Producing a sensitivity assessment method for visual forest landscapes

Landscape and Urban Planning 144 (2015) 128–141

Contents lists available at ScienceDirect

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

Research Paper

Producing a sensitivity assessment method for visual forest landscapes Ron Store a,∗ , Eeva Karjalainen b , Arto Haara c , Pekka Leskinen d , Vesa Nivala e a

Natural Resources Institute Finland (Luke), Economics and Society, Silmäjärventie 2, FI-69100 Kannus, Finland Natural Resources Institute Finland (Luke), New Business Opportunities, Jokiniemenkuja 1, FI-01370 Vantaa, Finland Natural Resources Institute Finland (Luke), Economics and Society, Yliopistonkatu 6, FI-80100 Joensuu, Finland d Finnish Environment Institute, P.O. Box 111, FI-80100 Joensuu, Finland e Natural Resources Institute Finland (Luke), Economics and Society, Eteläranta 55, FI-96300 Rovaniemi, Finland b c

h i g h l i g h t s • • • •

A method and a model were developed to assess the sensitivity of visual forest landscapes. In the case study a sensitivity classification for a land area of more than 27 000 km2 was produced. The most sensitive areas are located in high places and experience intensive outdoor recreation use. Sensitivity values estimated by the model are quite similar to values calculated from expert opinions.

a r t i c l e

i n f o

Article history: Received 27 February 2014 Received in revised form 1 June 2015 Accepted 5 June 2015 Available online 22 October 2015 Keywords: Landscape sensitivity Landscape planning Spatial multi-criteria evaluation Geographical information systems Expert knowledge modeling

a b s t r a c t A landscape sensitivity index provides information about the location of the most sensitive forest areas in terms of visual alteration. This information is needed to recognize those areas which require special attention in terms of management policy decisions and in directing landscape management activities and subsidies. The main goal of this study was to develop and test a GIS-based method to enable the production of a sensitivity index map on a regional scale. To accomplish this, sensitivity criteria, a model and calculating techniques were developed for the landscape province of the Kainuu and Kuusamo hill area in Finland. Sensitivity was described using three main criteria: (i) visibility, (ii) the amount of potential users (use pressure) and (iii) the attractiveness of the landscape – which are further defined by several sub-criteria. The calculation method was based on spatial multicriteria evaluation (SMCE), where cartographic modeling and expert knowledge modeling are utilized. The method was demonstrated and tested by a case study, where a visual landscape sensitivity map was produced for one municipality in the selected landscape province. The results were evaluated by forest and environment experts. The evaluation process showed that the sensitivity values estimated by the sensitivity model were quite similar to the values calculated from the expert map and field evaluations. © 2015 Elsevier B.V. All rights reserved.

1. Introduction The quality of the visual landscape is important not only to individual citizens and their health and well-being (Richardson et al., 2012; van Dillen, de Vries, Groenewegen, & Spreeuwenberg, 2012), but also to the livelihood of rural areas. Landscapes are the central attraction in nature-based tourism, and an appealing landscape can

∗ Corresponding author. Tel.: +358 29 532 3423. E-mail addresses: ron.store@luke.fi (R. Store), eeva.karjalainen@luke.fi (E. Karjalainen), arto.haara@luke.fi (A. Haara), Pekka.leskinen@ymparisto.fi (P. Leskinen), vesa.nivala@luke.fi (V. Nivala). http://dx.doi.org/10.1016/j.landurbplan.2015.06.009 0169-2046/© 2015 Elsevier B.V. All rights reserved.

attract other livelihoods and new residents to rural areas. Some 52% of Finnish forestry land is owned by private forest owners and 35% by the state (Finnish Statistical Yearbook of Forestry, 2012). Private forest owners appreciate the landscape provided by their forests. Around 20–30% of them emphasize the recreational and scenic values in their forest ownership (Karppinen, 1998a, 1998b), and 40% have areas in their forests that they will not fell because of the scenic values (Hänninen & Kurttila, 2007). Legislation has imposed general societal obligations on Metsähallitus (the authority responsible for the care of state forests), which include the responsibility for promoting outdoor recreation. Finland has also ratified the European landscape convention, which commits the authorities to protect, plan and manage landscapes.

R. Store et al. / Landscape and Urban Planning 144 (2015) 128–141

While the majority of the forest area is used for timber production, there is a need to map areas where special attention should be paid to strategic and operative planning because of their sensitivity to visual alteration such as forest fellings. In Finland, the forestry organizations often have information and databases on the locally valuable forest stands and small-scale destinations. However, until now forestry professionals and forest owners have not had enough knowledge to support their decision-making in terms of the visibility and scenic attractiveness of specific forest areas on a regional scale. Thus, equal attention is paid to forest landscape management everywhere, despite the visual sensitivities of the various places concerned. However, it would be more appropriate both scenically and economically to identify the visually sensitive areas and to focus landscape management activities on these areas. The term landscape sensitivity has been used to indicate geomorphic sensitivity, which means how geomorphic systems respond to environmental change such as erosion, increasing temperature, winds and storms, or human activity (Harvey, 2001). It can imply both resilience to change and the ability to recover from change. In forest landscape planning, the concept of landscape sensitivity or visual sensitivity is often defined as the resilience or fragility of the visual forest landscape to changes, such as altering land-use or forest fellings (Forest Landscape Design Guidelines, 1994; Visual Landscape Inventory, 1997). Landscape sensitivity can be defined as the likelihood that implementing forestry practices or other activities would evoke criticism and concern from the public (Visual Landscape Inventory, 1997). In some countries, landscape sensitivity is assessed as a part of forest landscape design or the forest management planning process (Bell, 1998; Landscape Aesthetics, 1995; Visual Landscape Inventory, 1997; Visual Resource Management, 2013). Landscape sensitivity can be used either as such in the planning or in determining scenic classes, which in turn are utilized to address the visual values in management planning. The terms used, the variables and targets assessed and the methods all vary, but the key concepts of the sensitivity assessments are: (1) visibility and distance zones from viewing points, (2) types and amount of use, (3) scenic attractiveness and quality, and (4) viewers’ experiences. Common to these landscape design methods (Bell, 1998; Landscape Aesthetics, 1995; Visual Landscape Inventory, 1997; Visual Resource Management, 2013) is that they are developed for large-scale landscapes, they often concern only specific areas, and they require different types of expert assessments, visitor surveys and other fieldwork. These models cannot be directly applied to Finnish conditions because the character of forest landscapes, the use of forests, and the prerequisites for this type of work differ. In Finland, viewing distances are relatively short and within-forest views are common, and it is not possible to use a lot of expert assessment and other fieldwork. In Finland, there is a need for a cost-efficient, robust, repeatable and automated landscape sensitivity assessment method that would be suitable for the whole country and that could be updated easily. The method should produce information about the location of the most sensitive forest areas for visual alteration at the regional level, i.e., to recognize areas that require special attention in the strategic and operative forest management planning and decision-making of forest owners. Sensitivity information is also needed in decisions concerning management policy, and in directing landscape management activities and subsidies. When formulating a sensitivity assessment task as a form of a formal decision model, the real world situation has to be simplified, as is the case with any other modeling task. However, there are several different criteria that have an influence on landscape sensitivity and have to be taken into consideration in calculations. In this kind of circumstance, multi-criteria evaluation (MCE) can be used to help gather the information together to support the decision and

129

to combine the evaluation criteria in a commensurable way (e.g., Kangas, Kangas, Leskinen, & Pykäläinen, 2001). The MCE methods provide a basis for evaluating a number of alternative choices on the basis of multiple criteria (Nijkamp, Rietveld, & Voogd, 1990). With the help of MCE, it is also possible to combine decision criteria measured in different measurement units, such as money and hectares on a common utility scale. Geographic information system (GIS) applications have been adopted by suitability assessment concerning large areas (e.g., Carver, 1991; Jankowski, 1995; Siddiqui, Everett, & Vieux, 1996; Store, 2009; Store & Kangas, 2001). GIS is used for producing the data needed in evaluations and for a platform of calculations. GIS-based multi-criteria evaluation, sometimes called spatial multi-criteria evaluation (SMCE), has some advantages compared to traditional MCE, e.g., the Analytical Hierarchy Method (AHP) described by Saaty (1980), because with SMCE it is possible to analyze a huge amount of alternatives. In other words, when AHP is meant to analyze the performance of a few decision alternatives, it is also possible to use modified MCE approaches that enable the comparison of basically an infinite number of decision alternatives. In the MCE literature, one refers to discrete choice problems and continuous problems (e.g., Jankowski, 1995). Because the criteria of landscape sensitivity are not necessary equally important, a sensitivity model with appropriate weights assigned to different criteria has to be developed. In situations where objective information and applicable models based on empirical data are inadequate or unavailable, it is possible to use expert knowledge modeling. Methods and techniques for utilizing expert knowledge in handling natural resources have been developed and used in forestry, for example (Alho & Kangas, 1997; Alho, Kangas, & Kolehmainen, 1996; Kangas, Karsikko, Laasonen, & Pukkala, 1993; Kangas, Store, Leskinen, & Mehtätalo, 2000; Store & Kangas, 2001). The goal of the present study is to develop a landscape sensitivity assessment method that is suitable on a regional scale. For this purpose, the special aims are: (1) to compile the criteria and model for the landscape sensitivity assessment to one landscape province in Finland, and (2) to develop and test a new GIS-based method whereby it is possible to calculate a sensitivity index map for a visual landscape. The method is demonstrated and tested through a case study, where a landscape sensitivity map is produced for one municipality in the selected landscape province and the results are evaluated by forest and environment experts.

2. Study area Landscape features differ in different parts of Finland to the extent that it is not possible to assess the landscape sensitivity for the whole country based on the exact same criteria or model. In Finland, the country is divided into 10 landscape provinces, parts of which are further divided into different regions (Maisemanhoito, 1992; Fig. 1). Landscape provinces provide a good starting point for developing the sensitivity criteria. The landscape provinces have been determined based on the significant natural features and their variability, such as landforms, soil, vegetation, as well as on the cultural characteristics of rural landscapes, while city landscapes have not been paid attention to. The landscape province of the Kainuu and Kuusamo hill area was selected for this study to produce the sensitivity model because it includes a variety of landscape elements, such as variability in topography, a lot of lakes and watercourses, forests, mires and field areas. Besides permanent settlements, there are also a lot of second homes and tourist destinations. Because of the abundance of landscape features and different types of landscape users, it can be assumed that the sensitivity criteria developed for the Kainuu

130

R. Store et al. / Landscape and Urban Planning 144 (2015) 128–141

promontory surrounded by waterways and has special landscape beauty values. Tourism and forestry are important sources of livelihood in this area. Approximately 1 million tourists visit Sotkamo each year, and the Vuokatti district is one of the most popular winter tourism areas in Finland. Parts of Sotkamo belong to nationally and regionally valuable landscape areas. The landscape is diverse, consisting of forests, bodies of water, ridges and hills, settlement areas and pastureland. The difference in elevation within the area is 228 m, the highest point being 365 m and the lowest point 137 m above sea level.

3. Methods 3.1. Basic steps of the sensitivity assessment method The sensitivity assessment method of forest landscapes developed in this study consists of three main phases. The first step is the construction of a sensitivity model for visual landscapes that is based on earlier studies and other existing knowledge. In the second step, a sensitivity index is calculated in GIS and as a result, a sensitivity map is produced. In the third phase, a sensitivity map and other results are evaluated by statistical means and expert evaluation. More specifically, the sensitivity assessment method encompasses the following steps (Fig. 2): (1) Constructing the landscape sensitivity model: selecting the sensitivity criteria and presenting them in the form of a decision tree, which defines the main- and sub-criteria structure for landscape sensitivity. This phase also includes producing weights for the sensitivity criteria according to their relative importance. The determination of weights is based on expert knowledge modeling. (2) Producing a sensitivity index: producing map layers describing the sensitivity criteria in GIS – one map layer for each sub-criterion. After that, the map layers are standardized in an appropriate manner and combined with the help of a sensitivity model to produce a sensitivity index for the study area. Finally, in this phase the sensitivity index is described in the form of a sensitivity map, which covers the study area spatially. (3) Evaluation of results: results are validated by (a) a statistical evaluation of weights by examining the uncertainty of expert responses, and (b) comparing the estimated index values from the sensitivity classification model against expert judgments made directly from maps and in the field. Fig. 1. Case study area and Kainuu and Kuusamo hill area in Finland.

3.2. Sensitivity model

and Kuusamo hill area can provide a basis for the criteria for other landscape provinces as well. The sensitivity assessment method proposed in this study was demonstrated and tested by a case study in which a landscape sensitivity map was produced for one municipality in the landscape province of the Kainuu and Kuusamo hill area. The municipality of Sotkamo was chosen as a case study area. The area covers about 2950 km2 and is located in eastern Finland (64◦ 07 50 N and 28◦ 23 00 E) (Fig. 1). Around 300 km2 (about 10%) of the area of Sotkamo is inland water and the land area is clearly dominated by forests. The municipality is sparsely populated – the number of inhabitants is about 10 700 and the population density is only 4.0 inhabitants per km2 . Settlement is concentrated mainly in two villages, Sotkamo Municipal Center and Vuokatti, which is the center of tourism and sport. Sotkamo Municipal Center is located on a thin

3.2.1. Determining the sensitivity criteria The criteria for landscape sensitivity were determined by expert appraisal. The criteria were developed specifically for the Kainuu and Kuusamo hill area landscape province, bearing in mind that the same criteria could also be used with subtle modification in the other landscape provinces of Finland. A precondition for all criteria was that geographic information that was required to produce values should be found in applicable databases so that there would be practically no fieldwork, and that this information would not change over short periods. Data concerning growing stock was excluded. It was acknowledged that tree stands form an integral part of visual landscapes and play a major role in their attractiveness. However, the focus was on timber production forests, and if information of tree stands was included, the model would have needed constant updating and would have been practically never up to date. In addition, the sensitivity assessment method concerns

R. Store et al. / Landscape and Urban Planning 144 (2015) 128–141

1. Model construcon

Selecng criteria

Assessing criteria weights

2. Sensivity classificaon GIS Producing criteria layers

131

3. Evaluaon

Stascal

222

Combining map layers

Index map

Experts

Fig. 2. Basic steps of the landscape sensitivity assessment method.

the regional scale, in which the unit of analysis is much larger than an average forest compartment. The starting point was the criteria often used in forest landscape design to determine the landscape sensitivity: scenic attractiveness or quality, visibility, and the amount and type of viewers (Bell, 1998; Bell & Apostol, 2008; Visual Landscape Inventory, 1997). These criteria were used as the main criteria, and an expert group selected sub-criteria describing each criterion and meeting the preconditions described in the previous paragraph. In the expert group both the practice and research was represented as well as different disciplines: landscape architecture (including landscape planning and design), forest planning and management (including landscape planning and design), and expertise on the public’s landscape values and preferences. The selection of sub-criteria was based on available geographic information, concepts of Finnish forest landscape design (especially the sub-criteria of visibility, Komulainen, 2010), results of public preference studies (sub-criteria of landscape attractiveness, e.g., Gimblett, 1990; Hammitt, Patterson, & Noe, 1994; Heft & Nasar, 2000; Herzog, 1992; Herzog & Barnes, 1999; Herzog & Bosley, 1992; Kent, 1993; Strumse, 1994), and views of the expert group. The three main criteria in our model were labeled as visibility, potential users and attractiveness of the landscape (Fig. 3). All the main criteria were divided into several sub-criteria. Visibility of landscape was divided into three sub-criteria: (i) summit and slope forests, (ii) edge forests, and (iii) forests that are seen from certain vantage points. The forest landscape typology used here (summit forests, slope forests, edges, valleys, shores) is based on the concept of the landscape structure which is a commonly used approach in Finnish forest landscape design (Komulainen, 2010).

Summit, slope and edge forests (including shores) were chosen as indicators of visibility, because they are the most visible forests in the landscape structure in Finland. Summit forests are those on top of hills, and slope forests, which were divided into steep and gentle slopes, are usually at a lower altitude beneath them. Edge forests included forests next to open areas; lakes, ponds, large rivers, fields and open treeless mires. Because other forest types besides summit, slope and edge forests can be visible, the third sub-criterion described the forest areas that are seen from specific vantage points from which the area is often looked at. While it was not possible to determine the actual amount of viewers, the second main criterion described the amount of potential users of the landscape. Several sub-criteria that give information about the potential use were developed. These were: (i) the type of permanent settlement (population center, village, sparsely inhabited area) and distance from it, (ii) density of second homes, (iii) type of accommodation service (hotels, motels and camping areas) and distance from it, (iv) types of recreational constructions (marked trails, other constructions such as fireplaces, lean-to shelters and information boards) and outdoor activity areas (e.g., golf courses, downhill skiing areas) and distance from them, density of marked trails, and surroundings of certain nature conservation areas, (v) amount of traffic (indicated by road types: highway, main road, regional road, local road) and distance from the road. It was assumed that population centers generate more potential users than sparsely inhabited areas. Likewise, accommodation

Fig. 3. Criteria tree for visual sensitivity of forest landscape to forestry practices.

132

R. Store et al. / Landscape and Urban Planning 144 (2015) 128–141

services hosting a greater number of travelers meant a higher number of potential users or viewers of the nearby area than smaller accommodation. In addition, more frequently used road types create more viewers from the roadsides than little-used roads. In the same way, it was assumed that the higher the number of potential users, the greater the distances the use reaches. The third main criterion, attractiveness of the landscape, was divided into four sub-criteria: (i) already recognized valuable landscapes, (ii) small-scale attractions, (iii) closeness of water, and (iv) variability of the landscape. The already recognized valuable landscapes include areas that have been given a specific label by the national authorities. These were nationally and regionally valuable landscape areas, traditional rural biotopes, areas included in the National Esker Conservation Program, and valuable rock areas. In addition, national landscapes, nationally significant landscape conservation areas, national urban parks, and so on would have been included but none of these types existed in the study area. Areas that belong to nature conservation areas where forest felling is not allowed were not included, and these areas were outside of the scope of this study. Sub-criterion small-sized attractions included small islands, small water elements such as creeks, small rivers, rapids and springs, as well as different types of rock formations such as cliffs, slopes, open rocks, rocky areas and erratic boulders. In addition, small-sized places protected by law, such as natural monuments, protected stones and archeological sites, were included. It was assumed that people often pay attention to these types of elements, and that they are a central part of people’s nature experience and also bring variety. Sub-criterion water systems means here the closeness to larger water areas such as lakes, rivers and ponds. Shorelines were also included in the sub-criterion of edge forests under the main criterion of visibility. However, it was selected as a separate sub-criterion under attractiveness, because water is known to be an important predictor of people’s landscape preferences in natural environments (e.g., Hammitt et al., 1994; Herzog & Barnes, 1999; Herzog & Bosley, 1992; Koch & Jensen, 1988; Shafer & Brush, 1977). Sub-criterion variability mainly describes the variety of different landscape types in a larger area but also to some extent the variability inside a landscape (within-forest views). Variability was assessed based on the variability in topography, variability in site fertility, and the amount of different edge zones (zones between forest and water, forest and treeless mire). Many studies show that the variety of different types of landscapes is a central component in esthetic appreciation, even though the number of studies is limited (see, e.g., Gimblett, 1990; Herzog, 1985, 1992; Kent, 1993; Strumse, 1994). The study by Karjalainen (2000) showed that the existence of different types of forest stands was important to the visitors of recreation areas; they wanted to have even the less preferred forest stands in a larger area. In the study of Axelsson Lindgren and Sorte (1984), people evaluated routes that had more varied landscape types more positively. Some studies have shown that people particularly appreciated the places where the landscape type changed or new information was revealed (Gustke & Hodgson, 1980; Heft & Nasar, 2000). 3.2.2. Weighting and standardizing sensitivity criteria After defining the sensitivity criteria, the priorities of the main and the sub-criteria, as well as the importance of the distances to the targets, were estimated. The weights for the landscape sensitivity criteria were determined by utilizing statistical modeling of expert views (Alho & Kangas, 1997; Alho et al., 1996). An advantage of the statistical approach is that it is based on well-known estimation techniques and statistical inference. The preference assessment technique was based on a variant of analytic hierarchy process (AHP) and its pairwise comparisons

technique (Saaty, 1980). AHP is general theory used to measure subjective preferences and to aggregate the measurements over multiple decision criteria. The pairwise comparisons of the main criteria and the sub-criteria within each main criterion were used to study the priorities of a total of 28 landscape criteria. Pairwise comparisons were asked with the help of a questionnaire, which was sent to landscape experts of the Kainuu and Kuusamo hill area landscape province. In order to achieve a substantial amount of respondents, landscape and forest management organizations of the province area were also asked to name their potential experts. In each pairwise comparison, one criterion was compared to another in the context of landscape sensitivity. Furthermore, same-level sub-criteria were compared to other sub-criteria at same level in the context of main or former main criterion. Using a Bayesian interpretation of the statistical model and Monte Carlo simulation techniques, a posteriori distribution of the priorities of the sensitivity criteria was produced (see Alho & Kangas, 1997). The analysis of the landscape criteria contained responses from multiple experts. Thus, the models were extended to also incorporate the variation between respondents using interval judgment analysis (Leskinen & Kangas, 1998). The influence of the distance from eight different targets on the landscape sensitivity was studied by an expert questionnaire. This included questions on how far (in meters) from a certain target a possible change in the target still affects the landscape. The mean, standard deviation and median of the distance were calculated for each of the eight targets from the experts’ answers. The calculated distances were used in sub-utility functions. The influence of distance was assumed to be linear. Thus, besides the target, e.g., road, the place at which sensitivity is estimated gets the whole priority of the road criterion, and then the priority linearly decreases as the distance to the target increases, and finally, from the distance respondents assessed it is zero. As distance influence on landscape sensitivity depends greatly on the target size, targets have been divided into classes based on their size, e.g., population center, village, sparsely inhabited area. 3.3. Producing sensitivity criteria maps The visual sensitivity of the landscape consisted of three main criteria, which in turn consisted of several sub-criteria, and the sub-criteria also had further sub-criteria (see Section 3.2.1). In the method proposed in this study, the lowest level sub-criteria were described as a form of GIS map layers. To produce these map layers we needed existing datasets with the required spatial cover and adequate accuracy and reliability. The numerical georeferenced materials used in this study were: – Digital elevation model, 25 m × 25 m raster grid, National Land Survey of Finland (NLS). – Topographic thematic database, vector, NLS. – Grid database, 250 m × 250 m grid, Statistics Finland. – Community structure database, 250 m × 250 m grid, Finnish Environment Institute. – Outdoor recreation areas and routes, vector, Municipality of Sotkamo. – National road and street database, vector, Finnish Transport Agency. – Multi-Source National Forest Inventory Data, raster grids, Finnish Forest Research Institute. – Traditional rural biotopes, vector, Centre for Economic Development, Transport and the Environment for Kainuu. – Business register, vector, Statistics Finland. – Valuable rock areas (OIVA), vector, Finnish Environment Institute.

R. Store et al. / Landscape and Urban Planning 144 (2015) 128–141

133

Table 1 Sub-criterion-specific parameters: classes, class limits, weights and distances for value functions. Sub-criteria

Classes and weights

Summit and slope forests Class limit Weight Edge forests Weight Width of edge Forest seen from vantage points Distance

Flat plain Gentle slope Slope < 5◦ Slope 5◦ –10◦ 0.286 0.486 Open mire Water 0.745 1 75 m 75 m Every vantage point had a point-specific weight and observation angle Vantage point located in summit or upper hill area: 3000 m Vantage point located in flat plain: 5000 m Population center Village 1 0.78 1247 767

Settlements Weight Distance Second homes Distance Accommodation services Class limit Weight Distance Outdoor recreation Weight Distance Traffic Weight Distance Water Class limit Weight Distance Variability Weight

265 m radius from cell centers Normal size Capacity > 100 1 560 Marked trail 1 150 Highway 1 194 Lake

Small size Capacity < 100 0.56 314 Other construction 0.96 215 Main road 0.91 194 Pond

1 160 Topography 1

0.88 76 Site fertility 1

– National Esker Conservation Program (OIVA), vector, Finnish Environment Institute. The ArcGIS Desktop 10 program and a cartographic modeling approach were used to produce the required map layers and to combine them with the landscape sensitivity index. The purpose of the GIS analysis was to produce a standardized sensitivity index for the fixed resolution of a 250 m × 250 m grid for each subcriterion. The analysis proceeded on a grid-by-grid basis. Most of the study material was first processed to 25 m × 25 m grid format and in 10 of 12 sub-criteria the areal unit in the calculation phase was 25 m × 25 m cell. In the case of two sub-criteria the calculations were made direct to 250 m × 250 m grids. The basic steps of GIS analysis for individual sub-criterion are described below, and the detailed sub-criterion-specific parameters are presented in Tables 1 and 2. To calculate sub-criterion summit and slope forest at the beginning all cells were classified according to their topographic position. Classification was done using the topographic positioning index method (TPI) (Weiss, 2001). In the calculation phase, the area of different topography classes in each of the 250 m × 250 m grids was Table 2 Factors, calculation distances and weights for the sub-criterion variability. Sub-sub-criteria

Distance

Weight

Topography

Altitude range Standard deviation Area of slope Area of summits Number of summit patches

Inside grid Inside grid 500 m radius 500 m radius 500 m radius

0.167 0.167 0.333 0.167 0.167

Site fertility

Shannon diversity index Shannon diversity index

Inside grid 750 m radius

0.5 0.5

Amount of edge

With open mire With water With open mire With water

Inside grid Inside grid 500 m radius 500 m radius

0.25 0.25 0.25 0.25

Steep slope Slope > 10◦ 0.87 Field 0.399 50 m

Summit 1 Non-edge 0.189

Sparsely inhabited area 0.29 352

Outdoor recreation area 0.8 270 Regional road 0.62 132 River Width ≥ 5 m 0.82 74 Amount of edge 1

Local road 0.53 132 Small river Width < 5 m 0.65 39

calculated (see Fig. 4). Then areas were multiplied by class-specific weights derived from expert judgments and finally weighted scores were summed up for each grid. To recognize edge forests, open mires, waters and fields were separated from the classified study material and expanded by class-specific widths. Pixels of the expanded area which were located in forest were taken as material to calculate the area of each edge class for every grid. The area of non-edge forest also formed a class. In the calculation phase the area of each class was calculated, weighted and finally summed up for each grid. In the case of sub-criterion forest seen from vantage points, a viewshed analysis (Fisher, 1996; Kim, Rana, & Wise, 2004; Lee, 1991) was performed from 10 vantage points using point-specific viewing angles, distances and weights. Vantage points were situated in places known to be used for landscape observation. After that, the Euclidean distance weight was calculated to visible pixels. The effects of distance diminish linearly outwards from vantage points, reaching zero at the point-specific maximum distance. Next, the analysis of separate vantage points was combined and cell-wise scores were summed up for each grid. In the sub-criterion permanent settlements, the overall score for each grid was based on the proximity and type of surrounding settlements. Settlements were classified into three settlement types: densely inhabited areas, villages, and sparsely inhabited rural areas. Each type had a type-specific weight and maximum impact distance. During analysis, a combined average of all settlement weights was calculated for each grid within the maximum distances using the linearly diminishing function. The weight was zero at the maximum distance. Finally, the averages for each grid were standardized to values between zero and one according to the maximum average of all grids. In the case of the sub-criterion second homes, the density of second homes around each 25-m × 25-m cell was calculated. After that, the sum of density values of cells located in forest was calculated for each grid. The sub-criteria accommodation services, outdoor recreation, traffic and closeness to water were calculated according to the same

134

R. Store et al. / Landscape and Urban Planning 144 (2015) 128–141

Fig. 4. Classes describing topographic variety (summit and slope forests) in the case study area and final sensitivity values for that criterion for 250 m grid cells.

basic principles. First, objects in each sub-criterion were classified according to the classification presented in Table 1. After that, the distance weight was calculated to forest cells using Euclidean distance and class-specific parameters. The effects of distance diminished linearly outwards from objects, beside the objects being one, reaching zero at the maximum distance quantified in parameter file (Table 1). Next, the standardized distance weights of separate classes were multiplied by corresponding class weights derived from expert judgments, and the weighted classes were combined on a cell-by-cell basis using the sum function. Finally, the cell-wise scores were summed up for each grid. Valuable landscapes and small-scale attractions were based on already recorded valuable landscapes and small-scale attractions. At the beginning the material described using different kinds of data formats were changed to a 25 m × 25 m cell format. After that, the forest cells were taken into consideration and the sums of valuable landscape pixels and small-scale attractive objects which are located in the forest area were calculated for each grid. The sub-criterion variability consists of three different factors: variability of topography, variability in site fertility and the amount of different edge zones (Table 2). This was calculated directly to 250 m × 250 m grids. Variability of topography is described by altitude range and standard deviation of altitude inside a grid. The area of slopes and summits and the number of summit patches within a 500-m radius of the grid center were calculated with the Fragstats 2.0 program (McGarigal & Marks, 1995). In the edge calculations, edge was defined as the borderline between open space and forest. Edge calculations were also made both inside the grid and within a 500-m radius of the grid center using the Fragstats program. In the case of site fertility, the Shannon diversity index was used to describe the variability in site fertility. Site fertility calculations were made both inside the grid and within a 750-m radius from the grid edges. After the calculations were made, different types of variability were multiplied by their relative importance and scaled between zero and one and finally summed up.

At the end of the calculation phase, for every sub-criterion where the overall score for a grid was a sum of 25 m × 25 m cells, the area was corrected. Area correction was required because of the forest area variation between grids. After that, all 12 sub-criteria were scaled between zero and one, sub-criterion by sub-criterion according to the maximum value of a given sub-criterion. In the next phase, the separate sensitivity maps were weighted according to the sensitivity model introduced in Section 3.2. Weighting was carried out using GIS by multiplying each sensitivity map by its weight coefficient. The separate sensitivity maps were first combined separately for the three main criteria by means of cartographic modeling and overlay analysis in GIS. The final sensitivity map was produced by combining the main criteria in the same manner. The resulting sensitivity index is a georeferenced variable, which is measured continuously in a range of between 0 and 100 so that the worst value is fixed to 0 and the best value is fixed to 100. 3.4. Validation of the model and sensitivity maps Results are validated by examining the uncertainty of expert responses and by comparing the estimated index values from the sensitivity classification model against expert judgments made directly from maps and in the field. The analysis of the priorities of landscape criteria contained responses from multiple experts. The uncertainty of the priorities of the multiple respondents was derived from the variation between responses and the inconsistency of the pairwise comparisons within each response. The priorities of the landscape elements and the uncertainty of the priorities were also studied separately for each expert by calculating an expert-specific coefficient of determination (R2 ) which measures how much the statistical model explains the expert-specific variation. Validation of the sensitivity maps was made by assessing the equivalence of the indexes produced by the model and the expert judgments made from maps. To do this, a set of 250-m × 250-m

R. Store et al. / Landscape and Urban Planning 144 (2015) 128–141

test squares were chosen from Sotkamo municipality. Firstly, all squares in the Sotkamo area were divided evenly into 10 groups, depending on their estimated sensitivity values. From each class a number of squares were then selected randomly. This resulted in a total of 30 test squares. From each test square, a topographic paper map expressed as a ratio as 1:20 000 with detailed and accurate graphical representation of cultural and natural features of the test square and its surroundings was created. The questionnaire, including maps of 30 test squares, was sent to 33 experts familiar with Sotkamo municipality. The experts were first asked to choose the most sensitive place from those 30 test squares (sensitivity value 100), and then to assess the sensitivity values of the other places with respect to it and to each other. Furthermore, the experts were asked to classify the test squares into five classes, from non-sensitive to highly sensitive. The questionnaire was fully completed by 19 respondents. The other equivalence assessment was made between the model and the field judgments. The field assessment was made from 10 plots by five experts familiar with the landscape of Sotkamo municipality. One field plot was assigned as a reference plot having the most sensitive predicted value of the model. The other nine plots were chosen in a way that they were evenly classified to five sensitivity classes based on their predicted value. Nine of the plots, including the reference plot, were the same as in the map judgment. Experts were asked to classify the plots and to give a sensitivity value to them. 4. Results The relative importance of landscape criteria were measured by utilizing a Bayesian approach to the regression technique. The material for priority analysis was acquired by an expert questionnaire. The questionnaire was answered completely by 30 experts from different fields. Based on the pairwise comparisons data, the priorities of the three main criteria; visibility, potential users and attractiveness of the landscape, were almost equal (Table 3). Edge forests, summit and slope forests and vantage points attained the highest priority from the sub-criteria (Table 3). The least valued sub-criterion was traffic. The results showed that it was possible to differentiate the more valued factors from the less valued ones, though the differences between the priorities of the criteria were sometimes small. There was some variation between judges. For example, the standard deviation of the priorities of main criteria of the experts varied

Table 3 Weights for the main criteria and sub-criteria according to the sensitivity model of the Kainuu and Kuusamo hill area and observed standard deviations of the 30 experts. Criteria

Weight

Stdev

Main criteria Visibility Potential users Attractiveness of landscape

0.345 0.325 0.330

0.091 0.110 0.093

Sub-criteria Summit and slope forests Edge forests Vantage points Settlements Second homes Accommodation services Outdoor recreation Traffic Valuable landscapes Small-scale attractions Water system Variability

0.114 0.116 0.115 0.074 0.066 0.066 0.075 0.044 0.112 0.068 0.087 0.064

0.050 0.056 0.043 0.040 0.025 0.034 0.030 0.024 0.051 0.028 0.039 0.023

135

from 0.09 to 0.11, being approximately one-third of the priorities, and from 0.023 (priority was 0.064) to 0.056 (priority was 0.116) for global sub-criteria (Table 3). However, when this variation was taken into account besides expert-specific inconsistency the magnitude of uncertainty of the predicted weights seemed to be tolerable for their further utilization in the predicted relative importance of the landscape sensitivity components. In-depth analysis of this variation was conducted in Haara et al., unpublished data. In the case study, landscape sensitivity index maps were calculated for the area of Sotkamo municipality. The resulting index maps for the three main criteria are shown in Figs. 5–7. The most important areas in terms of visibility are located at the top of the Vuokatti and Naapurivaara hill area (Fig. 5). In general, the highest elevations are most visible and they compose the broadest continuous spatial entities. The most visible areas are centered in the western part of Sotkamo. In the lowland areas, the most visible areas are mainly edge forests, which are located near water bodies and small islands, for example. However, the visible areas in lowlands are small and scattered across small spatial entities. The main criterion visibility is focused on the visibility arising from the natural circumstances and the amount of potential users has only a minimal effect. Potential use is focused on the two population centers of Sotkamo. Vuokatti population center has higher index values and a larger area for intensive potential use than Sotkamo, because recreational activities bring potential users to Vuokatti. In addition, other tourist areas have high values, such as Naapurivaara. Besides housing activities, traffic and recreation cause potential use alongside the routes. For example, main roads and the UKK hiking route stand out on the index map describing the amount of potential users (Fig. 6). However, there are many areas in the municipality of Sotkamo where the main criteria of potential use has only very low values. The Vuokatti and Naapurivaara hill region also stands out in terms of the attractiveness of the landscape (Fig. 7). It is a nationally and regionally valuable landscape area and there are some valuable rock areas as well. In addition, the attractiveness of the landscape gets high indexes for small islands and areas near water bodies which have high variability of natural circumstances, such as site productivity, topography and the amount of edge forest. The three main criteria were combined according to the sensitivity model to achieve the final landscape sensitivity map for the municipality of Sotkamo (Fig. 8). The hill areas of Vuokatti and Naapurivaara stand out clearly as visually highly sensitive areas. The Vuokatti hill area is also the largest continuous spatial high sensitivity entity in the municipality of Sotkamo. Areas alongside water bodies have high sensitivity values, especially when they are located near settlements. In addition, some separate high altitude areas stand out even if they do not have very high potential use. Small islands and the surroundings of major roads also have sensitivity values that are higher than average. In the south-eastern part of Sotkamo, there are visually sensitive areas because of the high values of landscape attractiveness. Both the reliability and validity of the sensitivity model were evaluated. The reliability of the model was assessed by examining two sources of uncertainty, namely (i) the inconsistency of judge-specific evaluations, and (ii) the differences between responses. The model was validated by comparing the equivalence of the predictions derived from the model to the expert judgments made directly from maps and in the field. In the case study, the sensitivity values predicted by the model and the means of expert judgments made from maps were compared, and in addition, the variation between the experts was calculated. The predicted sensitivity values of the model and expert judgments were quite similar (Fig. 9). The latter values

136

R. Store et al. / Landscape and Urban Planning 144 (2015) 128–141

Fig. 5. Visibility main criteria for Sotkamo municipality.

were slightly lower (2.7 units) and the mean difference between sensitivity indexes was 7.2. There was a strong correlation between the means of the judgments and the sensitivity values predicted using the ‘GIS sensitivity model’ (correlation coefficient 0.945, P < 0.000). A paired t-test was used to compare whether the predicted sensitivity values differed from the expert judgments. The null hypothesis that they did not differ was accepted, since the two-tailed P-value was 0.102. There was a variation between the experts, as the standard deviation of the assessed sensitivity values of the test squares was 17.1. In the other equivalence assessment, where the classifications were based on model-predicted values and field assessments, the assessments clearly correlated to each other (correlation coefficient 0.917, P < 0.001). The values of the field judgments were nearer to the model predictions than the judgments based only on maps (Fig. 10). There was no bias of field judgments, whereas map judgments were higher than model predictions (5.2 units), and the mean differences between sensitivity indexes were 11.9 for field and 12.9 for map assessments. A paired t-test was used to compare whether predicted sensitivity values differed from expert field judgments. The null hypothesis that they did not differ was accepted since the p-value was 1.00.

5. Discussion and conclusion The landscape sensitivity assessment method described in this study provides tools for pinpointing the most sensitive forest areas for visual alteration and information about the mutual sensitivities of different forest areas. The method is automated in GIS insofar as it is possible to develop a sensitivity evaluation for large areas or even for a whole country. The generated method was demonstrated and tested through the use of a case study, and the results were evaluated further by forest and environmental experts. The selected experts were particularly aware of the case study area landscapes (Sotkamo municipality), and were very familiar with the variation in visual forest landscapes there. Based on map and field judgments, the method was assessed to be satisfactory for the estimation of landscape sensitivity at a regional level. The selection of the criteria to the model and positioning them in the decision hierarchy is a crucial task of the sensitivity assessment method. This selection and the weights assigned to the selected criteria can have major impact on results. In this study, a substantial amount of effort was directed at the criteria selection phase. The criteria were selected based on existing research and existing methods in an interactive process where experts of forestry and visual

R. Store et al. / Landscape and Urban Planning 144 (2015) 128–141

137

Fig. 6. Amount of potential users in the study area.

landscape and researchers improved the model step by step. As the sensitivity assessment method was a data driven approach, some important criteria could have been missing, thus they could not have been produced from available data. According to the freeform feedback received from experts, the selection of the main criteria succeeded. Furthermore, the details noted by the experts in different map evaluation targets in the validation phase were included quite effectively in the selected sub-criteria. In general, the judges were pleased with the model formulation and the factors used in it. However, there is a certain amount of subjectivity involved in criteria selection, and some new perspectives came up, such as whether the recovery capability of the visual landscape should be taken into consideration. In addition, some experts have criticized the decision to exclude the growing stock information from the criteria set. Expert knowledge modeling was utilized when giving weights to the criteria. As expected, there was a variation between landscape experts. When dealing with such subjective preferences as landscape quality, and in this case landscape sensitivity as well, there are no single judges or groups to follow, but instead the compromise, or an average from a target population, is usually called

for (Hagerhall, 2001). In the analysis of the importance of different sensitivity criteria, sources of uncertainty included the inconsistency of judge-specific evaluations and the differences between responses. These types of uncertainties might also have originated from the misspecification of the sensitivity model’s hierarchy. Further, utilizing a Bayesian interpretation of the statistical models enables the measurement and illustration of the level of uncertainty in a way that is more understandable to decision-makers compared to classical statistical testing (e.g., Kangas & Leskinen, 2005). The sensitivity assessment method is based on utilizing existing georeferenced data and therefore the quality and cover of the material has a big influence on the produced index. Most of the material used in this case study was of good quality and covered the whole country, but some problems did occur. For example, in the case of recreational constructions and outdoor activity areas, the VIRGIS database, which covers the whole country, includes only a part of the required routes and constructions. The missing part of that data was acquired from the municipality. In the case of Sotkamo this was possible, but if the classification method were to be used for larger areas, this might become a problem. In addition, some extra

138

R. Store et al. / Landscape and Urban Planning 144 (2015) 128–141

Fig. 7. Attractiveness of landscape for the municipality of Sotkamo.

checks and manual work would be needed with the sub-criteria of accommodation services. Map layers describing sub-criteria were produced in ArcGIS by utilizing different kinds of calculation methods and spatial analysis functions. The results were evaluated separately for every sub-criterion and methods were developed further when problems occurred. However, a few sub-criteria calculation methods still need some modification. For example, visibility analysis, used with sub-criteria forests that are seen from viewing points, was implemented in a manner where landscape vantage points and viewing angles were selected manually. The content and calculation principles would need some modification if the sensitivity classification is made for larger areas. In future, it is worth studying the possibilities offered by the human-centered viewshed approach, where the purpose is not only to decide if a place is visible or not but also to measure the degree of visibility and simulate what is possible to see by moving through the landscape (Chamberlain & Meitner, 2013). When utilizing the sensitivity index in practical decisionmaking, it is very important to understand how the index is visually described in the form of thematic maps. Therefore, some special attention should be paid to the index classification process. In an

optimal situation, there would be an appropriate number of classes in a thematic map, and furthermore, instead of equal-width grouping, the class limits would have meaningful interpretations which are possible to connect to forest management practices. However, the determination of class limits in such a way is challenging. The evaluation phase showed that the sensitivity values estimated by the sensitivity model were quite similar to the values calculated from the expert map and field evaluations. During the case study, both the single criteria and the final sensitivity index were subjects of continuous evaluation and improvements. According to the opinion of the experts, some targets deserve higher sensitivity values on the final sensitivity map. In this context, the opinions were partly discordant. However, with some improvements there was a clear consensus. For example, according to the experts, the sensitivity values of shorelines were too low and therefore the weights of shorelines were increased to the final index calculation. This was done by making minor revisions to the calculation process and by gradually increasing the weight of shoreline criteria until the change was noticeable on the index map. Certain targets located at high altitude where there was no permanent settlement and which had minor recreation use had relatively high sensitivity values. According to some experts, sensitivity values for

R. Store et al. / Landscape and Urban Planning 144 (2015) 128–141

Fig. 8. Landscape sensitivity index for the municipality of Sotkamo.

Fig. 9. Sensitivity values of the 30 test squares estimated by the model and assessed by experts, and their absolute differences, from Sotkamo municipality.

139

140

R. Store et al. / Landscape and Urban Planning 144 (2015) 128–141

Fig. 10. Model estimated sensitivity values, and expert assessments from maps and in the field of eight test squares from Sotkamo municipality.

those areas were too high. This could be corrected by increasing the weight of the main criterion “amount of potential users” in relation to the main criteria “visibility” and “attractiveness of the landscape”, for example. However, the need for this correction and the best way of accomplishing it are subjects of future research. During the pilot evaluation, some methodological improvement needs were noted. For example, some of the decision criteria may not be preferentially independent, which means that a contribution of one single criterion to sensitivity cannot be estimated without taking into account the level of some other decision criteria. We had not taken these kinds of potential dependencies into account, but it would be possible to incorporate them into the sensitivity assessment method by adding a new map layer, including an interaction correction term for previously specified situations, for example. However, the interactions that are worth including in the model and how big the interaction correction should be are aims for future research. The area of the basic calculation entity (250 m × 250 m cell) is such a size that it usually includes some variation in sensitivity and the calculated index value represents a mean value of the sensitivity. Therefore, the calculation resolution averages the sensitivity, and targets that are small in size with very low or high sensitivity values are not necessarily visible in the results. The developed sensitivity index can be used as supporting data in forestry planning and political decision-making to increase the cost-efficiency of landscape management practices. Results can be used by forest owners, forestry organizations, the tourist industry and political decision-makers, among others. The indexes are in such a format that they can be readily integrated into the planning systems of forestry and environmental organizations. The landscape sensitivity index can recognize the most important and most sensitive areas for the region in question, which in turn helps to direct the forest-stand level landscape activities (such as detailed landscape inventories and planning and implementing fellings which pay attention to the landscape) to the most sensitive areas. On the other hand, areas that are resilient to change can either be excluded from landscape management practices or only minor investments can be directed at them. Thus, by using the sensitivity index, a higher total utility can be attained with the same economic investment, i.e., the same visual outcome can be realized with a lower loss of income from timber sales. Sensitivity information can also promote the trade of landscape values while the forest owner can receive information on the visually sensitive areas

on his or her land. In addition, when the visually sensitive areas can be recognized beforehand, potential conflicts between timber production and other use forms (recreation, tourism, etc.) can possibly be avoided. If the sensitivity index is produced for the whole of Finland in the future, an appropriate strategy would be to implement it in one landscape province at a time. The first step would then be to generate the index for the whole of the Kainuu and Kuusamo landscape province where the sensitivity model already exists. In this connection, the known problems with the study material and calculation procedures should be identified, and some tests regarding simplifying GIS calculations and decentralized calculation techniques across several more powerful computers should be performed. Acknowledgements We would like to express our thanks to the co-operation partners who helped us with the material and provided us with their practical expertise. The funding from the Ministry of Agriculture and Forestry of Finland (project 311 179 and 311 296) and from the Finnish Forest Research Institute (Metla) made the project possible. Additionally the authors are very grateful to those numerous experts who gave their time to assist in the selection of the criteria and weights to the sensitivity model and to evaluate the results. References Alho, J. M., & Kangas, J. (1997). Analyzing uncertainties in experts’ opinions of forest plan performance. Forest Science, 43(4), 521–528. Alho, J., Kangas, J., & Kolehmainen, O. (1996). Uncertainty in expert predictions of the ecological consequences of forest plans. Applied Statistics, 45(1), 1–14. http://dx.doi.org/10.2307/2986218 Axelsson Lindgren, C., & Sorte, G. J. (1984). (Visually distinguishable vegetation characteristics like undergrowth: A case study from Järavallsskogen) Visuellt urskiljbara vegetationskaraktärer som planunderlag: Exemplet Järavallsskogen. Sveriges lantbruksuniversitet, Institutionen för landskapsplanering. Stencil 84:3. Alnarp. Bell, S. (1998). Forest design planning. A guide to good practice. Forestry practice guide. Forestry Authority. Forestry Commission, iv, 76 p. Available at: http:// www.forestry.gov.uk/pdf/fdp.pdf/$FILE/fdp.pdf Bell, S., & Apostol, D. (2008). Designing sustainable forest landscapes. New York: Taylor and Francis. Carver, S. (1991). Integrating multi-criteria evaluation with geographical information systems. International Journal of Geographical Information Systems, 5(3), 321–339. http://dx.doi.org/10.1080/02693799108927858 Chamberlain, B., & Meitner, M. (2013). A Route-based visibility analysis for landscape management. Landscape and Urban Planning, 111, 13–24. http://dx. doi.org/10.1016/j.landurbplan.2012.12.004

R. Store et al. / Landscape and Urban Planning 144 (2015) 128–141 Finnish Statistical Yearbook of Forestry. (2012). Official Statistics of Finland. Agriculture, Forestry and Fishery. Fisher, P. (1996). Extending the applicability of viewsheds in landscape planning. Photogrammetric Engineering & Remote Sensing, 62(11), 1297–1302. (1994). Forest Landscape Design Guidelines ISBN: 0-11-710325-X. The Forestry Authority. Forestry Commission. HMSO. Gimblett, H. R. (1990). Environmental cognition: The prediction of preference in rural Indiana. Journal of Architectural and Planning Research, 7(3), 222–324. Gustke, L. K., & Hodgson, R. W. (1980). Rate of travel along an interpretive trail. The effect of an environmental discontinuity. Environment and Behavior, 12(1), 53–63. http://dx.doi.org/10.1177/0013916580121004 Hagerhall, C. M. (2001). Consensus in landscape preference judgements. Journal of Environmental Psychology, 21, 83–92. http://dx.doi.org/10.1006/jevp.2000.0186 Hammitt, W. E., Patterson, M. E., & Noe, F. P. (1994). Identifying and predicting visual preference of southern Appalachian forest recreation vistas. Landscape and Urban Planning, 29, 171–183. http://dx.doi.org/10.1016/01692046(94)90026-4 Hänninen, H., & Kurttila, M. (2007). (Impressiveness and development needs of forest environment diversity guidance)) Metsäluonnon monimuotoisuusneuvonnan vaikuttavuus ja kehittämistarpeet. Metlan työraportteja/Working Papers of the Finnish Forest Research Institute 57., 72 p. Available at: http://www.metla.fi/ julkaisut/workingpapers/2007/mwp057.pdf Harvey, A. M. (2001). Coupling between hillslopes and channels in upland fluvial systems: Implications for landscape sensitivity, illustrated from the Howgill Fells, northwest England. Catena, 42, 225–250. http://dx.doi.org/10.1016/ S0341-8162(00)00139-9 Heft, H., & Nasar, J. L. (2000). Evaluating environmental scenes using dynamic versus static displays. Environment and Behavior, 32(3), 301–322. http://dx.doi. org/10.1177/0013916500323001 Herzog, T. R. (1985). A cognitive analysis of preference for waterscapes. Journal of Environmental Psychology, 5(3), 225–241. http://dx.doi.org/10.1016/S02724944(85)80024-4 Herzog, T. R. (1992). A cognitive analysis of preference for urban spaces. Journal of Environmental Psychology, 12(3), 237–248. http://dx.doi.org/10.1016/S02724944(05)80138-0 Herzog, T. R., & Barnes, G. J. (1999). Tranquillity and preference revisited. Journal of Environmental Psychology, 19(2), 171–181. http://dx.doi.org/10.1006/jevp. 1998.0109 Herzog, T. R., & Bosley, P. J. (1992). Tranquility and preference as affective qualities of natural environments. Journal of Environmental Psychology, 12(2), 115–127. http://dx.doi.org/10.1016/S0272-4944(05)80064-7 Jankowski, P. (1995). Integrating geographical information systems and multiple criteria decision-making methods. International Journal of Geographical Information Systems, 9(3), 251–273. http://dx.doi.org/10.1080/ 02693799508902036 Kangas, J., Kangas, A., Leskinen, P., & Pykäläinen, J. (2001). MCDM methods in strategic planning of forestry on state-owned lands in Finland: Applications and experiences. Journal of Multi-Criteria Decision Analysis, 10(5), 257–271. http://dx.doi.org/10.1002/mcda.306 Kangas, J., Karsikko, J., Laasonen, L., & Pukkala, T. (1993). A method for estimating the suitability function of wildlife habitat for forest planning on the basis of expertise. Silva Fennica, 27(4), 259–268. http://dx.doi.org/10.14214/sf.a15680 Kangas, J., & Leskinen, P. (2005). Modelling ecological expertise for forest planning calculations – Rationale, examples, and pitfalls. Journal of Environmental Management, 76(2), 125–133. http://dx.doi.org/10.1016/j.jenvman.2005.01.011 Kangas, J., Store, R., Leskinen, P., & Mehtätalo, L. (2000). Improving the quality of landscape ecological forest planning by utilising advanced decision-support tools. Forest Ecology and Management, 132(2–3), 157–171. http://dx.doi.org/10. 1016/S0378-1127(99)00221-2 Karjalainen, E. (2000). Metsänhoitovaihtoehtojen arvostus ulkoilualueilla [Preferred forest management practices in recreation areas)]. Metsäntutkimuslaitoksen tiedonantoja, 776, 123–136. Karppinen, H. (1998a). Objectives of non-industrial private forest owners: Differences and future trends in southern and northern Finland. Journal of Forest Economics, 4(2), 147–173.

141

Karppinen, H. (1998b). Values and objectives of non-industrial private forest owners in Finland. Silva Fennica, 32(1), 43–59. Kent, R. L. (1993). Determining scenic quality along highways: A cognitive approach. Landscape and Urban Planning, 27(1), 29–45. http://dx.doi.org/10. 1016/0169-2046(93)90026-A Kim, Y., Rana, S., & Wise, S. (2004). Exploring multiple viewshed analysis using terrain features and optimisation techniques. Computers and Geosciences, 30, 1019–1032. http://dx.doi.org/10.1016/j.cageo.2004.07.008 Koch, N. E., & Jensen, F. S. (1988). (Forest recreation in Denmark. Part IV: The preferences of the population) Skovens friluftsfunktion i Danmark. IV del. Befolkningens ønsker til skovens og det åbne lands udformning. Saertryk af Det forstlige Forsøgsvaesen i Danmark. Komulainen, M. (2010). Forestscapes – A forest landscape typology as an integrated planning process tool. Dissertationes Forestales, 98, 196 s. Available at: http://www.metla.fi/dissertationes/df98.htm Landscape Aesthetics. (1995). A handbook for scenery management. United States Department of Agriculture. Forest Service. Agriculture Handbook Number 701. Lee, J. (1991). Analyses of visibility sites on topographic surfaces. International Journal of Geographical Information Systems, 5(4), 413–429. http://dx.doi.org/ 10.1080/02693799108927866 Leskinen, P., & Kangas, J. (1998). Analysing uncertainties of interval judgement data in multiple-criteria evaluation of forest plans. Silva Fennica, 32(4), 363–372. Maisemanhoito. (1992). (Report I of the wor-king g-roup on landscape areas) Maisema-aluetyöryhmän mietintö I. Ympäristöministeriö, Raportti 66. McGarigal, K., & Marks, B. J. (1995). FRAGSTATS: Spatial pattern analysis program for quantifying structure. Gen. Tech. Rep. PNW-GTR-351. Portland, OR: U.S. Department of Agriculture, Forest Service, Pacific Northwest Research Station., 122 p. Nijkamp, P., Rietveld, P., & Voogd, H. (1990). Multicriteria evaluation in physical planning. Contributions to economic analysis. North-Holland. Richardson, E. A., Mitchell, R., Hartig, T., de Vries, S., Astell-Burt, T., & Frumkin, H. (2012). Green cities and health: A question of scale? Journal of Epidemiology and Community Health, 66(2), 160–165. http://dx.doi.org/10.1136/jech.2011. 137240 Saaty, T. L. (1980). The analytic hierarchy process. New York: McGraw-Hill. Shafer, E. L., & Brush, R. O. (1977). How to measure preferences of natural landscapes. Landscape Planning, 4(3), 237–256. http://dx.doi.org/10.1016/ 0304-3924(77)90027-2 Siddiqui, M., Everett, J., & Vieux, B. (1996). Landfill siting using geographic information systems: A demonstration. Journal of Environmental Engineering, 122(6), 515–523. http://dx.doi.org/10.1061/(ASCE)07339372(1996)122:6(515) Store, R. (2009). Sustainable locating of different forest uses. Land Use Policy, 26, 610–618. http://dx.doi.org/10.1016/j.landusepol.2008.08.013 Store, R., & Kangas, J. (2001). Integrating spatial multi-criteria evaluation and expert knowledge for GIS-based habitat suitability modelling. Landscape and Urban Planning, 55(2), 79–93. http://dx.doi.org/10.1016/S01692046(01)00120-7 Strumse, E. (1994). Environmental attributes and the prediction of visual preferences for agrarian landscapes in Western Norway. Journal of Environmental Psychology, 14(4), 293–303. http://dx.doi.org/10.1016/S02724944(05)80220-8 van Dillen, S. M. E., de Vries, S., Groenewegen, P., & Spreeuwenberg, P. (2012). Greenspace in urban neighbourhoods and residents’ health: Adding quality to quantity. Journal of Epidemiology and Community Health, 66(6), e8. http://dx. doi.org/10.1136/jech.2009.104695 (1997). Visual Landscape Inventory: Procedures & Standards Manual. BC Ministry of Forests, Forest Practices Branch for the Culture Task Branch, Resources Inventory Committee. British Columbia Forest Service, Forest Recreation., 71 p. Available at: http://www.ilmb.gov.bc.ca/risc/pubs/culture/visual (2013). Visual Resource Management. US Department of the Interior. Bureau of Land Management. Retrieved from: http://www.blm.gov/wo/st/en/prog/Recreation/ recreation national/RMS.html Weiss, A. D. (2001). Topographic position and landforms analysis. Conference poster. In ESRI International User Conference San Diego, CA, USA.