Landscape and Urban Planning 78 (2006) 465–474

Relationships between visual landscape preferences and map-based indicators of landscape structure W.E. Dramstad a,∗ , M. Sundli Tveit b,1 , W.J. Fjellstad a,2 , G.L.A. Fry b,3 b

a Norwegian Institute of Land Inventory, P.O. Box 115, N-1431 As, ˚ Norway ˚ Norway Department of Landscape Architecture & Spatial Planning, Norwegian University of Life Sciences, P.O. Box 5029, N-1432 As,

Received 11 May 2005; received in revised form 14 December 2005; accepted 15 December 2005 Available online 12 July 2006

Abstract There is increasing awareness of the need to monitor trends in our constantly changing agricultural landscapes. Monitoring programmes often use remote sensing data and focus on changes in land cover/land use in relation to values such as biodiversity, cultural heritage and recreation. Although a wide range of indicators is in use, landscape aesthetics is a topic that is frequently neglected. Our aim was to determine whether aspects of landscape content and configuration could be used as surrogate measures for visual landscape quality in monitoring programmes based on remote sensing. In this paper, we test whether map-derived indicators of landscape structure from the Norwegian monitoring programme for agricultural landscapes are correlated with visual landscape preferences. Two groups of people participated: (1) locals and (2) non-local students. Using the total dataset, we found significant positive correlations between preferences and spatial metrics, including number of land types, number of patches and land type diversity. In addition, preference scores were high where water was present within the mapped image area, even if the water itself was not visible in the images. When the dataset was split into two groups, we found no significant correlation between the preference scores of the students and locals. Whilst the student group preferred images portraying diverse and heterogeneous landscapes, neither diversity nor heterogeneity was correlated with the preference scores of the locals. We conclude that certain indicators based on spatial structure also have relevance in relation to landscape preferences in agricultural landscapes. However, the finding that different groups of people prefer different types of landscape underlines the need for care when interpreting indicator values. © 2006 Elsevier B.V. All rights reserved. Keywords: Agriculture; Landscape metrics; Monitoring; Photographs

1. Introduction Most people, if questioned, will have an opinion as to whether a particular landscape is aesthetically pleasing or not, and the role of everyday landscapes in the well being of people is receiving increased focus (Hartig et al., 2003; Kaplan et al., 1998). Aesthetic issues are controversial in many ways and studies of landscape aesthetics are no exception (see, for example, Ndubisi, 2002; Parsons and Daniel, 2002). Commonly, criticism in landscape aesthetics refers to subjectivity, lack of standardization

∗

Corresponding author. Tel.: +47 64 94 96 84; fax: +47 64 94 97 86. E-mail addresses: [email protected] (W.E. Dramstad), [email protected] (M.S. Tveit), [email protected] (W.J. Fjellstad), [email protected] (G.L.A. Fry). 1 Tel.: +47 64 96 53 65. 2 Tel.: +47 64 94 97 04. 3 Tel.: +47 64 96 53 62. 0169-2046/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.landurbplan.2005.12.006

in methodology, non-transparent application of values and lack of replicability (even among experts) (Bruns and Green, 2001; Daniel, 2001; Ndubisi, 2002; Terkenli, 2001). In spite of efforts to develop methods that can be accepted throughout the scientific community (see review by Ndubisi, 2002), none has hitherto gained general acceptance. One unfortunate consequence of this can be avoidance of the issue by excluding consideration of the visual aspects of landscape altogether. Despite these problems, there is an increasing demand for the visual landscape to be included in landscape policy, management and planning as well as landscape monitoring (Tahvanainen et al., 2002; Tress et al., 2001). Landscape issues have recently moved up the political agenda in Europe (Wascher, 2000), as expressed by the development and ratification of the European Landscape Convention (Council of Europe, 2000). The Convention requires each signatory party to identify its own landscapes, analyse their characteristics and the forces and pressures transforming them and take note of changes (Article 6: Council of

466

W.E. Dramstad et al. / Landscape and Urban Planning 78 (2006) 465–474

Europe, 2000). The Landscape Convention emphasizes its relevance to both “landscapes that might be considered outstanding as well as everyday or degraded landscapes” (Article 2). Furthermore, the Convention outlines that “Landscape must become a mainstream political concern.” Being able to measure changes in the visual landscape through already existing monitoring programmes could ease this task. The lack of an easily accessible methodology to deal with the visual landscape issue frequently hampers the inclusion of visual aspects entirely. One possible approach to meet this challenge is to search for indicators of visual landscape quality that can be derived from data on landscape structure. Our main aim in the study presented here was to determine whether aspects of landscape composition could be used as surrogate measures for visual landscape quality in monitoring programmes based on remote sensing. Studies of human landscape preferences have been based on several different approaches. Zube (1984) identifies three different paradigms in landscape assessment; ‘professional’ where the trained expert interprets the landscape, ‘behavioural’ where biological and evolutionary principles are used to explain landscape preferences and ‘humanistic’ where attitudes, beliefs and ideas of each individual observer are in focus. Dearden (1987) questioned whether beauty is inherent in objects, or in the eye of the beholder and there is discussion about the degree to which personal attributes and experience influence landscape perception, and the extent to which landscape preferences represent learned behaviour (Meinig, 1976). Daniel (2001) describes this history as a controversy of the objective and subjective models, i.e. whether the aesthetics quality is to be found in properties of the objective of study or in the subject studying it. More recently, we have seen approaches to landscape aesthetics that accept a mixture of cultural and biological forces as explaining human landscape preference (Tress et al., 2001). Rapid and wide-ranging changes in landscapes in general, and agricultural landscapes in particular, have caused politicians and management authorities to recognise a need for timely information on both landscape state and change. Many countries and agencies are therefore working to develop indicators and establish ways to monitor and report on agricultural landscape change (see, for example, Defra, 2004; European Environment Agency, 2004; Eurostat, 2003; OECD, 2004; Piorr, 2003). In 1998, the Norwegian Ministry of Food and Agriculture, in cooperation with the Ministry of the Environment, initiated a monitoring programme focusing on agricultural landscapes (Dramstad et al., 2002). In the Norwegian monitoring programme, as well as in other work on agri-environmental indicators, the visual landscape has been a problematic issue since quantitative indicators of visual quality have proven difficult to find (Defra, 2004; Dramstad and Sogge, 2003; OECD, 2000). Like numerous other monitoring programmes, the Norwegian monitoring programme for agricultural landscapes (known as the 3Q-programme) is based on data collected through aerial photography, i.e. viewing the landscape from a birds-eye perspective. Landscape preference studies, on the other hand, have to a very large extent used landscape photographs as landscape surrogates (see, for example, Clay and Daniel, 2000; Daniel and

Meitner, 2001; Scott and Canter, 1997; Wherrett, 2000). The ability of photographs to represent the dynamic multidimensionality of real landscapes in an adequate way has been doubted and criticized. However, despite limitations, colour photographs have been found to represent landscapes in a satisfactory manner when compared to preference rankings made in the field (Daniel, 2001 and references therein, Trent et al., 1987; Wherrett, 1998). The aim of this study was to search for a link between the map-based land cover data typical for monitoring programs, and perceived aesthetic quality as quantified through a photographybased preference study. To achieve this we used the results from a landscape preference study for Norwegian agricultural landscapes (Tveit, 2000), linked to a number of 1 km2 squares included in the 3Q-monitoring programme. Based on the landscape represented in the images, we calculated a number of map-based landscape indicators, aiming to see whether any of these indicators were correlated to preference rankings. 2. Methods 2.1. The 3Q-programme The monitoring programme involves the mapping and analysis of 1400 1 km × 1 km squares. The sample squares are located using the 3 km × 3 km grid already used for sampling in the Norwegian forest inventory. Where the grid point falls on land used for agricultural purposes (as defined by the Economic Map Series), a 1 km2 surrounding this point has been included in the sample. The programme follows a 5-year inventory cycle, with national coverage first available in 2003. Mapping is based primarily on interpretation of aerial photographs (scale 1:12,500, true colour). Land cover and land use within each square are digitised according to a hierarchical classification system; including a total of ca. 100 land types at the third and most detailed level. Among the more common land types in the area were cereal fields, meadows, built-up land and both deciduous and coniferous woodlands. In addition, linear elements and point objects are recorded, including both natural and cultural features. A number of indicators are calculated from the resulting maps, addressing the four main issues of interest defined by the Norwegian Ministries: landscape structure, biodiversity, cultural heritage and accessibility. For further details on the 3Q-monitoring programme, see Dramstad et al. (2002). 2.2. On-the-ground photography As a module linked to the 3Q-programme and research projects related to this programme, ground photography has been conducted on a 10% sub-sample of the 3Q-monitoring squares (20–30 squares annually). Around 15–25 photographs were taken by a professional photographer in each 1 km2 , the aim being to describe the landscape content and variation throughout the square. In addition to documenting the present appearance of the landscape, these photographs also provide a historic landscape archive for the future (Puchmann and Dramstad, 2003). To enable the same landscape section to be photographed in the future, the precise location from which each photograph was

467

W.E. Dramstad et al. / Landscape and Urban Planning 78 (2006) 465–474

Fig. 1. (a) Map of Norway, with circle surrounding Østfold and Akershus counties. (b) Example monitoring square illustrating the method of photography. Each arrow represents one photograph. Focal length and aperture are recorded next to each arrow. Since 1999, all positions are geo-referenced in the field through the use of GPS.

taken and the direction of the view were recorded on maps in the field and later digitised and stored in a GIS (see Fig. 1b). Details of the photographic equipment used were also recorded. These mapped positions were used as the starting point for mapping the area included in each image (see below). 2.3. Preference study In the preference study, 30 photographs were selected from 3Q-monitoring squares in Østfold and Akershus counties (see Fig. 1). The photographs were chosen to illustrate different degrees of openness in the landscape. Openness was not measured at this stage but pictures were chosen subjectively based on perceived openness. An open landscape is defined as a landscape with low vegetation allowing a clear view, as opposed to tall vegetation which obscures the view. Other aspects than openness, such as light, weather conditions and seasonal differences, were held constant as far as possible. An attempt was made to keep the pictures free from features believed to be particularly strong drivers of preference, such as water (Nasar and Li, 2004), and man-made features (Kaplan and Kaplan, 1989; Strumse, 1994). However, it was not possible to avoid all man made features, since almost all pictures had some man made features in the far distance. The preference study was conducted using commonly applied methods (as in, for example, Daniel and Boster, 1976, pp. 9–10; H¨agerh¨all, 2001; Herzog, 1984, 1987; Lynch and Gimblett, 1992). Using a projector, landscape slides were shown on a large screen in random order in a neutral room. Each slide was displayed for 10 s, after which the viewers were given 50 s to answer questions on a form. The assessment of preferences was conducted by asking participants to give a score to each landscape according to how much they liked the view (1 for least preferred and 5 for most preferred). A short break was allowed after every eight pictures to avoid fatigue effects. A total of 53 people from

Østfold and Akershus and 38 students from other parts of the country participated in this study (see Table 1). All students came from the Norwegian University of Life Sciences and, while their disciplines varied, they were all studying environmentally related subjects. All participants were invited according to random selection procedures. For the student group this was done by distributing invitations to every second mailbox, while for the public group every 50th entry in the local phonebook was mailed an invitation, also offering to cover travel costs. 2.4. Mapping the view By comparing a large display of each photograph with the maps showing the photographer’s location and angle of view, the boundaries of the area visible in each image were digitised using ArcViewTM (ESRI). In this paper, we refer to the area covered by each photograph as a viewshed. The presence of vegetated boundary-lines, telegraph poles, buildings, etc., aided the identification of the area represented in the image. An example of three images and their estimated viewsheds is shown in Fig. 2. Since, we did not have access to land cover data for the area outside the monitoring squares, all images that contained clearly visible area from outside the squares were excluded. This procedure left us with a total of 24 images. A selection of landscape metrics was calculated for each viewshed (Table 2), including Table 1 Information about the participants in the preferences study

Number of participants Gender (M/F) Average age Youngest/oldest participant

Locals

Students

53 33/20 50.8 9/76

38 18/20 24.6 20/44

468

W.E. Dramstad et al. / Landscape and Urban Planning 78 (2006) 465–474

Fig. 2. Using landscape elements such as telegraph poles and buildings, the area represented in the image was digitised and this ‘viewshed’ used for further analysis.

measures of both content (which land types were present) and spatial configuration (how patches of different land types are located in relation to one another). Landscape metrics included the total area of the viewshed, number of different land types present, number of land type patches present, total length of patch edges, area of open land types, grain size of the open landscape, landscape heterogeneity and land type diversity measured as Shannon’s diversity index (Magurran, 1988). These metrics were chosen because they are commonly implemented in various forms of landscape monitoring and are relatively simple to use and to interpret. The heterogeneity index (see Fjellstad et al., 2001) is less well known but is used in the 3Q-monitoring programme. The index is calculated by recording land type at a grid of points and then comparing every possible pair of points. The heterogeneity index is the proportion of points that are on different land types. The minimum heterogeneity value is zero, when all points fall on the same land type Table 2 The minimum, mean and maximum values for the different landscape metrics for all estimated viewsheds

Total area (m2 ) Number of land types Number of patches Heterogeneity index Shannon’s diversity index Open area (m2 ) Percent openness Grain size of open areas Length of edge (m)

Minimum

Mean

Maximum

3696 3 3 0.00 0.21 2318 32 0.05 621

31508 5.5 9 0.35 0.87 27432 81 0.40 2335

128861 9 24 0.73 1.65 126828 100 1.73 6580

(large-scale, homogeneous landscape), and the maximum value is one, when the points in every pair fall on different land types (small-scale landscape with a high degree of spatial division). Since the viewsheds were relatively small, a 10 m × 10 m grid of points was applied when measuring heterogeneity instead of the 100 m × 100 m grid used in the monitoring programme. The heterogeneity index is a measure of spatial division that is independent of the number of different land cover types in an area. As such it complements the Shannon’s diversity index, which is based on the relative areas of different land types. To measure the area of open land types and the grain size of the open landscape, it was necessary to recode all land types according to whether the land cover would enable a clear view (open) or would obscure the view (closed). All land types were transformed to a binary variable, zero representing open areas such as cereal fields and roads while one represented closed areas such as forest or built-up land. Grain size was calculated as the number of patches of open land types divided by the total area of open land. Since area was measured in metre square this produced a very small number and we multiplied by 1000 to make the index easier to grasp at a glance. This gives an index value, which for a given area, increases evenly with increasing number of patches. Using grain size in addition to total open area takes account of the fact that landscape elements such as a narrow hedgerow or a grass bank between two fields (i.e. that divide the landscape into more patches) may change the visual impression of a landscape, even though total area of open landscape may be almost identical. Spearman’s correlation coefficient was used to explore relationships between the landscape preference scores and the various spatial metrics.

W.E. Dramstad et al. / Landscape and Urban Planning 78 (2006) 465–474

3. Results 3.1. Landscape metrics Table 2 shows the degree of variation present in the viewsheds for the different landscape metrics. For comparison, the average heterogeneity index values calculated for the 1 km2

469

monitoring squares in this region (137 squares) is 0.49 (minimum 0.27, maximum 0.80), whilst the average Shannon’s diversity index is 1.62 (minimum 0.69, maximum 2.34). The viewsheds, having been chosen to avoid water and man-made objects, were thus slightly less heterogeneous and diverse than a typical landscape view from this region of Norway (Fig. 3).

Fig. 3. Photograph (a) received the lowest average preference score from the locals group and photograph (b) received the highest average preference score from this group. Photograph (c) received the lowest average preference score from the student group, while photo (d) received the highest average preference score from this group. Photograph (e) is an example of a viewshed containing water, while photograph (f) shows the most open viewshed. Photographs (g) and (h) were those over which there was greatest disagreement between students and locals; (g) was ranked number four for the students (i.e., fourth most preferred) and number 17 for the locals, (h) was ranked number 21 by the students and number 10 by the locals.

470

W.E. Dramstad et al. / Landscape and Urban Planning 78 (2006) 465–474

Fig. 4. Average preference score for each image for the two groups of participants, showing clear differences in the responses of students and locals to the images. The images are numbered in rank order from left to right according to increasing percent open area in the viewshed. Arrows point to images with water.

3.2. Preference scores A preference score value was calculated for each image based on the ranking (scale 1 to 5) given by the observers. Average preference scores for the two groups of participants were almost identical (students: 3.25 versus locals: 3.21). Ranges in scores were from 2.40 to 4.16 for students and from 2.57 to 3.62 for the local group, implying that the student group made more use of the full scale when giving preference scores. Distribution of preference scores for the 24 images is shown in Fig. 4. The student group and the locals group differed as to which image they ranked as highest preferred and lowest preferred (see Fig. 3a–d) and, overall, there was no significant correlation between the preference scores assigned by the two groups (Spearman’s rho = 0.358, p = 0.086). The most contested images – where the two groups showed the most disagreement – were photographs 8 (Fig. 3g) and 21 (Fig. 3h). Photograph 8 was ranked as number four (i.e. fourth most preferred) for the students and as number 17 for the locals. Photograph 21 was ranked number 21 by the students and number 10 by the locals. 3.3. Correlations Using the total dataset, we found a significant positive correlation between preferences and both the number of land types (Spearman’s rho: 0.677, p < 0.001) and the number of patches (Spearman’s rho: 0.623, p = 0.001) within the image area. We found no correlation in the total dataset between preferences and landscape openness or spatial heterogeneity, but found a significant correlation with land type diversity (Spearman’s rho: 0.407, p = 0.049) within the mapped image area (measured by the Shannon index). When the dataset was split into two groups, we found interesting differences between students and locals.

For the student group Shannon’s diversity index, the heterogeneity index, and the number of land type patches were all significantly positively correlated with preference score (Table 3). Percent open area was negatively correlated with preference score for this group. For the locals, total area and number of land types in the image were significantly correlated with preference score at the 0.01 significance level, whilst area of open land types and total length of edge were significant at a 0.05 level. Grain size of the open landscape was not found to be significantly correlated with preference scores for either group. In interpreting these correlations it is important to consider that there were also correlations between some of the different landscape metrics (Table 4). For example, the total area of the viewsheds was positively correlated with the total open area, percent open area, length of edge and number of patches and negatively correlated with grain size. There was no correlation, on the other hand, between total area of viewshed and heterogeneity, Shannon’s diversity index or number of land types. A comparison of preference scores for the five images containing the largest proportion of open area (>96% open area), showed that the student group (average score 2.70) had a significantly lower preference (one-tailed t-test, p < 0.001) for these large-scale open landscapes than the local group (average score 3.33). Although we excluded photographs in which water was clearly visible, water appeared on the list of land types present within the viewshed for seven of the 24 images (see Fig. 3e). The water could be seen in only one of the images (in the background). The five most preferred images in the dataset all contained water within their viewshed. The other two images containing water as a land type ranked as number 12 and 22. Images of viewsheds containing water were significantly more preferred than those without water (one-tailed t-test, p = 0.004) and this applied both for students (p = 0.006) and for locals (p = 0.026).

471

W.E. Dramstad et al. / Landscape and Urban Planning 78 (2006) 465–474 Table 3 Correlations between landscape metrics and average preference scores for the two groups of participants Landscape metric

Shannon’s diversity index Heterogeneity index Number of land types Number of patches Percent open area Total area Area of open land types Length of edge Grain size

Students

Locals

Spearman’s rho

Significance values

Spearman’s rho

Significance values

0.578 0.573 0.534 0.453 −0.452 −0.037 −0.093 0.154 0.327

0.003**

0.138 −0.083 0.583 0.506 0.246 0.532 0.510 0.481 −0.152

0.522 ns 0.701 ns 0.003** 0.012* 0.247 ns 0.007** 0.011* 0.017* 0.479 ns

0.003** 0.007** 0.026* 0.027* 0.864 ns 0.667 ns 0.474 ns 0.119 ns

Ns: not significant. * Significant at the 0.05 level. ** Significant at the 0.01 level.

4. Discussion In Norway, as in many other countries, population centres have grown up in those parts of the country with the best agricultural conditions. This means that agricultural landscapes are the “everyday-landscapes” for a large proportion of the population, and it is therefore important to be able to monitor how changes affect the visual appearance of these landscapes. A landscape rich in biodiversity and amenity values is not necessarily a by-product of current agricultural practice (Barnard, 2000; Green and Vos, 2001), and the public have come to realise that many landscapes are deteriorating (Council of Europe, 2000). In certain cases, the demand for these amenity byproducts is increasing to the extent that food is considered to be the by-product (see discussion by Hellerstein et al., 2002). The current lack of scientifically based indicators for measuring changes in the visual landscape, hampers the inclusion of this topic in monitoring programmes such as the Norwegian 3Q-programme. The exclusion of landscape aesthetic information from monitoring and management may lead to less soundly based decisionmaking than desirable. When considering the costs and benefits of landscape changes, for example, economic interests that are easily measured may receive more attention than visual quality, even though there is an acceptance that this “public good” also has an economic value in terms of, for example, human health and tourism income. If alternative development scenarios could be assessed using simple quantitative methods, the probability that landscape aesthetics would be taken into account in decision-making could be increased. Methods to capture landscape values are therefore needed so that these values can be integrated effectively with other kinds of data in designing, planning and managing landscapes (Ndubisi, 2002). Nassauer (1986) challenges us to work towards an understanding of how elements fit together in the visual landscape, so that change can be planned and managed in a way that leads to desirable future landscapes, instead of focusing too strongly on the past. This is also the message put forward by Green and Vos (2001), underlining the need to conceive, design, create and maintain new landscapes fit for the social, economic and environmental needs of the twenty-first

century. To achieve this, we need to be able to quantify and monitor landscape change, including changes in visual appearance. Landscape preference research points to certain general “rules of thumb” regarding features that influence preference scores (Kaplan and Kaplan, 1989; Nassauer, 1995; Ndubisi, 2002; Zube, 1987). Water, for instance, has been shown by many studies to be positively correlated with preference scores (see, for example, Herzog and Bosley, 1992; Herzog and Barnes, 1999; Kaltenborn and Bjerke, 2002; Purcell et al., 1994). In our study, water itself was visible in only one of the images. However, information from maps showed that water was present in seven of the viewsheds, and the results from the preference study revealed that these images received significantly higher preference scores than images of areas without water. In these images, there are clear vegetation belts that indicate presence of waterways meandering across the landscape, a landscape feature probably recognised by most people. Our results suggest that these patterns of vegetation, possibly combined with topographic variation, caused a general positive response (both for students and locals). The underlying reason for this response may be an evolutionary adaptation, i.e. that people read the landscape and interpret cues of the presence of water, as proposed in the information-processing theory of Kaplan and Kaplan (1989). However, the existence of such a deep, sub-conscious reason for a preference, presumably applying to all people everywhere, is extremely difficult to test. Whilst we cannot conclude from this study whether it is the vegetation that causes positive reactions or the suggestion of the presence of water, the result is nevertheless that waterways (with their associated vegetation and topography) are a strong predictor of aesthetic preference. For the group of students participating in this study, we found significant positive correlations between landscape preferences and landscape heterogeneity and diversity. Such relationships have been mentioned by other authors previously (Hunziker, 1995; Kaplan and Kaplan, 1989; Piorr, 2003; Zube, 1987). It was interesting to note, however, that these relationships did not apply for the group of locals. The photographs in Fig. 3 clearly illustrate the differences between the two groups. Neither the heterogeneity index, nor Shannon’s diversity index were significantly correlated with preference scores for the locals. The

472

−0.570 0.004** −0.201 0.347 0.040* 0.852 0.298 0.157 −0.650 0.001** *

0.498 0.013 0.131 0.541 0.312 0.138 0.515 0.010* 0.921 0.000** −0.614 0.001**

Total area

0.930 0.000** 0.707 0.000** −0.050 0.818 0.192 0.368 0.464 0.022* 0.836 0.000** −0.723 0.000**

Total open area

Significant at the 0.05 level. Significant at the 0.01 level. *

**

Grain size

Total length edge

No. of patches

No. of land types

Shannon’s diversity index

Percent open area

Total area

Total open area

Total closed area

Rho Significance value Rho Significance value Rho Significance value Rho Significance value Rho Significance value Rho Significance value Rho Significance value Rho Significance value Rho Significance value

0.536 0.007** −0.361 0.083 −0.274 0.195 −0.661 0.000** 0.801 0.000** 0.233 0.274 0.183 0.393 −0.045 0.834 0.482 0.017*

−0.049 0.821 0.171 0.424 −0.695 0.000** 0.724 0.000** 0.377 0.070 0.300 0.155 0.312 0.138 0.190 0.375

Total closed area Hetero-geneity index

Table 4 Correlations between the landscape metrics, showing correlation coefficients (Spearman’s rho) and significance values

Percent open area

0.553 0.005** 0.492 0.015* 0.349 0.095 0.366 0.079

Shannon diversity index

0.757 0.000** 0.462 0.023* 0.346 0.097

No. of land types

0.748 0.000** 0.197 0.357

No. of patches

−0.372 0.073

Total length edge

W.E. Dramstad et al. / Landscape and Urban Planning 78 (2006) 465–474

correlation between the locals’ landscape preferences and total length of edge, which could also be considered an aspect of landscape diversity, is probably attributable to the strong preference of locals for a large view and the fact that total length of edge is strongly correlated to total viewshed area. One interpretation of the difference between students and locals may be that a few of the students had attended landscape ecology lectures and were influenced by what they had learned about the benefits of diverse and heterogeneous landscapes for biodiversity. Another possible interpretation is that preference scores reflected different landscape familiarity for the two groups (Kaplan and Kaplan, 1989; and see, for example, discussion by Armstrong, 2002; Bourassa, 1991; Kjølen, 1998). The significantly higher preference by locals for the five most open landscapes compared with students may be due to the fact that the local group live in landscapes that are amongst the most open, large-scale, intensively managed agricultural landscapes in Norway, whereas the students came from different parts of the country where such landscapes are unusual. This also fits with the finding that locals preferred a large view (total viewshed area) whilst student preferences were not significantly influenced by size of view. More research is needed to investigate the role of familiarity in shaping visual landscape preferences and to explore other underlying reasons for differences in landscape preferences such as age and social background. Nevertheless, the results of this study suggest that care is needed to interpret indicators of landscape change within an appropriate regional context. Regions defined by landscape type are more likely to provide a necessary and meaningful context for the interpretation of landscape indicator values than administrative units. In addition, our study shows that students who may go on to careers in landscape planning and management, may not share the aesthetic preferences of local people. This is an important consideration when working to fulfil the requirements of the European Landscape Convention, where a “Landscape quality objective” for a specific landscape, is defined as “the formulation by the competent public authorities of the aspirations of the public with regard to the landscape features of their surroundings” (Article 1: Council of Europe, 2000). Clearly, a more widespread use of methods for public participation in landscape planning will be required if such landscape quality objectives are to be adequately formulated. The 3Q-monitoring squares used in this study contained both low preference and high preference views in close proximity. Since single photographs cannot represent an entire monitoring square, the viewshed approach used in this study was necessary to establish connections between map-derived indicators and landscape preferences. Due to the complexity of underlying reasons for human landscape preferences, we cannot expect to be able to accurately predict visual preferences for larger mapped areas. However, having some general guidelines of the aspects of landscape content and spatial configuration that are important in determining preferences will improve our ability to predict whether landscape changes have positive or negative effects on the appearance of the landscape, and thus improve interpretation of regional indicator values. While we fully appreciate the need to apply such indicators with care (Bell and Morse, 1999; Cale

W.E. Dramstad et al. / Landscape and Urban Planning 78 (2006) 465–474

and Hobbs, 1994; Gustafson, 1998; Leit˜ao and Ahern, 2002; Turner, 1989), it is no longer possible to deny their need. Further, there are few methods available for the objective description of the visual landscape. Highly subjective and expert approaches do not seem acceptable to policy-makers or the general public and make it difficult to compare landscapes or quantify changes over time. This is particularly important if, as our study suggests, the trained experts have different landscape preferences than local people. The present situation that excludes visual aspects of a landscape from consideration due to lack of methods, is not acceptable either. In this respect, we agree with the statement that “. . . indicators can distort priorities—those things which are being measured and reported are viewed as more important, while things which are less readily measured are omitted and given lower priority” (Detr, 2000, p. 6). Landscape aesthetics are among the issues less readily measured. Indicators have a large number of advantages, and have come to be highly demanded, for example, by politicians (OECD, 2001). We would therefore encourage the continued research effort in developing indicators also for these more challenging topics, such as the visual landscape. 5. Conclusion By establishing a link between the birds-eye view of remote sensing data and the on-the-ground perspective of landscape photographs, this study identified several aspects of landscape content and spatial configuration that are related to people’s landscape preferences and that may therefore be suitable as indicators for the visual landscape. In particular, presence of water and number of land types appear to be useful predictive variables. However, the finding that different groups of people (students and locals) prefer different types of landscape underlines the need for care when interpreting indicator values. There are many possible underlying reasons for people’s preferences for landscapes. Amongst other things, it may be that an education in environmental subjects leads to a more complex interpretation of landscape views, incorporating other values than pure visual aesthetics (for example, an assessment of conditions for biological diversity) or it may be that preferences are affected by the degree of familiarity with the landscape in question. Whatever the causes of differences, this study suggests a need for wide public participation in landscape planning decisions if visual aesthetics are to be adequately addressed. We also underline the importance of interpreting landscape indicator values in an appropriate regional context, with a focus on landscape types rather than administrative units. Acknowledgements This study was supported by the Norwegian Research Council. The authors would like to thank Oskar Puschmann, Norwegian Institute of Land Inventory (NIJOS) for all his help with the photos, as well as everyone who participated in the preference study. We would also like to thank two anonymous reviewers for valuable comments on an earlier version of the manuscript.

473

References Armstrong, H., 2002. Myth & the landscape. In: Gordon, E., (Ed.), Spatial experience: Media and the production of place. Special edition of Spectator, USC Journal of Film and Television Criticism, pp. 12– 25. Barnard, C.H., 2000. Urbanization affects a large share of farmland. Rural Conditions Trends 10, 57–63. Bell, S., Morse, S., 1999. Sustainability Indicators. Measuring the Immeasurable. Earthscan Publications, London. Bourassa, S.C., 1991. The Aesthetics of Landscape. Belhaven Press, London. Bruns, D., Green, B.H., 2001. Identifying threatened, valued landscapes. In: Green, B., Vos, W. (Eds.), Threatened Landscapes. Conserving Cultural Environments. Spon Press, London, pp. 119–127. Cale, P.G., Hobbs, R.J., 1994. Landscape heterogeneity indices: problems of scale and applicability, with particular reference to animal habitat description. Pac. Conserv. Biol. 1, 183–193. Clay, G.R., Daniel, T.C., 2000. Scenic landscape assessment: the effects of land management jurisdiction on public perception of scenic beauty. Landscape Urban Plan. 49, 1–13. Council of Europe, 2000. European Landscape Convention and Explanatory Report. T-LAND 6. Strasbourg. Daniel, T.C., Boster, R.S., 1976. Measuring Landscape Aesthetics: The Scenic Beauty Estimation Method. Forest service, US Department of Agriculture. Daniel, T.C., 2001. Whither scenic beauty? Visual landscape quality assessment in the 21st century. Landscape Urban Plan. 54, 267–281. Daniel, T.C., Meitner, M.M., 2001. Representational validity of landscape visualizations: the effects of graphical realism on perceived scenic beauty of forest vistas. J. Environ. Psychol. 21, 61–72. Dearden, P., 1987. Consensus and a theoretical framework for landscape evaluation. J. Environ. Manag. 34, 267–278. Defra, 2004. Quality of Life Counts. Environment Protection Statistics and Information Management Division, Department for Environment, Food and Rural Affairs, London, UK. Detr, 2000. Local Quality of Life Counts. Department for Environment, Transport and the Regions, London, UK. Dramstad, W.E., Fjellstad, W.J., Strand, G.H., Mathiesen, H.F., Engan, G., Stokland, J.N., 2002. Development and implementation of the Norwegian monitoring programme for agricultural landscapes. J. Environ. Manag. 64, 49–63. Dramstad, W.E., Sogge, C., 2003. Agricultural impacts on landscapes: developing indicators for policy analysis. NIJOS Report 7/03. NIJOS/Landscape ˚ http://www.nijos.no. Europe, As. European Environment Agency, 2004. EEA-signals 2004. European Environment Agency, Copenhagen. Eurostat, 2003. The Lucas survey. European statisticians monitor territory. Updated edition—June 2003. Working papers and studies Theme 5 Agriculture and fisheries. Eurostat, Luxembourg. Fjellstad, W.J., Dramstad, W.E., Strand, G.-H., Fry, G.L.A., 2001. Heterogeneity as a measure of spatial pattern for monitoring agricultural landscapes. Norway J. Geogr. 55, 71–76. Green, B., Vos, W. (Eds.), 2001. Threatened Landscapes. Conserving Cultural Environments. Spon Press, London. Gustafson, E.J., 1998. Quantifying landscape spatial pattern: what is the state of the art? Ecosystem 1, 143–156. H¨agerh¨all, C.M., 2001. Concensus in landscape preference judgements. J. Environ. Psychol. 21, 83–92. Hartig, T., Evans, G.W., Jamner, L.D., Davis, D.S., G¨arling, T., 2003. Tracking restoration in natural and urban field settings. J. Environ. Psychol. 23, 109–123. Hellerstein, D., Nickerson, C., Cooper, J., Feather, D., Gadsby, D., Mullarkey, D., Tegene, A., Barnard, C., 2002. Farmland protection: the role of public preferences for rural amenities. Agricultural Economic Report No. 815, Economic Research Service, USDA, Washington, DC. Herzog, T.R., 1984. A cognitive analysis of preference for field-and-forest environments. Landscape Res. 9, 10–16. Herzog, T.R., 1987. A cognitive analysis of preference for natural environments: mountains, canyons and deserts. Landscape J. 6, 140–152.

474

W.E. Dramstad et al. / Landscape and Urban Planning 78 (2006) 465–474

Herzog, T.R., Barnes, G.J., 1999. Tranquillity and preference revisited. J. Environ. Psychol. 19, 171–181. Herzog, T.R., Bosley, P.J., 1992. Tranquillity and preference as affective qualities of natural environments. J. Environ. Psychol. 12, 115–127. Hunziker, M., 1995. The spontaneous reafforestation in abandoned agricultural lands: perception and aesthetic assessment by locals and tourists. Landscape Urban Plan. 31, 399–410. Kaltenborn, B.P., Bjerke, T., 2002. Associations between environmental value orientations and landscape preferences. Landscape Urban Plan. 59, 1–11. Kaplan, R., Kaplan, S., 1989. The Experience of Nature: a Psychological Perspective. Cambridge University Press, Cambridge, UK. Kaplan, R., Kaplan, S., Ryan, R.L., 1998. With People in Mind. Design and Management of Everyday Nature. Island Press, Washington, DC. Kjølen, A., 1998. Stedsidentitet. Et forsøk p˚a a˚ gripe et uhandgripelig begrep. Landskap, pp. 34–38. Leit˜ao, A.B., Ahern, J., 2002. Applying landscape ecological concepts and metrics in sustainable landscape planning. Landscape Urban Plan. 59, 65–93. Lynch, J.A., Gimblett, R.H., 1992. Perceptual values in the cultural landscape: a computer model for assessing and mapping perceived mystery in rural environments. J. Comp.; Environ. Urban Syst. 16, 453–471. Magurran, A.E., 1988. Ecological Diversity and its Measurement. Croom Helm, London. Meinig, D.W., 1976. The Beholding Eye. Ten versions of the same scene. Landscape Architect. 66, 47–54. Nasar, J.L., Li, M.H., 2004. Landscape mirror: the attractiveness of reflecting water. Landscape Urban Plan. 66, 233–238. Nassauer, J.I., 1986. Caring for the countryside. A Guide to seeing and maintaining rural landscape quality. U.S. Department of Agriculture Soil Conservation Service, Minnesota. Nassauer, J.I., 1995. Culture and changing landscape structure. Landscape Ecol. 10, 229–238. Ndubisi, F., 2002. Ecological planning: a Historical and Comparative Synthesis. John Hopkins University Press, Baltimore. OECD, 2000. Environmental indicators for agriculture: methods and results—the stocktaking report. Landscape. Document 88332. Organisation for Economic Co-operation and Development, Paris. OECD, 2001. Environmental indicators for agriculture: methods and results. Organisation for Economic Co-operation and Development, Paris. OECD, 2004. Agriculture and the Environment: lessons learned from a decade of OECD work. Organisation for Economic Cooperation and Development, Paris. Parsons, R., Daniel, T.C., 2002. Good looking: in defense of scenic landscape aesthetics. Landscape Urban Plan. 60, 43–56.

Piorr, H.P., 2003. Environmental policy, agri-environmental indicators and landscape indicators. Agric. Ecosyst. Environ. 98, 17–33. Puchmann, O., Dramstad, W.E., 2003. Documenting landscape change through fixed angle photography. In: Dramstad, W.E., Sogge, C. (Eds.), Agricultural Impacts on Landscapes: Developing Indicators for Policy Analysis. ˚ pp. 251–259. NIJOS/Landscape Europe, As, Purcell, A.T., Lamb, R.J., Peron, E.M., Falchero, S., 1994. Preference or preferences for landscape? J. Environ. Psychol. 14, 195–209. Scott, M.J., Canter, D.V., 1997. Picture or place? A multiple sorting study of landscape. J. Environ. Psychol. 17, 263–281. Strumse, E., 1994. Environmental attributes and the prediction of visual preferences for agrarian landscapes in Western Norway. J. Environ. Psychol. 14, 293–303. Tahvanainen, L., Ihalainen, M., Hietala-Koivu, R., Kolehmainen, O., Tyrvainen, L., Nousiainen, I., Helenius, J., 2002. Measures of the EU AgriEnvironmental Protection Scheme (GAEPS) and their impacts on the visual acceptability of Finnish agricultural landscapes. J. Environ. Manag. 66, 213–227. Terkenli, T.S., 2001. Towards a theory of the landscape: the Aegean landscape as a cultural image. Landscape Urban Plan. 57, 197–208. Turner, M.G., 1989. Landscape ecology: the effect of pattern on process. Ann. Rev. Ecol. Syst. 20, 171–197. Tveit, M.S., 2000. Landscape preferences. Who sees what in the agricultural landscape. Master Thesis. Dept. of Biology and Nature Conservation. Agri˚ cultural University of Norway, As. Trent, R.B., Neumann, E., Kvashny, A., 1987. Presentation mode and question format artifacts in visual assessment research. Landscape Urban Plan. 14, 225–235. Tress, B., Tress, G., Decamps, H., d’Hauteserre, A.M., 2001. Bridging human and natural sciences in landscape research. Landscape Urban Plan. 57, 137–141. Wascher, D.M., (Ed.) 2000. The face of Europe. Policy perspectives for European landscapes. ECNC Technical Report Series. European Centre for Nature Conservation, Tilburg. Wherrett, J.R. 1998. Natural Landscape Scenic Preference: Techniques for Evaluation and Simulation. PhD Thesis. Aberdeen and Robert Gordon University, Aberdeen. Wherrett, J.R., 2000. Creating landscape preference models using internet survey techniques. Landscape Res. 25, 79–96. Zube, E., 1984. Themes in landscape assessment theory. Landscape J. 3, 104–110. Zube, E., 1987. Perceived land use patterns and landscape values. Landscape Ecol. 1, 37–45.