A prototypical tool for normative landscape scenario development and the analysis of actors’ policy preferences

Landscape and Urban Planning 137 (2015) 40–53

Contents lists available at ScienceDirect

Landscape and Urban Planning journal homepage: www.elsevier.com/locate/landurbplan

Research Paper

A prototypical tool for normative landscape scenario development and the analysis of actors’ policy preferences Enrico Celio ∗ , Michel Ott, Elina Sirén, Adrienne Grêt-Regamey ETH Zurich, Institute for Landscape and Spatial Development, Planning of Landscape and Urban Systems (PLUS), Stefano-Franscini-Platz 5, 8093, Zürich, Switzerland

h i g h l i g h t s • • • • •

We developed an interactive tool to support normative landscape scenario development. The tool combines a land use model and 3D visualizations. Participants’ prefer extensive agricultural and a limitation of urban sprawling. We show participants’ decision pathways deliberating about policy objectives. These pathways are connected to participants’ favorite landscape visualizations.

a r t i c l e

i n f o

Article history: Received 28 February 2014 Received in revised form 22 December 2014 Accepted 29 December 2014 Keywords: Land-use model 3D visualization Normative landscape scenario Landscape development

a b s t r a c t In order to find negotiable solutions for spatial planning and landscape development, scenario analysis would profit from knowledge about the trade-offs people make in their policy objectives given a certain landscape scenario. To facilitate the development of normative landscape scenarios, the output of a landuse decision model was visualized in 3D, establishing a relationship between policy influence factors (as model drivers) and the 3D visualizations. A survey tool was developed for eliciting the preferences for these 3D visualizations and the respective policy measures. In this contribution, we present the survey tool tested in a case study from a pre-Alpine area in central Switzerland. Results show participants’ preferences for extensive agricultural land use and limitation of urban sprawl. The trade-off analysis shows how participants compare and choose among different policy objectives to reach a possible solution given their choice of a future landscape. The trade-off pathway most often observed reinforces the chosen visualization in terms of policy measures. In the discussion, we consider the advantages and disadvantages of this survey tool and conclude that it is an effective way to support the development of normative landscape scenarios in participatory planning processes. © 2015 Elsevier B.V. All rights reserved.

1. Introduction Abandonment and rural exodus, as well as agricultural intensification and urbanization are leading to rapid changes in cultural landscapes across the world (Plieninger, van der Horst, Schleyer, & Bieling, 2014). These changes are driven by various policies, changing biophysical conditions, and local actor characteristics that have different impacts on the landscape (e.g., visual, functional, or cultural). Inducing landscape changes without consciously making choices among the policies that drive them can lead to socially

∗ Corresponding author. Tel.: +41 44 633 40 64. E-mail addresses: [email protected] (E. Celio), [email protected] (M. Ott), [email protected] (E. Sirén), [email protected] (A. Grêt-Regamey). http://dx.doi.org/10.1016/j.landurbplan.2014.12.013 0169-2046/© 2015 Elsevier B.V. All rights reserved.

undesirable and potentially ecologically unfriendly outcomes (cf. van Zanten et al., 2014 for agricultural intensification; cf. Hamidi & Ewing, 2014 for urban sprawl). To effectively form conceptions of an uncertain future, scenario studies can help provide insights into the preferences and decision-making of stakeholders. Research on scenario techniques has made substantial contributions in the form of structures, typologies, and methodological frameworks. Given a coherent and consistent set of assumptions about driving forces and relationships, scenarios are seen as simplified descriptions of how the future may develop (cf. Walz et al., 2013) or as description of a “hypothetical future state of a system” that also provides information concerning how to reach a given state (cf. Scholz & Tietje, 2002). Descriptive scenarios explore possible futures, while normative scenarios describe probable or preferable futures (van Notten, Rotmans, van Asselt, & Rothman,

E. Celio et al. / Landscape and Urban Planning 137 (2015) 40–53

2003). Börjeson, Höjer, Dreborg, Ekvall, and Finnveden (2006) suggested three slightly different scenario types: (a) predictive scenarios (“What will happen?”) including forecasts and “what-ifscenarios”; (b) explorative scenarios aiming at exploring possible outcomes as a reaction to external factors or behaviors; and (c) normative scenarios. According to Börjeson et al. (2006), normative scenarios focus on the current situation (“How can a specific target be reached?”). In this paper, we understand “normative landscape scenarios” as plausible future landscape developments that should be achieved (Nassauer & Corry, 2004). Furthermore, “normative scenario methods” describe “pathways to desired future outcomes or visions” (Rounsevell et al., 2012). In the process outlined by Nassauer and Corry (2004), stakeholders provide their knowledge and assess the (preliminary) results, while process moderators (e.g., scientists) facilitate comprehensible data. With the help of a prototypical tool, this paper examines how the development of normative landscape scenarios can be initiated. As stakeholder perspectives are inherently included in normative scenarios, these scenarios might be more relevant for policy decision-making and could also be helpful for elaborating on landscape plans in a participatory manner. Furthermore, if landscape development requires defining decisions that should be made in light of a preferred future development, exploring the feasibility of alternative futures may be just as important as determining their likelihood. Thus, scenario studies that show future landscape development and provide information related to societal and ecological implications are typical methods for supporting spatial decision-making (cf. Ervin, 2001; Verburg, Rounsevell, & Veldkamp, 2006; e.g. Rounsevell et al., 2006; Walz et al., 2013). To be valuable in planning, scenarios need to be drilled down to a suite of measures aligned on a timeline. With forecasting methods, we for example develop trend explorations from the past or consistent scenarios from a current starting point. In contrast, backcasting methods are explicitly normative (Robinson, 1982) and take a future state as given for the development of milestones up to that (visionary) state (Haslauer, Biberacher, & Blaschke, 2012). Some authors argue that forecasting and backcasting approaches should be used in combination (Kok, van Vliet, Bärlund, Dubel, & Sendzimir, 2011; van Berkel & Verburg, 2012). The participants’ choices with respect to the presented prototypical tool are normative and thus indicate a desired future. Therefore, they could potentially be used as future visionary states in a backcasting exercise. By including actors in a planning process (potentially using foreand back-casting), we seek to balance the stakes regarding the use of space. The actors’ way of decision-making – i.e., their rationality when participating in such a process – is usually a black box. However, this information could help in elaborating suitable compromises. In the presented choice situations using the presented tool, each participant follows her mindset, and thus monitoring these choice situations sheds light into this process. As a result, the question regarding whether participants are using cost-benefit deliberations or if they are driven by, e.g., emotional aspects, may be explored. In order to support the development and better understanding of normative scenarios, 3D landscape visualizations have been used by different authors (cf. Nassauer & Corry, 2004) as these visualizations represent a model of reality that helps in the understanding of the processes in landscape change (cf. Wissen Hayek, 2011). 3D visualizations were used in several case studies to assess the acceptance of projects or the preference for particular ˜ ˜ ˜ & Ruiz-Aviles, landscapes (Arriaza, Canas-Ortega, Canas-Madue no, 2004; Home, Bauer, & Hunziker, 2010; Ives & Kendal, 2013; Junge, Lindemann-Matthies, Hunziker, & Schüpbach, 2011; LindemannMatthies, Briegel, Schüpbach, & Junge, 2010). Such studies have demonstrated that 3D visualizations are particularly helpful for

41

sharing outputs of land-use models with decision makers and stakeholders (Koomen, Rietveld, & de Nijs, 2008 citing Beurden, Borsboom-van, Lammeren, Hoogerwerf, & Bouwman, 2006; Lloret, Rodríguez, Omtzigt, Koomen, & de Blois, 2008). Bateman, Day, Jones, and Jude (2009) found that virtual reality visualizations may allow participants to better evaluate their willingness to pay for landscape changes (e.g., decisions regarding the size of the freshwater nature reserve area). Furthermore, 3D visualizations may possibly help to understand the landscape scale as a functional unit (Lloret et al., 2008; Wissen Hayek, 2011). If 3D visualizations are based on the outputs of land-use models, they reflect a validated rationale for possible developments, which is valuable for communication and credibility. Landscape visualizations created through such a modelingvisualization process reflect the impact of implemented land-use drivers on the landscape. Different policy objectives potentially lead to different visual representations of the landscape. Hence, observing actors’ normative choices for policy objectives reveals trade-offs between the objectives of the policies and provides a better understanding of the decision-making processes in landscape development. Even though scenario analysis and preference assessments are usually the starting point for planning processes, the question of how stakeholders compare and choose among policy objectives is crucial to reach a suitable compromise. Here, we operationalize the link between landscape development and policy objectives by policy variables with discrete levels driving a landuse model. Similar to Gregory and Wellman (2001), participants are asked to state their preferences in a sequence of tasks. Here, visual representations of landscapes are the starting point, and the selection of a favorite visual representation of the landscape restricts participants’ possibilities later on in the survey; in contrast, in the Gregory and Wellman (2001) study, participants were asked to choose among policy alternatives in a first step. In the following, we present an interactive survey tool based on a land-use modeling approach using Bayesian networks (BN) and thereon built 3D visualizations to elicit normative landscape scenarios. Landscapes and policy measures are inherently connected with each other through the land-use modeling framework in which policy measures are drivers of changes in land use. Starting with the future landscape choices made by participants, trade-offs can be identified between the policy measures required to reach those desired futures. We illustrate the survey tool in a case study in central Switzerland showing how the prototypical tool allows preferences to be clarified for landscape development and facilitates the identification of policy measures required to attain desirable future outcomes. Referring to Nassauer and Corry (2004), this study thus provides a tool for addressing following the question: “How should the landscape change?” Furthermore, the tool is used to more precisely demonstrate how to find answers to the following sub-questions: “What are plausible goals for future landscape policy?” and “What characteristics of landcover [sic] [. . .] help to achieve those goals?”

2. Method 2.1. Study area The study area encompasses the Kleine Emme catchment (481 km2 ) in Switzerland. Elevation in this mountainous area ranges from approximately 440 m to 2350 m a.s.l. This region comprises the urbanized area of Lucerne and the rural area of the UNESCO Biosphere Reserve Entlebuch and has a total of approximately 50,000 inhabitants. The visualizations were showing the village of Schuepfheim (about 4000 inhabitants), which is located halfway between Lucerne and the end of the valley. In the

42

E. Celio et al. / Landscape and Urban Planning 137 (2015) 40–53

visualizations the center of the village was depicted in a distance of about 1 km from a viewpoint north of the village. The specific situation of this village may be seen as representative for pre-Alpine areas. Twenty-three percent of the jobs in this municipality are found in the primary sectors (agriculture and forestry). For a point of comparison, the percentage of jobs in Switzerland in the primary sector is 4.3% (LUSTAT, 2013; BFS, 2008). Besides the general importance of the agricultural sector in this location, discussions on (agricultural) land use versus residential development seemed indispensable as the municipality of Schuepfheim plans to enlarge the settlement area outside the settlement core area despite the existence of unbuilt areas in the core area (cf. Appendix A for more detailed information). Hence, for finding sustainable compromises between (agricultural) land use, acceptable population density, and the preferred (e.g., aesthetic or social) quality of the settlement areas, tools combining visualizations, and policy instruments may help to raise awareness for the land use trade-offs and lead to more conscious decision-making processes. 2.2. Prototypical survey tool for assessing landscape and policy preferences The prototypical survey tool developed for this study links a land-use model with 3D visualizations as visualizations help to understand landscape change. A key element of the survey design was to investigate participants’ preferences for and trade-offs between policy objectives to reach desired landscape developments in the future. The survey is composed of the following three main elements: (1) a land-use model providing a rationale for the influence of policy mixes on landscape scenarios; (2) 3D visual representation of the landscape; and (3) a web browser-based tool to assess how people make choices in policy mixes for reaching a chosen landscape. The combination of these three elements provides the material to monitor the decision-making processes participants engage in when considering different visual representations of the landscape, choosing between policy mixes resulting in certain landscape scenarios, and a combination of visual and policy stimuli. This monitoring may be the foundation for the transparent formulation of normative landscape scenarios. In the following sections, the three elements constituting the prototypical tool for normative landscape scenario development are presented in more detail. 2.2.1. Land-use model linking policy mixes and landscape scenarios The Bayesian network-based land-use decision modeling approach BLUMAP (Celio, Koellner, & Grêt-Regamey, 2014) was used to establish cause–effect relationships between policy variables and land-use change and thus to visual representations of the landscape. Hence, we could define which policy mixes result in different landscape scenarios (and vice versa). More specifically, agricultural, forest, and settlement land change were simulated for a 20-year period based on actor intentions, policy drivers, and biophysical characteristics. Modeling efforts were followed by creating visualizations of the landscapes from the land-use model outputs. BLUMAP models land-use decisions using the following three main types of land use: agriculture, forest, and settlement. The main characteristics of BLUMAP are as follows: (1) the participatory expert-based set-up and parameterization process of three Bayesian networks (Kjaerulff & Madsen, 2008; Marcot, Steventon, Sutherland, & McCann, 2006); (2) the inclusion of spatially-explicit questionnaire information about the characteristics of local actors; (3) the integration of policy measures, local actor characteristics and biophysical influences; and (4) the probabilistic simulation of the different land-use categories. The inputs and outputs of the model include three intensity levels of agricultural cultivation

Fig. 1. Set-up of 3D visualizations and questionnaire structure.

depicting extensive parcels (yield<30 dt dry matter/ha), mediumintensive parcels (30–100 dt dry matter/ha), and intensive parcels (>100 dt dry matter/ha); undeveloped and built settlement area; and coniferous and deciduous forest in a raster-based format (25 m). Moorland and unproductive areas such as rocks are held constant as moorland is strictly protected by federal regulations and local actors have low interest in unproductive areas. Landuse dynamics are simulated by transferring land-use probabilities iteratively into a new time step using different decision periods (recalculations: settlement after 2.5 years, agricultural after 5 years, forest after 25 years). Further details concerning this modeling approach may be found in Appendix B, Celio et al. (2012), and Celio et al. (2014). To generate plausible descriptive land-use scenarios, as a first step we used the full spectrum of policy mixes and generated a set of land-use scenarios (see step I in Fig. 1). We thus combined each policy variable state with all other states. Including six policy variables (general direct payments, ecological direct payments, opening of agricultural markets, Federal and Cantonal spatial planning policy, municipal settlement enlargement, location of possible new municipal settlement enlargement), this resulted in 432 model runs and therefore in 432 land-use scenarios. “Location of new settlement” did not significantly influence land-use change for the year 2030, so this variable was later eliminated. Thus, here we use “policy mix” to denote a consistent policy-level combination of these five different policies. In a second step, we classified the scenarios according to their land-use patterns, thereby reducing the number of scenarios to group similar outputs together and to be able to distinguish between ten groups of land-use patterns. Land-use patterns in one group were very similar and could be achieved through a variable set of policy mixes. In other words, for each land-use pattern group, a consistent mix of policy-levels served as the foundation for the visual representations of the landscape scenario. Further information regarding the analysis of the land-use scenarios may be found in Appendix B. 2.2.2. 3D visual representation of the landscape As we transformed 2D data into 3D visualizations, the level of detail had to be defined (Beurden et al., 2006). For the present application, it was important to find a compromise between landscape aesthetics and the functional information originating from BLUMAP, as participants should be able to read the landscape and connect emotionally to the landscape aesthetics (cf. Kaplan &

E. Celio et al. / Landscape and Urban Planning 137 (2015) 40–53

Kaplan, 1989). 3D landscape visualizations should combine aesthetics (“looks like”) and function (“acts like”) (cf. Ervin, 2001). In discussions with the municipal council, we chose a familiar viewpoint in the region (cf. Schindler, 2005) and removed elements that might influence the perceived aesthetic beauty of the landscape in order to prevent bias toward individual 3D visualizations not related to landscape change (e.g., animals on the meadows; cf. Manyoky et al., 2012). The role of the foreground, middle ground, and background was defined in the visualizations: in the foreground we primarily displayed change in agriculture, in the middle ground change in settlement area, and in the background change in forest. The level of detail was adapted to this scheme so that the changing landscape elements were shown at a lower level of detail in the middle ground and background. Given our ignorance about future building styles, we depicted new houses in the middle ground as abstract cubes. We created the 3D visualizations from the land-use model output in two major steps: (1) a preliminary preparation and analysis of the output data (see I., II., and III. in Fig. 1); and (2) the actual visualization, which was elaborated using a game engine (see IV. in Fig. 1). Further information regarding the preparation of the 3D visualizations and the techniques used may be found in Appendix B. 2.2.3. Web browser-based tool for communicating scenarios and conducting surveys The survey design was elaborated in an iterative process including researchers, a computer scientist, a web designer, and local councils. This resulted in the following in a four-step survey for assessing preferences for visual representations of the landscape scenario and policy objectives: Step 1. Pairs of 3D visualizations scenarios were compared (Fig. 1, step 1 of the questionnaire). For 10 pairs of scenarios, participants had to choose their favorite visual representation of the landscape. We made sure that the chosen visualizations covered the whole spectrum of possible land-use changes modeled with BLUMAP. This step aimed at ascertaining participants’ preferences for a visual representation of the landscape. Step 2. Policy measures were introduced (Fig. 2). First, participants had to choose an unrestricted favorite policy mix (i.e., all levels of a policy could be chosen). Second, participants chose a restricted policy mix, in which only those policylevels that corresponded to the favorite landscape scenario of step 1 could be chosen (Fig. 1, step 2 of the questionnaire). We took policy measures from the model variables and illustrated them with icons symbolizing their different levels (Fig. 2). The goal of this step was to ascertain participants’ policy preferences. Step 3. In order to understand the importance of visual stimuli in participants’ decision-making, we presented them with a conflict/choice exercise in the third step of the survey. Participants had to compare and choose between (1) their favorite visual representation of the landscape from step 1 combined with the respective policy mix and (2) their favorite policy mix of step 2 combined with the respective visual representation of the landscape scenario. We call this task a “conflict/choice exercise” because the visual representation of the landscape the participants chose during step 1 did not need to align with the policy preferences of step 2; hence, an inner conflict was provoked. This step aimed at better understanding participants’ trade-off between policy mixes when confronted with a given representation of a landscape.

43

Step 4. The participants conducted a pairwise comparison of 3D visualizations with policy mix information. This was very similar to the task performed in step 1; however, this time both sources of information (the landscape scenario visualization and the policy mix) were available. The aim of this step was to discover the preferences of participants if both sources of information were given. A certain mental fatigue given through the length of the survey could be expected and participants were expected to concentrate on one of the sources of information (landscape scenario or policy mix). Data was stored in a database for every participant and every choice situation. As data regarding visual and policy choices were dependent on each other, dependencies (e.g., preferences for visual representation of the landscape could be related to policy mixes) and trade-offs of policy mixes could be revealed. 2.3. Survey participants Participants completed the survey using either a computer with an internet connection or a tablet PC (e.g., iPad) with a local network connection. On average, surveys took between 10 and 15 min to complete. Table 1 shows the different sub-sample groups we used to test the tool. The different subsamples are groups of the broader public for whom the tool has been designed. For the “shops” subsample, we sampled both in the morning and in the afternoon at different locations. The order of the appearance of the visualizations was always randomized (cf. Bogartz, 1994) as sequencing might influence choices. 2.4. Survey analysis For further analysis of the survey, we counted the number of participants choosing a certain option. We show the movement from a free choice of policy-levels to a situation where participants were restricted by their own landscape preference, a process that provided information concerning the trade-off participants make in policy options. For analysis and diagram construction we used R (3.0.2). 3. Results 3.1. Preferences for the visual representations of the landscape scenarios Fig. 3 shows participants’ preferences for the visual representations of the landscape scenario. Scenarios show variance in agricultural intensity depicted by meadow types, settlement development, and forest cover differences. Participants preferred extensive use of agricultural land as well as restricted urban sprawl, considering in their choices more strongly the restriction of urban sprawl than the extensive use of agricultural land (cf. Fig. 3, VisID: B and Vis-ID: C). The most distinct difference in the preference ratings of the first step (beginning of the survey) was explained by the intensity of the agriculture in the foreground. Intensively cultivated agriculture was least accepted by the participants. The ranking of the scenarios stayed constant over the survey (i.e., during the first and the last step), suggesting that the ranking of the visual representations of the landscapes did not occur by chance. For this reason, we assume that distinctions between the visualizations were clear and allowed participants to make well-reasoned choices. Comparing first and last step in the survey, the spread of points per visualization remained similar from the top-ranked to the lowest-ranked visualization. The difference in points per visualization between the least-liked visualization and the rest of

44

E. Celio et al. / Landscape and Urban Planning 137 (2015) 40–53

Fig. 2. Set of policy measures used in the survey with a description of their levels.

the visualizations was particularly pronounced in the first step of the survey. The difference between the sub-samples regarding the ranking was only marginal (cf. Appendix C for the sub-sample split and characterization of the sub-samples). In step 3, participants were presented with their favorite visualization and their favorite policy mix. Participants more frequently chose the situation with the policy mix they had chosen than the situation with the preferred visualizations (Table 2). In other words, policy measures were slightly more important to participants than

the stimulus from the visualization. The only sub-sample that had a very slight tendency of focusing on the visualizations was the high school sub-sample. 3.2. Policy measures preferences The participants’ choices of policy measures reveal a relatively clear tendency regarding the most-preferred and the least-preferred policy mix (see Table 3). Whereas participants were

Table 1 Sub-sample size and description. Locations with 3D visualizations showing Schuepfheim view Sample name

Sub-sample size

Local shops

30

High school

64

Summer camp

8

Students

27

Web

15

TOTAL

144

Location

Participants

Sample was collected in front of two local shops on three days between 10 a.m. and 4 p.m. using iPads. Sample was collected in workshops with students using iPads.

Between 17 and 89 year old (mean 52). Female: 17, Male: 11, NA: 2. Persons living in study area: 26. Between 16 and 20 years old (and three teachers aged 29, 36, 51). All persons living in the study area. Between 17 and 19 years old (and two coaches aged 21 and 33). Female: 6, Male: 2. No persons living in the study area. Between 21 and 33 year old (mean 24). Female: 9, Male: 16. No persons living in the study area.

Using four iPads and four notebook PCs, participants of an ESRI summer camp were sampled. Bachelor’s and Master’s degree students at the Department of Civil, Environmental and Geomatic Engineering ETH Zurich using PCs. Sample consists of persons who had no time in front of shops or were participating because they heard about the survey from other sources.

Between 19 and 52 years old (mean: 42). Female: 2, Male: 8, NA: 5. Persons living in study area: 1.

E. Celio et al. / Landscape and Urban Planning 137 (2015) 40–53

45

Fig. 3. Ranking of the visualizations. The bar plot indicates the points per visualization and reveals how preferences changed between the first round (only visualizations) and the end of the survey (visualizations and policy measures are shown to participants). (For interpretation of the colors in this figure, the reader should refer to the web version of this article.)

against settlement enlargements, their choices were less clear with regard to the Federal and Cantonal spatial planning policies. The survey results regarding direct payments reveal that participants are more in favor of payments compensating for ecological

management than payments compensating for cultivated areas or for the number of animals. An ambiguous result, however, shows “agricultural policy” where choices were almost equally distributed between both options.

Table 2 Summary of choices in the third step of the survey: Comparison of favorite visual representation of the landscape to favorite policy mix. Quantity

>Preference for favorite visualization >Preference for favorite policy

58 45.3% 70 54.7%

Samples coming from . . . Shops

High School

11

31

Students 9

Others 7

17

29

15

9

Sixteen participants were consistent by choosing a visualization and the respective policy mix resulting in the same visualization. Hence, these cases were excluded in this analysis.

46

E. Celio et al. / Landscape and Urban Planning 137 (2015) 40–53

Table 3 Policy preferences of the unrestricted choice situation.

Green indicates the most preferred combination. Red indicates the least preferred combination.

3.3. Trade-off between policy measures Fig. 4 shows two examples (Vis-ID: A, Vis-ID: E) with the possible policy-levels that could be chosen once the landscape preference was revealed; i.e., the restricted policy choice situation (step 2 of the questionnaire). The colored circles show which consistent combination of policy-levels were given as one policy mix that could be chosen in the questionnaire. The parameters ecological direct payments, general direct payments, and Federal and Cantonal spatial planning policy show variability whereas agricultural policy and settlement enlargement are determined by binary states (settlement yes/no, forest yes/no) and are taken as fixed. In the two scenario examples shown in Fig. 4, we see the small sensitivity of the visual representation of the landscape (the landuse mosaic) with respect to spatial planning policy. For a visual representation of a given landscape scenario, spatial planning policy varies over all possible states. Analyzing the modeling output confirms this pattern, as land use itself only changes with respect to the probability of occurrence but not in terms of land-use categories. Looking at the variation of the policy-levels for the scenario with extensive foreground (vis-ID: A), we see a frontier at a level of general direct payments corresponding to 100% of 2012 general direct payments (the second level). In contrast, in the visualization with intensive foreground (vis-ID: E), four of five policy mixes show ecological payments at the lowest level. Furthermore, the connection between the visualized landscapes and the policy measures is displayed in Fig. 4. A visualization with extensive foreground is based

on rather high policy-levels for ecological direct payments and low levels of general direct payments while a visualization with intensive foreground is based, among other things, on low levels of ecological direct payments. Fig. 5 shows the choice pathways for the same two visualizations as above (i.e., for the varying policy measures depicted in Fig. 4), starting from an unrestricted choice situation to a situation where participants were restricted by their own visualization choice (possible combinations indicated through colors in Fig. 4). In other words, we show the trade-offs between policy mixes of the unrestricted choices (red) and the restricted choices (green). The arrows in Fig. 5 depict the choice paths participants have made and show participants’ movement in directions of possible policy mixes in a restricted situation, and thus show the movement from one policy mix to another. Visualization A (vis-ID: A) and Visualization E (vis-ID: E) in Fig. 5 represent the most-preferred and least-preferred visualizations, respectively. Therefore, the subsample size is different: n = 60 in landscape scenario visualization A and n = 10 in landscape scenario visualization E. The diagrams representing the other visualizations may be found in Appendix D. Visualization A showed an extensively used agricultural foreground with no settlement enlargement in the middle ground. In contrast, Visualization E was characterized by an intensively cultivated foreground and no settlement enlargement. In Fig. 5(a) and (b), participants modify their choices toward more ecological direct payments, reinforcing the link they seem to create between policy measures and visualizations (an extensive cultivated foreground).

E. Celio et al. / Landscape and Urban Planning 137 (2015) 40–53

47

Fig. 4. Possible policy-levels to choose from for the restricted policy choice situations. Colored circles indicate the combinations of levels that lead to the respective landscape. (For interpretation of the colors in this figure, the reader should refer to the web version of this article.)

Fig. 5. Choice pathways of the most- and least-preferred visualization. Bubble size is relative to the sub-sample size. (a), (b), and (c) show the decision pathways for the most-preferred landscape visualization. (d), (e), and (f) show the decision pathways for the least-preferred landscape visualization. (a) and (d) show the decision pathways for the policies ecological direct payments and general direct payments. (b) and (e) show the decision pathways for ecological direct payments and spatial planning policy. (c) and (f) show the decision pathways for spatial planning policy and general direct payments. (For interpretation of the colors in this figure, the reader should refer to the web version of this article.)

48

E. Celio et al. / Landscape and Urban Planning 137 (2015) 40–53

Analogously, the most frequent movements in Fig. 5(d) and (e) are in the direction of fewer ecological direct payments. Fig. 5(c) reveals that two-thirds of the respondents chose the same combination of general direct payments and spatial planning policy in both the unrestricted and the restricted choice situation; this is the highest number when comparing all five visualizations. Furthermore, in Fig. 5(a) and (c), we find choice pathways that head in opposite directions (exchanging the level of direct payments). However, in Fig. 5(f) the most-preferred policy mix cannot be chosen in the restricted choice situation; hence, participants had to move to another policy mix. Here, movement is found again toward the upper right corner (that is, toward the chosen visualizations). Similarly, in Fig. 5(d) and (e) of visualization E, participants chose the unrestricted choice situation, a “too extensive” policy mix; thus, their policy choice would have led to a less intensive agriculture than their choice in the visual choice situation. They were forced through their own visualization choice to move to another position because the “too extensive” policy mix did not lead to their favorite visualization. In other words, the gap between their favorite visualization and their favorite policy mix forced adaptation in the restricted policy choice situation. When comparing choice pathways of similar visualizations (e.g., foreground with medium-intensive agriculture), the pattern of the pathways is similar. We find a tendency toward the middle states for the two visualizations with medium-intensive cultivated land in the foreground (vis-ID B and D in Appendix D), and movements toward high ecological payments in visualizations with an extensively cultivated foreground (vis-ID A, and C in Appendix D).

4. Discussion This paper presents a prototypical tool for elaborating normative landscape scenarios in spatial planning and landscape development by establishing a rational relationship between policy instruments and landscape scenario visualizations. Results show how participants change their preferences for policy mixes related to landscape changes when restricted in their choices. While debate exists regarding preferences for landscape functions and aesthetics (Howley, 2011; Ives & Kendal, 2013), our results support Howley’s findings that respondents prefer extensive agricultural land use over medium-intensive and intensive land use. The fact that participants stated preferences for certain types of agricultural cultivation and restricted settlement development may be used for discussions on the local level when planning processes are initiated. The trade-off analysis showed how participants move to possible policy mixes based on their landscape scenario preferences, allowing discussions about their mindset. In general, participants reinforced the links they made between extensive agriculture and high ecological direct payments, and between intensive agriculture and lower ecological direct payments. By connecting restricted and unrestricted choice situations through a land-use model, one can inform planners about negotiable choices. In particular, this connection establishes a rationale between the possible choices. The connection of policy measures to the visualization brought participants to a conflict situation (step 2) that had been driven by their own choices and thus triggered more thoughtful and motivated choices. A portion of participants (on average 31%) repeated their choice from the unrestricted situation if the policy mix was still available. On the one hand, these participants had the chance to do so because they had already chosen a possible policy-level mix for their favorite visualization. On the other hand, their choice might have been the same because they could remember their choice in the unrestricted round and wanted to show consistency within the survey (cf. Falk & Zimmermann,

2013). The fact that only the high school sub-sample had a slight tendency of selecting preferred visualization above a preferred policy mix could be attributed to their age (Balling & Falk, 1982 in: Hunziker, Buchecker, & Hartig, 2007). This result requires further investigation that examines whether high school students are used to means of visual communication (through media consumption) and possibly lack understanding of policy mechanisms (cf. Berglund, 2008 for an overview of the possibilities and obstacles in childrens’ participation in planning). The analysis of the choice (or trade-off) pathways shows how participants move in directions of possible policy mixes and reveals if they are actors that maximize their utility. In order to classify participants’ movements in the diagrams and to reveal their rationale, the survey would need to be set up so that it controls for specific characteristics of the participants (e.g., their relationship to the visualized landscape or their occupation). The resulting typology of trade-off behavior could eventually help in landscape development as process leaders could work with stakeholder groups focusing on a given set of policy measures or visualizations as means of communication. These policies and visualizations help to answer some of the focal questions for developing normative landscape scenarios, such as “How should the landscape change?” (cf. Nassauer & Corry, 2004). Using this prototypical tool, we find favorite landscape scenarios and acceptable policy mixes that can serve as a guideline for landscape development. As this tool directly involves actors, the resulting directions for landscape development are inherently normative. While BLUMAP allows for the linking of different land-use categories with policy measures in a validated model, it only includes three main land-use types and does not allow full depiction of the landscape of the case study region. A first point for the revision of the approach would thus require the inclusion of small-scale objects, such as hedges and fruit trees (cf. Ives & Kendal, 2013). However, this is difficult not only from the point of view of the allocation of the landscape elements, but also due to the design of the elements. Procedural modeling techniques may provide support through the definition of design parameters for objects and grammars for distributing the objects in landscapes (GrêtRegamey, Celio, Klein, & Wissen Hayek, 2013; Neuenschwander, Wissen Hayek, & Grêt-Regamey, 2012). Another issue is that landscape elements change their value for people over time, introducing additional complexity to scenario studies (Hunziker et al., 2008). While we have focused on setting up a prototypical approach, more research is needed in linking land-use modeling and 3D visualizations in a generic manner as advocated by Beurden et al. (2006). The prototypical character of the tool is also reflected in the biased sample. We learned that, first, the complexity of the survey should not be underestimated as discussions with participants after their participation revealed that completing it was a great effort for them; second, this complexity is potentially not appropriate for a survey designed for passers-by and should probably be used in a workshop setting with more time to reflect upon the survey tasks. One other challenge that surfaced is that the different levels of realism in the foreground, middle ground, and background need to be improved (e.g., the inclusion of abstract representations for new houses or the relatively abstract borders of different agricultural intensities). The unnatural appearance of these abstract structures was a potential distraction for participants, and in general is known to significantly influence communication and motivation in participatory planning processes (Wissen Hayek, 2011). Some studies judge photo-realistic scenario representations as effective (cf. van Berkel & Verburg, 2014), arguing that they would potentially enable participants to better understand the services landscapes provide. Further pitfalls in generating visualizations could not completely be omitted (cf. compilation in Rehr, Williams, & Levin, 2014) as

E. Celio et al. / Landscape and Urban Planning 137 (2015) 40–53

differences in participants’ perception of images could not be controlled and are probably reflected in the participants’ preference statements. Participants may have interpreted our visualizations differently than intended, and responses might also have been biased by, e.g., a local social desirability. However, this was not evident when comparing local participants with other sub-samples which might potentially be due to abstract elements in the visualizations and hence a reduced identification with the visualizations. We did not closely follow the normative landscape scenario approach (Nassauer & Corry, 2004), as we used, e.g., a modeling approach for generating the basic scenarios. In this study, we focused on part two of their framework for the development of normative landscape scenarios. Therefore, the relevance and performance of future landscapes remain to be elaborated. However, through the restricted policy choice situation we determined the freedom of action concerning possible futures (cf. Robinson, 2003) as the sampled participants revealed both their favorite landscape visualization and their policy trade-offs. Combining the presented results with the development of storylines can be a valuable next step (cf. Kok & Van Delden, 2013). Feasible policy measures and future landscape development can serve as cornerstones for the storylines and can help make the link between policy and the overall appearance of a landscape more explicit. Adopting the perspective of a landscape scenario to refer a policy instrument, in other words, the ability to use the BN approach to back-propagate from an observation to its causes, allows exploration of the feasibility of possible future options as well as their probability. This, in turn, allows identification of the values of the input variables necessary to generate a given result. Back propagation is related to a backcasting that could be achieved through determining feasible policy measures. Traditionally, backcasting exercises require multiple rounds of respondent feedback regarding their preferences and policy goals. Cost and time limited our ability to complete a full backcasting exercise. In order to overcome this limitation, more sampling is needed to elicit multiple points in time to create a pathway (i.e., a landscape development process with specific milestones and policy measures). With land-use models, the feasibility for landscape developments can be explored through determining the combination of factors that will result in a desired future landscape. van Berkel and Verburg (2012) applied a similar approach to a cultural landscape in the Netherlands. They used an agent-based model for forecasting alternative socio-economic developments and backcasted these scenarios in a workshop where they defined local interventions to achieve predefined landscape goals. In order to develop backcasts solely with a quantitative model, an inverse modeling approach can be applied (e.g. Grêt-Regamey & Crespo, 2011). Grêt-Regamey and Crespo’s approach based on geographically weighted regression does not consider normative statements, which are often used in backcasting.

5. Conclusion This study focused on the question “How should the landscape change?” and presented a tool to support the development of normative landscape scenarios. The prototypical survey tool links policy measures and visual representations of the landscape for understanding choices and trade-offs between policy measures related to landscape development scenarios. An important characteristic of the tool is that participants choose their preferred future first while the respective possible policy options are linked to those future landscape developments via a landuse model and later presented to the participants in a second step. Participants revealed preferences for extensively used agricultural land and restrictive settlement planning. Choice pathways

49

in directions of possible policy mixes revealed the preferences of actors given a restricted choice situation. Participants had to choose options constrained through their landscape preference, revealing their rationale between the different choices. The resulting choice pathways were not purely driven by cost-benefit deliberations; rather, these pathways confirmed the participants’ favorite visualizations. One can thus postulate that once participants select their visual representation of a landscape, they are potentially willing to choose supportive policy options regarding their landscape choice. Acknowledgments The authors are grateful to the National Research Program NRP 61 (“Sustainable water management”) for funding the project “HydroServ”, to the Cantonal administration of Lucerne for address information and geodata provision, and to the representatives of the municipalities of Schuepfheim and Hasle as well as to the respondents of the survey who invested their time for this contribution. The authors would like to thank the team members who have supported this work: Ralph Sonderegger for web and icon design, Madeleine Manyoky, Ulrike Wissen Hayek, Thomas M. Klein, Noemi Neuenschwander for their advice and support regarding visualizations, and Maarten van Strien regarding the manuscript, Stefanie Müller for preparing 3D visualizations. Furthermore, Florian Knaus supported this work with his ideas and introduced us to the municipalities and to the Netzwerk Schweizer Pärke. Marcel Hunziker supported this work as an expert regarding the survey set-up and the visualizations. We would like to thank the three anonymous reviewers for their valuable comments on the manuscript. Appendix A. Information regarding settlement planning in the municipality of Schuepfheim The municipality of Schuepfheim has 65.5 ha of built settlement area (residential and mixed-use areas) and 13.8 ha of unbuilt settlement area (in 2010; Canton of Lucerne, Bauzonenstatistik); 7.92 ha are being held in reserve for residential use, enough for a potential 558 new inhabitants (Schuepfheim, 2011). The demand for unbuilt residential area is, of course, highly uncertain; nevertheless, extrapolating the population growth from 1991 to 2009, additional residential area for about 218 inhabitants is needed. Despite these reserves, the general principle for settlement development proposes enlarging the settlement area by 7.8 ha, or by a capacity of about 480 inhabitants (Schuepfheim, 2011). References Schuepfheim, Gemeinde. (2011). Siedlungsleitbild (pp. 41). Schuepfheim: Gemeinde Schuepfheim. Appendix B. Technical specification and detailed set-up process of the Bayesian network based land-use modeling approach (BLUMAP) and its use for the questionnaire set-up B.1. BLUMAP set-up This model set-up description is based on Celio et al. (2014) and shows therein used stepwise approach (Fig. A1). For setting up BLUMAP, eight consecutive (partially iterative) steps are passed which are based on existing guidelines for setting up Bayesian networks (BN) (Cain, 2001; Carmona & Varela-Ortega, 2007; Chen & Pollino, 2012). First, land-use related problems were defined and possible driving factors of land-use decisions were gathered through a literature review (e.g. Althaus, 2008; Beedell & Rehman, 2000; Bolliger et al., 2007; Edwards-Jones, 2006) and

50

E. Celio et al. / Landscape and Urban Planning 137 (2015) 40–53

Fig. A1. BLUMAP set-up process. Source: Celio et al. (2014).

eight expert interviews according to the guidance in Bromley (2005). For step two to five, an expert group consisting of six land-use experts was formed (two for agriculture, forestry and settlement, respectively). Workshop sessions with the experts were held alternately with one single expert and with a group of experts being able to track uncertainties in model construction and parameterization. Step 2 was dedicated to the weighting of land-use drivers (according to the swing weights method, cf. Scholz & Tietje, 2002) and step 3 to the definition of causal relation (according to an impact matrix, cf. Scholz & Tietje, 2002). In step 4, experts defined reasonable states for all nodes and in step 5, experts elicited conditional probabilities of intermediate nodes. In step 6, questionnaire data was used to update local actor characteristics (such as attitude toward federal agricultural programs, environmental awareness, etc.) in the BN. For each local actor characteristic, a Likert item box was included in the previously pretested questionnaire. Step 7 and 8 were dedicated to the explorative calibration with the help of the experts and quantitative and subjective validation (Celio et al., 2012). B.2. BLUMAP modeling Land use vector data was converted into raster cells. We used ArcGIS 10.1 and 10.2 (ESRI Inc.) for data preparation, Netica 4.16 and Netica-J 4.18 (Norsys Software Corp.) for BN modeling. BN consist of a set of variables (called nodes) and a set of directed graphs. Each node possesses a finite set of mutually exclusive states and a conditional probability table (CPT). Nodes are connected mathematically through Bayes’ rule of conditional probability: P(x|e) = P(x,e)/P(e), where x represents the probability of a parent node and e represents evidence for a child node. Geodata and the three BN were integrated in a Java program (elaborated in Eclipse) allowing to loop cell-by-cell (using ArcObjects), through different time steps. For each cell, the BN is called

and configured with the relevant data at this location and in the following probabilities of the land uses at t1 are then calculated. We used different time dynamics for the three different land-use sectors: settlement cells were recalculated after 2.5 years, agricultural cells after 5 years and forest cells after 25 years. B.3. Use of BLUMAP and its evaluation for the questionnaire design We used a multi-processing set-up to run the 432 policy-level combinations in parallel, thus to calculate the different raster outputs. For each model run, metadata was collected (variable combination, the simulation duration, etc.) to allow for tracking of the run properties. Python was used to organize the overall workflow coordinating each run to be called with the correct set of variables and its characteristics on the right time. For every scenario the indicators ‘allocation disagreement’ and ‘quantity disagreement’ (Pontius & Millones, 2011) were calculated based on land-use model outputs using ‘gdal’ and ‘numpy’ (www.gdal.org, www.numpy.org). The comparison of present state and future scenario allowed for determining the magnitude with which a scenario differed from the current land-use mosaic. As not every policy-level change resulted in a visible change in the land-use map, we grouped the different outputs by finding appropriate quantiles with respect to allocation and quantity disagreement. Grouping similar values of allocation and quantity disagreement allowed distinguishing relevant land-use changes. These groups were used as selection criterion for choosing the land-use mosaics to be visualized. B.4. Creation of 3D visualization (1) The raster output was converted to polygons and raster edges were smoothed. These polygons were then differentiated according to their land use. These land use separated polygons were used as input for the 3D visualization procedures. All data handling was done with ESRI ArcGIS 10.1 and 10.2. Heights of swisstopo

E. Celio et al. / Landscape and Urban Planning 137 (2015) 40–53

dataset swissBuildings3D (Swisstopo, 2013) was joined to building footprints from the survey office of Canton of Lucerne (“Amtliche Vermessung”). These footprints with building heights were used as input for the 3D visualization. (2) We used the game engine Crytek CryEngine 3.4.5 (and the editor Sandbox) to create the 3D visualizations based on the land-use model output. The game engine allowed us to establish the connection between the geo-referenced modeling output and 3D visualizations (more details about the approach to import spatially-reference data in the editor is given in Manyoky et al. (2014)). Finally, Adobe Photoshop (CS 5) editing was used to reduce the artificial appearance of the semi-realistic colored orthophotos by masking agricultural land uses in the middle- and back-ground and introducing texture for these.

B.5. Web-based tool The web-based tool was set-up in Python and uses sqlite (version 2.0b1) for data storage. For the portable version (for conducting the survey with tablet-PCs) a local wireless network was established and data was stored on a portable server. Data storage and retrieving was key for the tool design as participants received feedback to their choices real-time. References Althaus, P. (2008). Entscheidungsverhalten junger Landwirte— analyse mittels ökonomischer Experimente. ETH Zürich. Beedell, J., & Rehman, T. (2000). Using social-psychology models to understand farmers’ concervation behaviour. Journal of Rural Studies, 16, 117–127. Bolliger, J., Kienast, F., Soliva, R., & Rutherford, G. (2007). Spatial sensitivity of species habitat patterns to scenarios of land use change (Switzerland). Landscape Ecology, 22, 773–789. doi: 10.1007/s10980-007-9077-7. Bromley, J. (2005). Guidelines for the use of Bayesian networks as a participatory tool for Water Resource Management. NERC Open Research Archive, 137. Cain, J. (2001). Planning improvements in natural resources management. Guidelines for using Bayesian networks to support the planning and management of development programmes in the water sector and beyond. Centre for Ecology and Hydrology.

51

Carmona, G., & Varela-Ortega, C. (2007). Integration of Bayesian Networks and agro-economic models as a decision support system for water management in the upper Guadiana Basin. CAIWA, International Conference on Adaptive and Integrated Water Management, 12–15. Celio, E., Brunner, S. H., & Grêt-Regamey, A. (2012). Participatory land use modeling with Bayesian networks: A focus on subjective validation. Proceedings of the 2012 International Congress on Environmental Modelling and Software, Sixth Biennial Meeting, Leipzig, Germany. Celio, E., Koellner, T., & Grêt-Regamey, A. (2014). Modeling land use decisions with Bayesian networks: spatially explicit analysis of driving forces on land use change. Environmental Modelling & Software. http://dx.doi.org/10.1016/j.envsoft.2013.10.014. Chen, Serena H., & Pollino, Carmel A. (2012). Good practice in Bayesian network modelling. Environmental Modelling & Software, 37(0), 134–145. http://dx.doi.org/10.1016/j.envsoft.2012.03.012. Edwards-Jones, G. (2006). Modelling farmer decision-making: concepts, progress and challenges. Animal Science, 82(06), 783790. Manyoky, Madeleine, Wissen Hayek, Ulrike, Heutschi, Kurt, Pieren, Reto, & Grêt-Regamey, Adrienne. (2014). Developing a GISbased visual-acoustic 3D simulation for wind farm assessment. ISPRS International Journal of Geo-Information, 3(1), 29–48. Pontius, R.G., & Millones, M. (2011). Death to Kappa: Birth of quantity disagreement and allocation disagreement of accuracy assessment. International Journal of Remote Sensing, 32(15), 4407–4429. http://www.tandfonline.com/doi/abs/10.1080/ 01431161.2011.552923. Scholz, R.W., & Tietje, D. (2002). Embedded case study methods. Integrating quantitative and qualitative knowledge: Sage Publications. Swisstopo. SwissBuildings3D. (2013). http://www.swisstopo. admin.ch/internet/swisstopo/en/home/products/landscape/ swissBUILDINGS3D.html Retrieved 01.09.13.

Appendix C. Result of the sub-sample analyses

Fig. A2. Visualization and combined preferences of sub-samples. First and last step of survey.

52

E. Celio et al. / Landscape and Urban Planning 137 (2015) 40–53

Appendix D. Choice pathways per visualization. Bubble size is relative to the sub-sample size

Fig. A3. Choice pathways of visualizations B, C, and D. Bubble size is relative to the sub-sample size. (For interpretation of the colors used in this figure, see the web version of this article.).

References ˜ ˜ ˜ J. A., & Ruiz-Aviles, P. (2004). Arriaza, M., Canas-Ortega, J. F., Canas-Madue no, Assessing the visual quality of rural landscapes. Landscape and Urban Planning, 69(1), 115–125, http://dx.doi.org/10.1016/j.landurbplan.2003.10.029 Balling, John D., & Falk, John H. (1982). Development of visual preference for natural environments. Environment and Behavior, 14(1), 5–28. http://dx.doi. org/10.1177/0013916582141001 Bateman, Ian J., Day, Brett H., Jones, Andrew P., & Jude, Simon. (2009). Reducing gain–loss asymmetry: A virtual reality choice experiment valuing land use change. Journal of Environmental Economics and Management, 58(1), 106–118, http://dx.doi.org/10.1016/j.jeem.2008.05.003. Berglund, U. (2008). Using children’s GIS maps to influence town planning. Children, Youth and Environments, 18(2), 110–132. Beurden, J. A. M., Borsboom-van, Lammeren, R. J. A., Hoogerwerf, T., & Bouwman, A. A. (2006). Linking land use modelling and 3D visualisation. In J. Leeuwen, & H. P. Timmermans (Eds.), Innovations in design & decision support systems in architecture and urban planning (pp. 85–101). Netherlands: Springer. BFS. Betriebszählung—business census. (2008). http://www.bfs.admin.ch/ bfs/portal/en/index/infothek/erhebungen quellen/blank/blank/bz/10.html. Bogartz, R. S. (1994). An introduction to the analysis of variance. Westport: Praeger Publishers.

Börjeson, Lena, Höjer, Mattias, Dreborg, Karl-Henrik, Ekvall, Tomas, & Finnveden, Göran. (2006). Scenario types and techniques: Towards a user’s guide. Futures, 38(7), 723–739, http://dx.doi.org/10.1016/j.futures.2005.12.002. Celio, E., Brunner, S.H., & Grêt-Regamey, A. (2012). Participatory land use modeling with bayesian networks: a focus on subjective validation. Proceedings of the 2012 International Congress on Environmental Modelling and Software, Sixth Biennial Meeting, Leipzig, Germany. Celio, E., Koellner, T., & Grêt-Regamey, A. (2014). Modeling land use decisions with Bayesian networks: Spatially explicit analysis of driving forces on land use change. Environmental Modelling & Software, http://dx.doi.org/10.1016/ j.envsoft.2013.10.014. Ervin, Stephen M. (2001). Digital landscape modeling and visualization: A research agenda. Landscape and Urban Planning, 54(1–4), 49–62, http://dx.doi.org/ 10.1016/S0169-2046(01)00125-6. Falk, Armin, & Zimmermann, Florian. (2013). A taste for consistency and survey response behavior. CESifo Economic Studies, 59(1), 181–193. http://dx.doi.org/10.1093/cesifo/ifs039 Gregory, Robin, & Wellman, Katharine. (2001). Bringing stakeholder values into environmental policy choices: A community-based estuary case study. Ecological Economics, 39(1), 37–52, http://dx.doi.org/10.1016/S0921-8009(01)00214-2. Grêt-Regamey, A., Celio, E., Klein, T. M., & Wissen Hayek, U. (2013). Understanding ecosystem services trade-offs with interactive 3D procedural

E. Celio et al. / Landscape and Urban Planning 137 (2015) 40–53 modeling for sustainable urban planning. Landscape and Urban Planning, 109, 107–116. Grêt-Regamey, A., & Crespo, R. (2011). Planning from a future vision: inverse modeling in spatial planning. Environment and Planning B-Planning & Design, 38(6), 979–994. http://dx.doi.org/10.1068/b37083 Hamidi, Shima, & Ewing, Reid. (2014). A longitudinal study of changes in urban sprawl between 2000 and 2010 in the United States. Landscape and Urban Planning, 128(0), 72–82, http://dx.doi.org/10.1016/j.landurbplan.2014.04.021. Haslauer, Eva, Biberacher, Markus, & Blaschke, Thomas. (2012). GIS-based backcasting: An innovative method for parameterisation of sustainable spatial planning and resource management. Futures, 44(4), 292–302, http://dx.doi.org/10.1016/j.futures.2011.10.012. Home, Robert, Bauer, Nicole, & Hunziker, Marcel. (2010). Cultural and biological determinants in the evaluation of urban green spaces. Environment and Behavior, 42(4), 494–523. http://dx.doi.org/10.1177/0013916509338147 Howley, Peter. (2011). Landscape aesthetics: Assessing the general publics’ preferences towards rural landscapes. Ecological Economics, 72(0), 161–169, http://dx.doi.org/10.1016/j.ecolecon.2011.09.026. Hunziker, Marcel, Buchecker, Matthias, & Hartig, Terry. (2007). Space and place—two aspects of the human-landscape relationship. In F. Kienast, O. Wildi, & S. Ghosh (Eds.), A changing world (8) (pp. 47–62). Netherlands: Springer. Hunziker, Marcel, Felber, Patricia, Gehring, Katrin, Buchecker, Matthias, Bauer, Nicole, & Kienast, Felix. (2008). Evaluation of landscape change by different social groups. Mountain Research and Development, 28(2), 140–147. http://dx.doi.org/10.1659/mrd.0952 Ives, Christopher D., & Kendal, Dave. (2013). Values and attitudes of the urban public towards peri-urban agricultural land. Land Use Policy, 34(0), 80–90, http://dx.doi.org/10.1016/j.landusepol.2013.02.003. Junge, Xenia, Lindemann-Matthies, Petra, Hunziker, Marcel, & Schüpbach, Beatrice. (2011). Aesthetic preferences of non-farmers and farmers for different land-use types and proportions of ecological compensation areas in the Swiss lowlands. Biological Conservation, 144(5), 1430–1440, http://dx.doi.org/10.1016/j.biocon. 2011.01.012. Kaplan, Rachel, & Kaplan, Stephen. (1989). The experience of nature. A psychological perspective. Cambridge: Cambridge University Press. Kjaerulff, U. B., & Madsen, A. L. (2008). Bayesian networks and influence diagrams. Springer. Kok, Kasper, & Van Delden, H. (2013). Linking narrative storylines and quantitative models to combat desertification in the Guadalentín watershed (Spain). In M. Giaoutzi, & B. Sapio (Eds.), Recent developments in foresight methodologies, complex networks and dynamic systems (1) (pp. 203–214). New York: Springer Science+Business Media. Kok, Kasper, van Vliet, Mathijs, Bärlund, Ilona, Dubel, Anna, & Sendzimir, Jan. (2011). Combining participative backcasting and exploratory scenario development: Experiences from the SCENES project. Technological Forecasting and Social Change, 78(5), 835–851, http://dx.doi.org/10.1016/j.techfore.2011.01.004. Koomen, E., Rietveld, P., & de Nijs, T. (2008). Modelling land-use change for spatial planning support. Annals of Regional Science, 42(1), 1–10. http://dx.doi.org/10.1007/s00168-007-0155-1 Lindemann-Matthies, Petra, Briegel, Reinhold, Schüpbach, Beatrice, & Junge, Xenia. (2010). Aesthetic preference for a Swiss alpine landscape: The impact of different agricultural land-use with different biodiversity. Landscape and Urban Planning, 98(2), 99–109, http://dx.doi.org/10.1016/j.landurbplan.2010.07.015. Lloret, J., Rodríguez, Omtzigt, N., Koomen, E., & de Blois, F. S. (2008). 3D visualisations in simulations of future land use: Exploring the possibilities of new, standard visualisation tools. International Journal of Digital Earth, 1(1), 148–154. http://dx.doi.org/10.1080/17538940701782718 LUSTAT. Statistik luzern— lucerne statistics. Statistikportal. (2013). http://www. lustat.ch/index.htm. Manyoky, M., Wissen Hayek, U., Klein, T.M., Pieren, R., Heutschi, K., & Grêt-Regamey, A. (2012). Concept for collaborative design of wind farms facilitated by an

53

interactive GIS-based visual-acoustic 3D simulation. Paper presented at the Digital Landscape Architecture, Berlin and Offenbach. Marcot, B., Steventon, J. D., Sutherland, G. D., & McCann, R. (2006). Guidelines for developing and updating Bayesian belief networks applied to ecological modeling and conservation. Canadian Journal of Forest Research, 36, 3063–3074. Nassauer, Joan Iverson, & Corry, Robert C. (2004). Using normative scenarios in landscape ecology. Landscape Ecology, 19(4), 343–356. http://dx.doi.org/10.1023/B:LAND. 0000030666.55372.ae Neuenschwander, N., Wissen Hayek, U., & Grêt-Regamey, A. (2012). Integrated multi-criteria modeling and 3D visualization for informed trade-off decision making on urban development options. Paper presented at the eCAADe Prague [12th–14th September 2012]. Plieninger, Tobias, van der Horst, Dan, Schleyer, Christian, & Bieling, Claudia. (2014). Sustaining ecosystem services in cultural landscapes. Ecology and Society, 19(2) http://dx.doi.org/10.5751/ES-06159-190259 Rehr, Amanda P., Williams, Gregory D., & Levin, Phillip S. (2014). A test of the use of computer generated visualizations in support of ecosystem-based management. Marine Policy, 46(0), 14–18, http://dx.doi.org/10.1016/j.marpol.2013.12.012. Robinson, J. (1982). Energy backcasting: A proposed method of policy analysis. Energy Policy, 10(4), 337–344, http://dx.doi.org/10.1016/0301-4215(82)900489. Robinson, J. (2003). Future subjunctive: Backcasting as social learning. Futures, 35(8), 839–856, http://dx.doi.org/10.1016/S0016-3287(03)00039-9. Rounsevell, M. D. A., Pedroli, Bas, Erb, Karl-Heinz, Gramberger, Marc, Busck, Anne Gravsholt, Haberl, Helmut, et al. (2012). Challenges for land system science. Land Use Policy, 29(4), 899–910, http://dx.doi.org/10.1016/j.landusepol.2012.01.007. Rounsevell, M. D. A., Reginster, I., Araújo, M. B., Carter, T. R., Dendoncker, N., Ewert, F., et al. (2006). A coherent set of future land use change scenarios for Europe. Agriculture, Ecosystems & Environment, 114(1), 57–68, http://dx.doi.org/ 10.1016/j.agee.2005.11.027. Schindler, Richard. (2005). Landschaft verstehen. Industriearchitektur und Landschaftsästhetik im Schwarzwald. Freiburg i.Br.: modo Verlag GmbH. Scholz, R. W., & Tietje, D. (2002). Embedded case study methods. Integrating quantitative and qualitative knowledge. Sage Publications. van Berkel, Derek B., & Verburg, Peter H. (2012). Combining exploratory scenarios and participatory backcasting: Using an agent-based model in participatory policy design for a multi-functional landscape. Landscape Ecology, 27(5), 641–658. http://dx.doi.org/10.1007/s10980-012-9730-7 van Berkel, Derek B., & Verburg, Peter H. (2014). Spatial quantification and valuation of cultural ecosystem services in an agricultural landscape. Ecological Indicators, 37, 163–174 (Part A). http://dx.doi.org/10.1016/j.ecolind.2012.06.025. van Notten, Philip W. F., Rotmans, Jan, van Asselt, Marjolein B. A., & Rothman, Dale S. (2003). An updated scenario typology. Futures, 35(5), 423–443, http://dx.doi.org/10.1016/S0016-3287(02)00090-3. van Zanten, Boris T., Verburg, Peter H., Espinosa, Maria, Gomez-y-Paloma, Sergio, Galimberti, Giuliano, Kantelhardt, Jochen, et al. (2014). European agricultural landscapes, common agricultural policy and ecosystem services: A review. Agronomy for Sustainable Development, 34(2), 309–325. http://dx.doi.org/ 10.1007/s13593-013-0183-4 Verburg, Peter H., Rounsevell, Mark D. A., & Veldkamp, A. (2006). Scenario-based studies of future land use in Europe. Agriculture, Ecosystems & Environment, 114(1), 1–6, http://dx.doi.org/10.1016/j.agee.2005.11.023. Walz, A., Braendle, J. M., Lang, D. J., Brand, F., Briner, S., Elkin, C., et al. (2013). Experience from downscaling IPCC-SRES scenarios to specific national-level focus scenarios for ecosystem service management. Technological Forecasting and Social Change, http://dx.doi.org/10.1016/j.techfore.2013.08.014. Wissen Hayek, U. (2011). Which is the appropriate 3D visualization type for participatory landscape planning workshops? A portfolio of their effectiveness. Environment and Planning B: Planning & Design, 38(5), 921–939. http://dx.doi.org/ 10.1068/b36113