeat the analysis for an arbitrary number of viewpoints, calculation was automated by generating a GIS-based model writ- ten in Python 2.5 (Python Software Foundation, NH, USA) and using standard routines provided with ArcGIS 9.3TM (ESRI, Redlands, CA, USA). 2.5. Landscape metrics Landscape metrics were calculated for the land cover mosaic using FRAGSTATS Version 3.3. (McGarigal et al., 2002) which includes a variety of metrics describing area, patch, edge and shape properties as well as diversity on three different levels: patch, class or landscape. Selection of landscape metrics can be based on exper- tise or on statistical approaches (Lausch & Herzog, 2002; Riitters et al., 1995). In line with comparable studies (Dramstad et al., 2006; Franco et al., 2003; Hunziker & Kienast, 1999; Palmer, 2004), we selected 60 landscape metrics at landscape level (see Appendix A for details). The land cover mosaics of the 5565 viewpoints were all of the same size. Non-visible areas, classified as background, were assumed to be ‘outside’ the landscape of interest and had no influence on area-based metrics (McGarigal et al., 2002). The selected landscape metrics were subsequently reduced by principal component analysis (PCA). 2.6. Perception survey Landscape metrics can describe landscape in terms of hetero- geneity, diversity, and composition, but they do not reflect human perceptions. The areal extent of the visible landscape is positively correlated to perceived scenic beauty (Germino et al., 2001; Sander & Manson, 2007) and can be assessed by area-based metrics. Land- scape metrics were related to human perceptions through a survey investigating people’s perception of scenic beauty. The survey was based on a questionnaire presenting a set of photographs and con- taining six series related to (1) landscape structure, (2) settlement pattern, (3) forest pattern, (4) presence of water, (5) forest density, and (6) view zones. Each series was made up of four images: one real photograph and three different versions of the original photograph modified with Adobe PhotoShopTM. A seventh series was added at the end of the questionnaire repeating the six original photographs from series 1 to 6. Additionally, we included questions related to demographical information (age, gender, origin). The questionnaire was translated in German and Italian by a professional translator. The respondents were selected in public locations in the study sites on the basis of an equal distribution of age, gender, origin (inhabi- tants and tourist) to represent perceptions of the whole community (Lothian, 1999). A total of 253 persons were interviewed by pre- senting the questionnaire. The respondents were asked to rank U. Schirpke et al. / Landscape and Urban Planning 111 (2013) 1– 12 5 g diff t 1 a q n a s w t p t t i v m P m a t z s w b u T C s Fig. 3. Mean scenic beauty values of picture series 6, representin he four pictures of each series according to scenic beauty (from = least beautiful to 4 = most beautiful). The response rate was 89%, nd the respondents employed in average 5–10 min to fill out the uestionnaire. The rankings of the six different picture series are ot comparable with each other because each series is related to specific theme and the pictures were ranked only within each eries. To compare the different themes represented by series 1–6, e used the seventh series, which repeated the six original pho- ographs from series 1 to 6. The seventh series consisted of six ictures which led to a ranking scale ranging from 1 = least beautiful o 6 = most beautiful. To obtain comparable scenic beauty values for he series 1–6, we calculated a modified scenic beauty value for each mage by multiplying each original value with the scenic beauty alue of the related photograph of series 7. The sixth series was anipulated to obtain different combinations of the view zones: icture 1 shows all three zones, Picture 2 represents the near and iddle zone, Picture 3 shows the near zone, and Picture 4 contains ll three zones but has no foreground elements (Fig. 3). To quan- ify the influence of each view zone, we assigned a weight to each one. First, we computed a scenic beauty value for each zone by ubtracting scenic beauty value of the pictures. Subsequently, the eight of each zone was calculated by dividing each mean scenic eauty value by the scenic beauty value of all zones to obtain val- es between 0 and 1 (Table 2). We used these weights to calculate able 2 alculation of mean scenic beauty value for each view zone, based on mean scenic beauty cenic beauty value by the scenic beauty value of all zones. View zone Calculation Near zone Picture 3 (1.09) Middle zone Picture 3 (1.09) subtracted from Picture 2 (1.80) Far zone Picture 2 (1.80) subtracted from Picture 1 (2.25) All zones Picture 1 (2.25) erent view zones (1 for least beautiful and 4 for most beautiful). the total weighting factors for all images of the questionnaire. After visually identifying the number of view zones of all images accord- ing to the distinguishability of landscape elements as defined in Section 2.2, we obtained a total weighting factor for each image by summing up the weights of the contained view zones. Finally, the weighting factor was applied to the modified scenic beauty val- ues of each picture to take into account the number of view zones present and their influence on scenic beauty. All photograph positions were geo-referenced in the field with GPS. By setting the appropriate view angle and direction of the picture, the views were located on the land cover map (Fig. 4). Non- visible areas were excluded in calculating the viewshed based on the DSM from the position of the photograph. According to the dif- ferent versions of the original photographs, also different land cover maps were created. Based on the adapted land cover maps, land- scape metrics were calculated for all picture views of the survey. 3. Results 3.1. Perception survey The survey suggests that view zones play an important role for the perception of scenic beauty. The higher the number of visi- ble view zones, the better the picture was liked, and foreground values of picture series 6 (see Fig. 3). Weight was obtained by dividing each mean Mean scenic beauty value Weight 1.09 0.48 0.71 0.32 0.45 0.20 2.25 1 6 U. Schirpke et al. / Landscape and Urban Planning 111 (2013) 1– 12 F ap ove m ible ar e c s b ( i 3 a c s o c E e F s ig. 4. (a) The original photograph and (b) the same view in Google Earth with a m ap and the viewshed delimitated by setting view angle and direction. (d) Non-vis lements were preferred to the middle and far zone (Fig. 3). By cal- ulating the weights of each zone, we assessed their influence on cenic beauty. While the near zone contributes by 48% to scenic eauty, the middle zone reaches 32% and the far zone only 20% Table 2). The distribution of the scenic beauty values for the 24 mages is shown in Fig. 5. .2. Statistical analysis For the 5565 viewpoints, we calculated 60 landscape metrics nd selected explanatory variables by means of a principal omponent analysis (PCA) with varimax rotation. The rotated, tandardized components are described by the covariance of the riginal variables and reflect the input variables in few but signifi- ant variables that are absolutely independent (Riitters et al., 1995). leven components with an eigenvalue above 1 were extracted and xplain 93% of the total variance (Appendix B). The first and fourth 0 2 4 6 8 10 12 1-1 1-2 1-3 1-4 2-1 2-2 2-3 2-4 3-1 3-2 3-3 3 Landscape structure Settlement pattern Forest patte M e a n s c e n ic b e a u ty v a lu e Pictu Original preference score Modified prefere most beautiful least beautiful ig. 5. Mean scenic beauty value for each image, showing original, modified (multiplied cenic beauty value weighted by the view zones). rlay. (c) The position of the viewpoint (yellow circle) was placed on the land cover eas were excluded. components consist mainly of area metrics quantifying the area and extent of patches. Whereas the second component comprises dif- ferent types of metrics expressing complexity of patches within landscape, the third component includes only diversity metrics representing richness and evenness to quantify diversity of land- scape. The fourth component consists of different area metrics. Components five, seven and eight are dominated by different shape metrics describing landscape configuration by representing the complexity of patch shape, patch size and patch compaction. The sixth component contains different indices describing landscape fragmentation. Components nine and ten include shape metrics, while the eleventh component is represented by the number of patches. Based on the scenic beauty values of the perception survey and the landscape metrics related to the pictures, we applied a step- wise linear regression analysis to build a model for estimating scenic beauty. Scenic beauty values were entered as a dependent -4 4-1 4-2 4-3 4-4 5-1 5-2 5-3 5-4 6-1 6-2 6-3 6-4 rn Presence of water Forest density View zones re number nce score Weighted preference score with related photograph of series 7) and weighted scenic beauty value (modified U. Schirpke et al. / Landscape and Urban Planning 111 (2013) 1– 12 7 Table 3 Linear regression result with beta-coefficients and significance of the components. Unstandardized coefficients Unstandardized coefficients T Sig Variable B SEB Beta − v p b ( c t 3 w f 3 a w 5 o v a ( u i r t c ( u B s s m d l t s H b ( w i a p 4 b e ( & r f e v w o Component 7 0.519 0.108 Component 3 −0.579 0.129 ariable, while the eleven selected components were used as inde- endent variables. The model identified two predictors of scenic eauty (Table 3) and a good level of prediction was achieved R2 = 0.72, adjusted R2 = 0.69). The first predictor corresponds to omponent 7 (Appendix B) with highest loadings for shape and frac- al index distribution representing shape complexity. Component was selected as second predictor expressing landscape diversity ith high loadings for all six diversity metrics and negative loading or contagion. .3. Scenic beauty Scenic beauty was estimated for all viewpoints outside forests nd settlement areas by applying the regression model. Viewpoints ithin forests returned scenic beauty values from the survey (series ). Viewpoints within settlement areas, which were less than 1% f all viewpoints, were set to no data because no scenic beauty alues were available from the survey. Finally, area-wide maps with raster size of 500 m × 500 m were created for all study sites (Fig. 6). Scenic beauty of the viewpoints ranges between 0.9 and 55.4 Table 4). To examine the spatial variations of scenic beauty, we sed landscape units as applied by Tasser et al. (2009). While sim- lar land cover does not always correspond to the same elevation ange for all study sites, due to diverse climate and agricultural use, he landscape units reflect land cover related to elevation: (1) agri- ulturally used valley bottom, (2) agriculturally used valley slopes, 3) montane forest belt, (4) subalpine forest belt, (5) agriculturally sed alpine pastures, (6) natural alpine grassland, and (7) nival belt. ased on the scenic beauty maps and the delimitation of the land- cape units, we calculated mean values for each unit of all study ites (Table 4). Generally, viewpoints in the valley bottom indicate ean values of scenic beauty or below and landscape pattern is ominated by settlements and grasslands but the visible area is imited by slopes, trees, or buildings (Fig. 7a). The forest belt (mon- ane forest and subalpine forest belt) is characterized by very low cenic beauty due to the view being impaired by trees (Fig. 7b). igh scenic beauty can be found for viewpoints above the forest elt, especially within natural alpine grassland and the nival belt Fig. 7c). The visible area of viewpoints above the tree-line increases ith increasing elevation. The viewsheds are mostly character- zed by complex topography, heterogeneous landscape patterns of lpine pastures in the vicinity, and more homogeneous landscape atterns in the distance. . Discussion and conclusions Daniel (2001) indicated that, in contrast to just perception- ased methods (e.g. Arriaza et al., 2004; Cañas et al., 2009; Hunziker t al., 2008; Zube & Pitt, 1981) or purely expert-based approaches e.g. Bishop & Hulse, 1994; Germino et al., 2001; Herbst, Förster, Kleinschmit, 2009), merging the two opposing approaches could esult in a more effective approach that better represents landscape eatures and human judgments. Accordingly, our GIS-based mod- ling approach combined an automated assessment of the specific iew properties of mountain landscapes and landscape patterns ith a perception-based method, investigating human perceptions f scenic beauty. In a first step, the area seen from a viewpoint was 0.617 4.813 0.000 0.575 −4.492 0.000 examined and, by considering different distance zones, the model accounted for the influence of distance on perceived size, shape, and color of landscape features. The visible area was intersected with land-cover maps, and landscape patterns, expressed by landscape metrics, were related to perceived scenic beauty out of a percep- tion survey by a regression analysis. In line with other studies (de la Fuente de Val et al., 2006; Dramstad et al., 2006; Hunziker & Kienast, 1999; Palmer, 2004), our results confirm the relationship between landscape pattern and scenic beauty. We found that scenic beauty is positively correlated to complexity of patch shape, diversity and structural richness of landscape, whereas large homogeneous areas reduce scenic beauty. The regression model was developed and established for our study region, the Central Alps. In contrast to other studies in the European Alps (Grêt-Regamey et al., 2007; Hunziker & Kienast, 1999) our model can be applied for any viewpoint in mountain regions within Europe, which is considered a human entity, shar- ing common area, culture and behavior patterns (Jordan-Bychkov & Bychkova-Jordan, 2008). Thus, the scenic beauty of any view- point can be compared to all other points throughout Europe. The GIS-based model can also easily be transferred to other regions with similar topographic properties all over the world. It might be necessary to repeat the perception study for other cultural regions where people might perceive scenic beauty differently (Zube & Pitt, 1981). Although the perception of scenic beauty can vary between diverse social groups or different generations within one landscape region (Dramstad et al., 2006; Hunziker et al., 2008; Tveit, 2009), landscape variations are generally much greater than the variations between observer’s judgments (Daniel, 2001). Input data are based on digital elevation models and land cover maps, which are usually available for most areas in Europe and comparable satellite-based data exist for many regions world-wide. There are no restrictions regarding spatial and non-spatial resolution of the data. Availability of high resolution data used for the near zone is more difficult and might necessitate new mapping. The high resolution data can be substituted e.g. by CORINE land cover (EEA, 2009) for first assess- ments but lower resolution of input data reduces the quality of the model and smoothes the values for scenic beauty. Another advan- tage of our method is that scenic beauty can be predicted for any viewpoint which allows different applications: (1) it is possible to perform area-wide mapping by distributing viewpoints over the whole area, or (2) to explore selected zones, for instance those of touristic interest, along roads or hiking trails by placing the view- points along defined features. The quality of assessments depends on the resolution of the input data, which determines the highest possible density of viewpoints because view properties and land- scape composition are highly variable and can change within very short distances. In comparing scenic beauty of diverse landscape units, major differences can be observed between viewpoints above the forest belt, characterized by high scenic beauty, and viewpoints within the forest belt, to which low scenic beauty was attributed. Supported by Ribe (2009), the survey indicated that structure and diversity influence the perception of scenic beauty in timber stands. Open forests are generally preferred. On the other hand, forests are highly appreciated for recreational activities and are related to spiritual, esthetic, cultural, and educational values (Scarpa et al., 2000). Close 8 U. Schirpke et al. / Landscape and Urban Planning 111 (2013) 1– 12 lley a t l u h a L f 2 g v p T M f Fig. 6. Scenic beauty of (a) Lech Valley, (b) Pustertal, (c) Stubai Va o average scenic beauty was found for the agriculturally used val- ey bottom and the natural alpine grassland belt. In these landscape nits, landscape pattern and structure are strongly influenced by uman activities. In many Alpine regions, considerable changes in griculture and forestry could be observed (Rutherford et al., 2008). and abandonment mainly affected alpine pastures, and natural orest re-growth leads to altered landscape patterns (Sitzia et al., 010), a general shift from a patchy mosaic toward a more homo- eneous scenery. The increase in forest not only means restricted iew and a loss of viewpoints but affects scenic beauty of any view- oint, because scenic beauty is conditioned by the composition and able 4 ean values of scenic beauty for different landscape units: (1) agriculturally used valley orest belt, (5) agriculturally used alpine pastures, (6) natural alpine grassland, and (7) ni Landscape unit 1 2 3 Area [km2 ] 98 45 146 Mean elevation [m a.s.l.] 1272 1306 1352 Scenic beauty (N = 5565) Mean 10.0 5.9 1. SDa 4.1 4.7 2. Minimum 0.9 0.9 0. Maximum 16.0 16.9 15. a Standard deviation. nd (d) Vinschgau. High values correspond to great scenic beauty. pattern of the whole visible area. As a consequence, a decrease of scenic beauty, especially along hiking tracks, might affect the attractiveness of the area. Resulting maps offer a basis for various applications, especially in landscape planning or tourism geogra- phy. The GIS-based model can support scenic beauty assessments in the decision making process for future policies or to evaluate already implemented measures, e.g. Tasser et al. (2012) emphasize that mountain farming is important to maintain the cultural land- scapes of tourist destinations and abandonment of agricultural land can be avoided by payments for landscape preservation. Regarding the Europe 2020 Strategy, the Common Agricultural Policy (CAP) bottom, (2) agriculturally used valley slopes, (3) montane forest belt, (4) subalpine val belt. High values correspond to great scenic beauty. 4 5 6 7 Total 261 391 372 78 1389 1850 2129 2590 3019 2081 8 2.1 9.7 14.8 17.5 9.1 9 3.8 5.8 4.8 5.2 7.1 9 0.9 0.9 0.9 5.3 0.9 8 46.5 48.2 48.5 55.4 55.4 U. Schirpke et al. / Landscape and Urban Planning 111 (2013) 1– 12 9 used p v l G w B s s m a i Fig. 7. Typical landscape patterns seen from viewpoints in (a) the agriculturally roposes to relate financial support in the future to ecosystem ser- ices (European Commission, 2010). As cultural ecosystem services ike the recreational value are often expressed by scenic beauty (de root et al., 2010), our proposed GIS-based model allows region- ide assessments for evaluating payments for ecosystem services. y calculating scenic beauty maps based on future land-use/-cover cenarios, future impacts can be visualized and management deci- ions adapted. For the tourism sector in particular, our proposed odel offers great potential to strengthen the competitiveness of region by preserving the landscape or by creating the necessary nfrastructure to access places of great scenic beauty. Acronym Landscape metrics TA Total area NP Number of patches PD Patch density LPI Largest patch index TE Total edge ED Edge density LSI Landscape shape index AREA MN Mean patch area distribution AREA AM Area-weighted mean patch area distribution AREA MD Median patch area distribution AREA RA Range patch area distribution AREA SD Standard deviation patch area distribution AREA CV Coefficient of variation patch area distribution GYRATE MN Mean radius of gyration distribution GYRATE AM Area-weighted mean radius of gyration distribution GYRATE MD Median radius of gyration distribution GYRATE RA Range radius of gyration distribution GYRATE SD Standard deviation radius of gyration distribution GYRATE CV Coefficient of variation radius of gyration distribution SHAPE MN Mean shape index distribution SHAPE AM Area-weighted mean shape index distribution SHAPE MD Median shape index distribution SHAPE RA Range shape index distribution SHAPE SD Standard deviation shape index distribution SHAPE CV Coefficient of variation shape index distribution FRAC MN Mean fractal index distribution FRAC AM Area-weighted mean fractal index distribution FRAC MD Median fractal index distribution FRAC RA Range fractal index distribution FRAC SD Standard deviation fractal index distribution valley bottom, (b) the subalpine forest belt, and (c) the natural alpine grassland. Acknowledgements We would like to thank the three anonymous reviewers for help- ing to improve the manuscript. We also thank Brigitte Scott for language editing. Special thanks to Sonja Hölzler who has carried out the perception survey, as well as to everyone who partici- pated in the preference study. This study was supported by the ERA-Net BiodivERsA, with the national funder FWF, part of the 2008 BiodivERsA call for research proposals and the KuLaWi project (INTERREG IV – EU project (Agri)cultural landscape – Strategies for the cultural landscape of the future, project n. 4684, CUP: B26D09000010007). Appendix A. Variations of landscape metrics for the 5565 viewpoints. Mean Min Max S.D. 14212.8 255.0 756241.0 16383.4 125.1 21.0 527.0 62.2 2.7 0.0 95.9 5.4 24.3 4.0 89.3 10.8 65956.5 3220.0 1904140.0 45019.5 8.2 0.5 83.4 7.3 8.5 3.4 17.7 1.6 136.9 1.0 3959.4 137.0 1434.1 7.3 222591.7 3180.3 13.2 0.1 200.0 29.5 3047.9 26.0 290400.0 4620.1 402.4 2.6 29421.9 508.7 344.5 117.4 1106.1 117.4 294.1 28.1 1097.1 182.6 1694.8 116.1 24779.2 869.0 114.6 14.1 745.9 110.6 2930.1 204.2 32334.5 1567.1 475.0 38.6 3318.6 245.7 174.6 74.4 426.7 39.0 1.4 1.2 1.8 0.1 2.0 1.2 5.7 0.3 1.3 1.0 1.6 0.1 3.2 1.1 15.1 1.4 0.6 0.3 1.1 0.1 39.2 22.1 67.7 5.7 1.1 1.0 1.1 0.0 1.1 1.0 1.2 0.0 1.1 1.0 1.1 0.0 0.2 0.1 0.4 0.1 0.1 0.0 0.1 0.0 1 A w b 0 U. Schirpke et al. / Landscape and Urban Planning 111 (2013) 1– 12 Acronym Landscape metrics Mean Min Max S.D. FRAC CV Coefficient of variation fractal index distribution 5.3 3.2 7.3 0.7 PARA MN Mean perimeter-area ratio distribution 677.5 84.6 1362.8 282.3 PARA AM Area-weighted mean perimeter-area ratio distribution 48.8 5.8 506.8 35.3 PARA MD Median perimeter-area ratio distribution 488.3 30.0 1500.0 362.3 PARA RA Range perimeter-area ratio distribution 1966.8 388.1 1995.5 109.6 PARA SD Standard deviation perimeter-area ratio distribution 614.1 112.5 852.7 115.7 PARA CV Coefficient of variation perimeter-area ratio distribution 103.3 40.9 247.0 32.9 CONTIG MN Mean contiguity index distribution 0.6 0.3 0.9 0.1 CONTIG AM Area-weighted mean contiguity index distribution 1.0 0.7 1.0 0.0 CONTIG MD Median contiguity index distribution 0.7 0.2 1.0 0.2 CONTIG RA Range contiguity index distribution 1.0 0.2 1.0 0.0 CONTIG SD Standard deviation contiguity index distribution 0.3 0.1 0.4 0.1 CONTIG CV Coefficient of variation contiguity index distribution 52.9 7.3 104.6 18.7 PAFRAC Perimeter-area fractal dimension 1.1 1.0 1.4 0.0 CONTAG Contagion 79.9 50.4 95.4 5.5 PLADJ Percentage of like adjacencies 97.6 74.7 99.7 1.8 IJI Interspersion & juxtaposition index 39.9 3.9 74.0 11.2 COHESION Patch cohesion index 99.0 89.5 100.0 0.8 DIVISION Landscape division index 0.9 0.2 1.0 0.1 MESH Effective mesh size 1434.1 7.3 222591.7 3180.3 SPLIT Splitting index 10.6 1.3 56.2 5.1 PR Patch richness 13.7 4.0 42.0 4.9 PRD Patch richness density 0.3 0.0 10.2 0.7 RPR Relative patch richness 137.3 40.0 420.0 49.4 SHDI Shannon’s diversity index 1.0 0.2 2.8 0.3 SIDI Simpson’s diversity index 0.5 0.1 0.9 0.1 MSIDI Modified Simpson’s diversity index 0.8 0.1 2.6 0.3 SHEI Shannon’s evenness index 0.4 0.1 0.9 0.1 SIEI Simpson’s evenness index 0.6 0.1 1.0 0.1 MSIEI Modified Simpson’s evenness index 0.3 0.0 0.8 0.1 ppendix B. Rotated component matrix. Selection of the 60 landscape metrics by means of principal component analysis (PCA) ith varimax rotation and Kaiser normalization. Rotation converged in 11 iterations. The resulting components are described y the covariance of the ingoing variables. All values above 0.5 or beyond −0.5 are displayed in bold. Component Communalities Variablesa 1 2 3 4 5 6 7 8 9 10 11 Cumulative % of variance explained 20.1 34.0 46.1 57.2 65.5 72.4 78.4 82.5 86.2 89.5 92.5 TA −0.405 −0.169 −0.179 0.791 −0.024 0.265 −0.133 0.033 0.025 −0.003 0.087 0.948 NP 0.460 0.173 0.040 0.018 0.356 0.304 −0.210 0.057 0.038 0.203 0.623 0.941 PD 0.169 0.901 0.157 −0.073 −0.014 −0.044 −0.094 0.014 −0.012 0.090 0.087 0.896 LPI 0.140 0.083 −0.151 0.040 0.001 −0.910 0.043 0.134 0.026 0.011 0.036 0.901 TE −0.071 −0.262 0.123 0.642 0.218 0.390 −0.023 0.254 0.009 0.121 0.259 0.848 ED 0.275 0.763 0.390 −0.145 0.158 0.024 0.132 0.192 −0.033 0.077 0.022 0.918 LSI −0.202 0.251 0.049 0.108 0.217 0.620 −0.160 0.271 0.053 0.138 0.485 0.906 AREA MN −0.632 −0.197 −0.272 0.603 −0.073 0.136 −0.104 0.028 −0.008 −0.083 −0.075 0.925 AREA AM −0.008 −0.050 −0.063 0.971 −0.028 −0.080 0.023 0.004 −0.024 −0.040 −0.030 0.961 AREA MD −0.740 0.008 −0.157 0.171 0.042 0.180 −0.250 −0.107 −0.007 −0.009 0.098 0.720 AREA RA −0.153 −0.112 −0.125 0.960 −0.025 −0.040 −0.039 0.052 0.000 −0.024 0.007 0.980 AREA SD −0.285 −0.146 −0.184 0.916 −0.052 −0.015 −0.029 0.051 −0.012 −0.059 −0.055 0.989 AREA CV 0.532 0.181 −0.098 0.041 0.249 −0.517 −0.081 0.143 0.039 0.168 0.482 0.945 GYRATE MN −0.790 −0.297 −0.256 0.287 −0.100 0.196 −0.102 0.029 −0.003 −0.097 −0.103 0.940 GYRATE AM −0.350 −0.383 −0.349 0.644 −0.004 −0.195 −0.049 0.276 0.038 −0.016 0.007 0.923 GYRATE MD −0.854 −0.116 −0.160 0.188 −0.037 0.203 −0.172 −0.078 −0.010 −0.059 0.022 0.887 GYRATE RA −0.445 −0.382 −0.288 0.569 0.012 −0.072 −0.120 0.272 0.063 0.008 0.060 0.852 GYRATE SD −0.612 −0.400 −0.334 0.447 −0.079 0.055 −0.064 0.177 0.027 −0.073 −0.113 0.909 GYRATE CV 0.659 −0.113 −0.157 0.096 0.229 −0.443 −0.057 0.262 0.037 0.145 0.304 0.916 SHAPE MN −0.135 −0.077 0.051 0.017 0.356 −0.053 0.852 0.167 −0.010 −0.055 −0.171 0.942 SHAPE AM −0.074 0.010 0.001 0.244 0.158 −0.203 0.093 0.900 −0.032 0.001 0.048 0.953 SHAPE MD −0.136 −0.072 0.042 −0.002 −0.054 −0.020 0.926 0.055 0.013 −0.001 −0.018 0.891 SHAPE RA 0.268 0.021 0.110 −0.011 0.808 0.053 −0.051 0.049 −0.007 0.059 0.228 0.801 SHAPE SD 0.139 −0.037 0.085 0.016 0.916 −0.025 0.223 0.190 −0.016 −0.010 −0.046 0.956 SHAPE CV 0.202 −0.023 0.081 0.012 0.931 −0.010 −0.006 0.167 −0.017 0.006 0.008 0.942 FRAC MN 0.466 0.233 0.111 −0.119 0.246 −0.166 0.747 0.006 0.031 0.072 −0.009 0.951 FRAC AM 0.009 0.287 0.105 0.092 0.122 −0.183 0.189 0.872 −0.040 −0.001 0.039 0.949 FRAC MD 0.298 0.180 0.118 −0.099 0.019 −0.154 0.852 0.057 0.017 0.061 0.058 0.906 FRAC RA 0.486 0.127 0.054 −0.053 0.729 0.065 0.103 −0.106 0.062 0.144 0.133 0.859 FRAC SD 0.615 0.183 0.026 −0.075 0.624 −0.039 0.254 −0.063 0.071 0.083 −0.034 0.889 FRAC CV 0.615 0.174 0.019 −0.071 0.636 −0.029 0.214 −0.066 0.073 0.083 −0.037 0.882 PARA MN 0.830 0.363 0.077 −0.135 0.252 −0.123 −0.054 −0.053 0.062 0.130 0.133 0.968 PARA AM 0.249 0.886 0.314 −0.164 0.039 −0.010 0.058 0.098 −0.031 0.055 0.031 0.992 a R A B B B C C D d d D D D E E F F U. Schirpke et al. / Landscape and Urban Planning 111 (2013) 1– 12 11 Component Communalities Variablesa 1 2 3 4 5 6 7 8 9 10 11 PARA MD 0.732 0.389 0.088 −0.132 0.269 −0.131 −0.080 −0.066 −0.020 0.091 0.312 0.919 PARA RA 0.154 −0.047 −0.014 −0.008 0.007 0.000 0.024 −0.020 0.969 0.021 0.007 0.967 PARA SD 0.806 0.030 −0.006 −0.071 0.194 0.045 −0.034 −0.032 0.334 0.122 −0.281 0.902 PARA CV −0.840 −0.305 −0.124 0.182 −0.158 0.207 −0.104 0.009 0.122 −0.109 −0.106 0.965 CONTIG MN −0.831 −0.368 −0.081 0.141 −0.238 0.135 0.053 0.052 −0.058 −0.131 −0.137 0.972 CONTIG AM −0.255 −0.878 −0.323 0.172 −0.036 0.008 −0.060 −0.101 0.032 −0.051 −0.034 0.990 CONTIG MD −0.747 −0.391 −0.089 0.137 −0.254 0.149 0.078 0.063 0.021 −0.088 −0.284 0.923 CONTIG RA 0.142 −0.089 −0.030 0.002 0.011 0.003 0.014 −0.032 0.967 0.022 0.017 0.966 CONTIG SD 0.788 −0.017 −0.020 −0.065 0.169 0.058 −0.038 −0.044 0.342 0.118 −0.328 0.900 CONTIG CV 0.841 0.292 0.048 −0.111 0.280 −0.059 −0.124 −0.062 0.118 0.118 0.079 0.942 PAFRAC 0.293 0.567 0.270 −0.148 0.283 −0.244 0.154 0.312 −0.147 0.022 0.295 0.872 CONTAG 0.007 −0.358 −0.892 0.136 −0.057 −0.014 −0.094 −0.013 0.048 0.124 −0.006 0.973 PLADJ −0.249 −0.886 −0.314 0.164 −0.039 0.010 −0.058 −0.098 0.031 −0.055 −0.031 0.992 IJI 0.285 0.461 0.086 −0.182 0.272 −0.262 0.294 −0.291 −0.014 −0.115 0.231 0.715 COHESION −0.256 −0.860 −0.301 0.167 −0.011 −0.054 −0.017 0.140 0.017 −0.068 −0.071 0.957 DIVISION −0.165 −0.079 0.165 −0.035 0.004 0.909 −0.070 −0.119 −0.006 −0.001 −0.017 0.908 MESH −0.008 −0.050 −0.063 0.971 −0.028 −0.080 0.023 0.004 −0.024 −0.040 −0.030 0.961 SPLIT −0.322 −0.037 0.126 0.014 0.000 0.810 −0.175 −0.177 0.044 0.065 0.084 0.853 PR 0.391 0.267 0.094 −0.075 0.105 0.036 0.028 0.007 0.038 0.849 0.072 0.979 PRD 0.084 0.863 0.131 −0.067 −0.070 −0.103 −0.024 0.024 −0.028 0.203 −0.116 0.846 RPR 0.391 0.267 0.094 −0.075 0.105 0.036 0.028 0.007 0.038 0.849 0.072 0.979 SHDI 0.221 0.426 0.805 −0.147 0.099 0.059 0.108 0.000 0.026 0.207 0.037 0.969 SIDI 0.215 0.168 0.915 −0.108 0.067 0.130 0.014 0.026 0.025 0.161 0.023 0.973 MSIDI 0.189 0.272 0.889 −0.114 0.074 0.136 0.048 0.030 0.021 0.184 0.001 0.974 SHEI −0.040 0.294 0.914 −0.127 0.046 0.014 0.084 −0.015 −0.044 −0.131 0.006 0.968 SIEI 0.148 0.139 0.948 −0.102 0.053 0.121 0.001 0.023 0.003 0.093 0.016 0.976 MSIEI −0.020 0.142 0.960 −0.091 0.024 0.099 0.012 0.019 −0.036 −0.088 −0.022 0.972 For full names of acronyms see Appendix A. eferences rriaza, M., Cañas-Ortega, J. F., Cañas-Madueño, J. A., & Ruiz-Aviles, P. (2004). Assessing the visual quality of rural landscapes. Landscape and Urban Planning, 69, 115–125. eza, B. B. (2010). The aesthetic value of a mountain landscape: A study of the Mt. Everest Trek. Landscape and Urban Planning, 97, 306–317. ishop, I. D., & Hulse, D. W. (1994). Prediction of scenic beauty using mapped data and geographic information systems. Landscape and Urban Planning, 30, 59–70. ishop, I. D. (2003). Assessment of visual qualities, impacts, and behaviours, in the landscape, by using measures of visibility. Environment and Planning B: Planning and Design, 30, 677–688. añas, I., Ayuga, E., & Ayuga, F. (2009). A contribution to the assessment of scenic quality of landscapes based on preferences expressed by the public. Land Use Policy, 26(4), 1173–1181. hhetri, P., & Arrowsmith, C. (2008). GIS-based modelling of recreational potential of nature-based tourist destinations. Tourism Geographies, 10(2), 233–257. aniel, T. C. (2001). Whither scenic beauty? Visual landscape quality assessment in the 21st century. Landscape and Urban Planning, 54, 267–281. e Groot, R. S., Alkemade, R., Braat, L., Hein, L., & Willemen, L. (2010). Challenges in integrating the concept of ecosystem services and values in landscape planning, management and decision making. Ecological Complexity, 7, 260–272. e la Fuente de Val, G., Atauri, J. A., & de Lucio, J. V. (2006). Relationship between landscape visual attributes and spatial pattern indices: A test study in Mediterranean-climate landscapes. Landscape and Urban Planning, 77, 393–407. ramstad, W. E., Tveit, M. S., Fjellstad, W. J., & Fry, G. L. A. (2006). Relationships between visual landscape preferences and map-based indicators of landscape structure. Landscape and Urban Planning, 78, 465–474. udley, N., Baldock, D., Nasi, R., & Stolton, S. (2005). Measuring biodiversity and sustainable management in forests and agricultural landscapes. Philosophical Transactions of the Royal Society of London Series B: Biological Sciences, 360, 457–470. ullinger, S., Dirnböck, T., & Grabherr, G. (2003). Patterns of shrub invasion into high mountain grasslands of the Northern Calcareous Alps, Austria. Arctic, Antarctic and Alpine Research, 35, 434–441. EA. (2009). Corine land cover 2000 (CLC2000) seamless vector database – version 10/2009. Available from http://www.eea.europa.eu/data-and-maps/ data/corine-land-cover-2000-clc2000-seamless-vector-database-1 (accessed 23.04.10). uropean Commission. (2010). The CAP towards 2020: Meeting the food, natural resources and territorial challenges of the future. Available from http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=COM:2010:0672: FIN:en:PDF (accessed 13.07.12). ischer, M., Rudmann-Maurer, K., Weyand, A., & Stöcklin, J. (2008). Agricultural Franco, D., Franco, D., Mannino, I., & Zanetto, G. (2003). The impact of agroforestry networks on scenic beauty estimation: The role of a landscape ecological net- work on a socio-cultural process. Landscape and Urban Planning, 62, 119–138. García, L., Hernández, J., & Ayuga, F. (2003). Analysis of the exterior colour of agroin- dustrial buildings: A computer aided approach to landscape integration. Journal of Environmental Management, 69, 93–104. Germino, M. J., Reiners, W. A., Blasko, B. J., McLeod, D., & Bastian, C. T. (2001). Esti- mating visual properties of Rocky Mountain landscapes using GIS. Landscape and Urban Planning, 53, 71–83. 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. Environment and Planning B: Planning and Design, 34, 50–67. Herbst, H., Förster, M., & Kleinschmit, B. (2009). Contribution of landscape metrics to the assessment of scenic quality – The example of the landscape structure plan Havelland/Germany. Landscape Online, 10, 1–17. Horvath, H. (1995). Estimation of the average visibility in central Europe. Atmospheric Environment, 29(2), 241–246. Hunziker, M., Felber, P., Gehring, K., Buchecker, M., Bauer, N., & Kienast, F. (2008). Evaluation of landscape change by different social groups. Mountain Research and Development, 28, 140–147. Hunziker, M., & Kienast, F. (1999). Potential impacts of changing agricultural activi- ties on scenic beauty – A prototypical technique for automated rapid assessment. Landscape Ecology, 14, 161–176. Jarvis, A., Reuter, H.I., Nelson, A., & Guevara, E. (2008). Hole-filled SRTM for the globe version 4. Available from the CGIAR-CSI SRTM 90 m database http://srtm.csi.cgiar.org Jordan-Bychkov, T., & Bychkova-Jordan, B. (2008). The European culture area: A sys- tematic geography. Maryland: Rowman & Littlefield. Kearney, A. R., Bradley, G. A., Petrich, C. H., Kaplan, R., Kaplan, S., & Simpson-Colebank, D. (2008). Public perception as support for scenic quality regulation in a nation- ally treasured landscape. Landscape and Urban Planning, 87(2), 117–128. Kim, Y., Rana, S., & Wise, S. (2004). Exploring multiple viewshed analysis using terrain features and optimisation techniques. Computers & Geosciences, 30, 1019–1032. Lamb, R. J., & Purcell, A. T. (1990). Perception of naturalness in landscape and its relationship to vegetation structure. Landscape and Urban Planning, 19, 333–352. Lausch, A., & Herzog, F. (2002). Applicability of landscape metrics for the monitoring of landscape change: issues of scale, resolution and interpretability. Ecological Indicators, 2(1-2), 3–15. Litton, R. B. (1977). River landscape quality and its assessment. In Proceedings of the symposium on River Recreation Management and Research. Gen. Tech. Re NC-28, Northcentral For. Exp. Stn. US Department of Agriculture St. Paul, MN, (pp. 46–54). Lothian, A. (1999). Landscape and the philosophy of aesthetics: Is landscape quality inherent in the landscape or in the eye of the beholder? Landscape and Urban Planning, 44, 177–198. McGarigal, K., Cushman, S.A., Neel, M.C., & Ene, E. (2002). FRAGSTATS: Spatial land use and biodiversity in the Alps. Mountain Research and Development, 28, 148–155. liri, F. (1984). Synoptische Klimatographie der Alpen zwischen Mont Blanc und Hohen Tauern (Synoptic climatology of the Alps between Mont Blanc and Hohen Tauern). Innsbruck: Wagner. pattern analysis Program for categorical maps. Computer software program produced by the authors at the University of Massachusetts, Amherst. Avail- able from http://www.umass.edu/landeco/research/fragstats/fragstats.html (accessed 24.11.11). Meitner, M. J. (2004). Scenic beauty of river views in the Grand Canyon: Relating perceptual judgments to locations. Landscape and Urban Planning, 68, 3–13. http://www.eea.europa.eu/data-and-maps/data/corine-land-cover-2000-clc2000-seamless-vector-database-1 http://www.eea.europa.eu/data-and-maps/data/corine-land-cover-2000-clc2000-seamless-vector-database-1 http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=COM:2010:0672:FIN:en:PDF http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=COM:2010:0672:FIN:en:PDF http://srtm.csi.cgiar.org/ http://www.umass.edu/landeco/research/fragstats/fragstats.html 1 and U M O P R R R R S S S Weinstoerffer, J., & Girardin, P. (2000). Assessment of the contribution of land use 2 U. Schirpke et al. / Landscape ichel, N., Burel, F., Legendre, P., & Butet, A. (2007). Role of habitat and landscape in structuring small mammal assemblages in hedgerow networks of contrasted farming landscapes in Brittany, France. Landscape Ecology, 22, 1241–1253. de, Å., Fry, G., Tveit, M. S., Messager, P., & Miller, D. (2009). Indicators of perceived naturalness as drivers of landscape preference. Journal of Environmental Man- agement, 90, 375–383. almer, J. F. (2004). Using spatial metrics to predict scenic perception in a changing landscape: Dennis, Massachusetts. Landscape and Urban Planning, 69, 201–218. aitz, K., & Dakhil, M. (1988). Recreational choices and environmental preference. Annals of Tourism Research, 15, 357–370. ibe, R. G. (2009). In-stand scenic beauty of variable retention harvests and mature forests in the U.S. Pacific Northwest: The effects of basal area, density, retention pattern and down wood. Journal of Environmental Management, 91, 245–260. iitters, K. H., O’Neill, R. V., Hunsaker, C. T., Wickham, J. D., Yankee, D. H., Timmins, S. P., et al. (1995). A factor analysis of landscape pattern and structure metrics. Landscape Ecology, 10, 23–39. utherford, G. N., Bebi, P., Edwards, P. J., & Zimmermann, N. E. (2008). Assessing land-use statistics to model land cover change in a mountainous landscape in the European Alps. Ecological Modelling, 212, 460–471. ander, H. A., & Manson, S. M. (2007). Heights and locations of artificial structures in viewshed calculation: How close is close enough? Landscape and Urban Planning, 82, 257–270. carpa, R., Chilton, S. M., Hutchinson, W. G., & Buongiorno, J. (2000). Valuing the recreational benefits from the creation of nature reserves in Irish forests. Eco- logical Economics, 33, 237–250. chneeberger, N., Bürgi, M., & Kienast, F. (2007). Rates of landscape change at the northern fringe of the Swiss Alps: Historical and recent tendencies. Landscape and Urban Planning, 80, 127–136. rban Planning 111 (2013) 1– 12 Scott, D., Jones, B., & Konopek, J. (2007). Implications of climate and environmental change for nature-based tourism in the Canadian Rocky Mountains: A case study of Waterton Lakes National Park. Tourism Management, 28, 570–579. Shang, H., & Bishop, I. D. (2000). Visual thresholds for detection, recogni- tion and visual impact in landscape settings. Environmental Psychology, 20, 125–140. Sitzia, T., Semenzato, P., & Trentanovi, G. (2010). Natural reforestation is changing spatial patterns of rural mountain and hill landscapes: A global overview. Forest Ecology and Management, 259, 1354–1362. Tasser, E., Ruffini, F., & Tappeiner, U. (2009). An integrative approach for analysing landscape dynamics in diverse cultivated and natural mountain areas. Landscape Ecology, 24, 611–628. Tasser, E., Schermer, M., Siegl, G., & Tappeiner, U. (2012). Wir LandschaftMacher – Vom Sein und Werden der Kulturlandschaft in Nord-, Süd- und Osttirol (We landscape maker – development of the cultural landscape in North, South and East Tyrol). Bozen: Athesia. Tveit, M. S. (2009). Indicators of visual scale as predictors of landscape prefer- ence; a comparison between groups. Journal of Environmental Management, 90, 2882–2888. Wallentin, G., Tappeiner, U., Strobl, J., & Tasser, E. (2008). Understanding alpine tree line dynamics: An individual-based model. Ecological Modelling, 218, 235–246. pattern and intensity to landscape quality: Use of a landscape indicator. Ecolog- ical Modelling, 130, 95–109. Zube, E. H., & Pitt, D. G. (1981). Cross-cultural perceptions of scenic and heritage landscapes. Landscape Planning, 8(1), 69–87. Predicting scenic beauty of mountain regions 1 Introduction 2 Methods 2.1 Study sites 2.2 Distance zones 2.3 Data collection 2.4 Visibility analysis 2.5 Landscape metrics 2.6 Perception survey 3 Results 3.1 Perception survey 3.2 Statistical analysis 3.3 Scenic beauty 4 Discussion and conclusions Acknowledgements Appendix A Variations of landscape metrics for the 5565 viewpoints. Appendix B Rotated component matrix. Selection of the 60 landscape metrics by means of principal component analysis (PCA) ... References