Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Toward human-centered approaches in landscape planning : exploring geospatial and visualization techniques… Chamberlain, Brent Charles 2011

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata


24-ubc_2011_fall_chamberlain_brent.pdf [ 7MB ]
JSON: 24-1.0072308.json
JSON-LD: 24-1.0072308-ld.json
RDF/XML (Pretty): 24-1.0072308-rdf.xml
RDF/JSON: 24-1.0072308-rdf.json
Turtle: 24-1.0072308-turtle.txt
N-Triples: 24-1.0072308-rdf-ntriples.txt
Original Record: 24-1.0072308-source.json
Full Text

Full Text

TOWARD HUMAN-CENTERED APPROACHES IN LANDSCAPE PLANNING: EXPLORING GEOSPATIAL AND VISUALIZATION TECHNIQUES FOR THE MANAGEMENT OF FOREST AESTHETICS   by   Brent Charles Chamberlain   M.Sc., Forestry, University of British Columbia 2007 B.B.A., Business Administration, Pacific Lutheran University, 2002 B.A., Computer Science, Pacific Lutheran University, 2002    A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY   in    THE FACULTY OF GRADUATE STUDIES   (Forestry)    THE UNIVERSITY OF BRITISH COLUMBIA  (Vancouver)   October 2011   © Brent Charles Chamberlain, 2011  ii Abstract As sustainable forest management continues to influence forest planning and the balance of social, economic and ecological goals is evaluated, managers must find ways to ensure that forests provide for a variety of products and services for people and the environment. This dissertation focuses on finding new ways to more effectively manage forest aesthetics, including the development of methods that are human-centered; methods that are based on human perception and empirical research. The research explores a variety of geospatial and visualization technologies designed to aid managers in the process of planning for the conservation of visual resources. The first development is based on findings from empirical research that present a quantifiable expression of how the shape of harvest blocks can influence preferences. A study was conducted which required individuals to rate 52 near photo-realistic images which simulated different possible harvests in a forested landscape. Three difference shape characteristics were controlled for: geometric primitive (atomic shape), complexity (irregularity), and aspect ratio (elongation). The results indicate that geometric primitive has the largest effect on preferences followed closely by complexity. The research shows that complex, rounded- edged circular shapes are most preferable, and regardless of shape, moderate levels of complexity dramatically increase preferences. The second development is the Human-Centered Viewshed (HCV). Viewsheds are used in landscape management, but may lack important landform detail. The HCV combines an efficient algorithm, XDraw, with three characteristics of landform to provide a measure space an object occupies within an individual’s field of view. The three characteristics involve the effect of slope, aspect and distance from an observer to a target location. The output is a simple, discrete 2D image that supplies a degree of visibility for each location in space which can be used to determine how an individual may experience the landform as they move through the landscape.   iii Applications of these discoveries on the management of forest aesthetics are presented, followed by a discussion of management trade-offs with ecology. The research in this dissertation can improve the current visual resource management process by providing planners with new information to help them more effectively manage forest aesthetics.  iv Preface This dissertation developed from my passion and interest in caring for our forests and by using technology to make the management of these ecosystems better for both people and the world we live in. So often, much of what society experiences about forests is through the exposure we have from living near, traveling through and recreating within them. This research represents new methods and insights for managing forests in places of high scenic or cultural value while also being sensitive to the ecological processes. The dissertation consists of six chapters. The introduction and conclusion were written wholly by me. No appendices are provided, but raw data are available upon request. A version of Chapter 2 has been submitted for publication, except section 2.5 was left out of the submission as it was completed as an extension to the empirical research. I wrote the manuscript and developed the survey interface used for this research. I also took the lead role in supervising the development of the 3D images. Mr. Oliver Lane, M.F., provided technical support in rendering the images and facilitated the computer survey with my supervision. I supervised Ms. Kristen Lambke who helped compile statistics using Fragstats for the analysis discussed in section 2.5. Dr. Michael Meitner, who is a second author on the paper, helped instruct the experimental design and statistical analysis of the data. He also provided a substantial review of the manuscript. The research required the approval of UBC’s Behavioural Research Ethics Board and was approved with Certificate Number H10-01209. A version of Chapter 3 has been submitted for publication. I wrote the manuscript and I am the sole author of the algorithm and software discussed in the chapter. Dr. Michael Meitner is the second author on the publication, and provided a substantial review of the manuscript. Mr. Craig Chamberlain provided help with the illustrations and both he and Dr. David Tait helped with the preparation of mathematical notation. Chapter 4 was written entirely by me, including all analysis and figures, except where and where Alec Patrizio helped in the creation of figures in section 4.3.1.  v Table of Contents Abstract .................................................................................................................................... ii Preface ..................................................................................................................................... iv Table of Contents .................................................................................................................... v List of Tables ........................................................................................................................ viii List of Figures ......................................................................................................................... ix List of Abbreviations ........................................................................................................... xiii Acknowledgements .............................................................................................................. xiv Dedication ............................................................................................................................. xvi Chapter 1: Introduction ......................................................................................................... 1 1.1 Managing Forested Scenic Landscapes ................................................................................ 2 1.2 Integrating Geographic Information Systems for Visual Resource Management ................ 6 1.3 Forests and Aesthetics in British Columbia .......................................................................... 8 1.4 Research Theme, Purpose and Goals .................................................................................. 11 1.5 Research Structure .............................................................................................................. 13 Chapter 2: Quantifying the Effects of Harvest Block Design on Aesthetic Preferences 16 2.1 Introduction ......................................................................................................................... 16 2.2 Materials and Methods ........................................................................................................ 22 2.3 Results ................................................................................................................................. 28 2.4 Discussion ........................................................................................................................... 35 2.4.1 The Influence of Geometric Primitive ............................................................................ 35 2.4.2 The Influence of Complexity .......................................................................................... 36 2.4.3 The Influence of Aspect Ratio ........................................................................................ 37 2.4.4 The Influence of the Interaction between Geometric Primitive and Complexity ........... 38 2.4.5 Implications for Forest Management .............................................................................. 38 2.5 Using Shape Indicators to Measure Aesthetics ................................................................... 40  vi Data Preparation ..................................................................................................... 41 2.6 Conclusions ......................................................................................................................... 48 Chapter 3: The Human-Centered Viewshed: An Efficient Algorithm to Evaluate Landscape Visual Magnitude............................................................................................... 50 3.1 Introduction ......................................................................................................................... 50 3.1.1 Visibility Analysis and Algorithms ................................................................................ 52 3.1.2 Visual Magnitude ........................................................................................................... 53 3.2 Methods ............................................................................................................................... 54 3.2.1 Notation and Assumptions .............................................................................................. 54 3.2.2 Visibility and Viewshed ................................................................................................. 55 3.2.3 Visual Magnitude: The Human-Centered Dimension .................................................... 57 3.2.4 Details of the Study Area ................................................................................................ 61 3.2.5 The Process ..................................................................................................................... 62 3.3 Results ................................................................................................................................. 63 3.3.1 Outputs ............................................................................................................................ 64 3.3.2 Computational Examples ................................................................................................ 66 3.4 Discussion ........................................................................................................................... 67 3.5 Conclusion .......................................................................................................................... 69 Chapter 4: Geospatial Techniques and Considerations for the Management of Visually Sensitive Areas in British Columbia ................................................................................... 71 4.1 The Visual Landscape Inventory ........................................................................................ 73 4.1.1 Using Visual Magnitude for Visual Landscape Inventory ............................................. 76 4.1.2 Provincial Level Analysis and Regional Comparison .................................................... 83 4.2 The Management and Assessment of Harvests in Visually Sensitive Areas ...................... 90 4.2.1 Effectiveness Evaluation ................................................................................................ 92 4.2.2 Automating the Calculation of the Perspective Amount of Alteration ........................... 94 How the Calculation is Completed using Visual Magnitude ................................. 94 How the Calculation is Completed Using 3D Simulations .................................... 97 Validating Visual Magnitude: a Comparison with 3D Simulations ....................... 99 4.3 Considerations and Problems of Using Single Viewpoint Evaluations ............................ 105 4.3.1 Concerns with Viewshed Analysis ............................................................................... 107 4.3.2 Significance and Subsequent Weighting of a Viewpoint ............................................. 114  vii 4.3.3 Representing a Viewpoint as a Spatial Distribution ..................................................... 115 4.4 Conclusions ....................................................................................................................... 117 Chapter 5: Trade-offs and Limitations for Visual Resource Management .................. 118 5.1 Important Aesthetic and Ecological Considerations for Forest Management ................... 118 5.1.1 Variable Retention: Ecological and Aesthetic Implications on Forestry ...................... 120 5.1.2 Implications for Emulating Natural Forest Disturbance on Aesthetics ........................ 123 5.1.3 Ecological Considerations of Using Visual Magnitude to Guide Planning .................. 127 5.1.4 Conclusions .................................................................................................................. 129 Chapter 6: Conclusion ........................................................................................................ 131 6.1 Summary of Conclusions .................................................................................................. 131 6.2 Future Directions .............................................................................................................. 135 6.3 Concluding Remarks ......................................................................................................... 137 References ............................................................................................................................ 138 Appendices ........................................................................................................................... 152 Appendix A: Survey Questionnaire ............................................................................................... 152  viii List of Tables Table 2.1 Study design matrix for harvest designs ................................................................. 24 Table 2.2 The list of variables used in the development of the harvest designs. .................... 25 Table 2.3 ANOVA results of harvest attributes on preference ratings for the 52 harvest designs and 40 subjects surveyed ........................................................................................... 29 Table 2.4 Correlation of mean ratings and spatial metrics of 52 rendered scenes .................. 43 Table 2.5 Correlation coefficients of complexity metrics and mean ratings categorized by geometric primitive ................................................................................................................. 44 Table 3.1 Comparison of results for XDraw and Direct methods. Results are based on four different spatial resolutions (12 m – 100 m) and a varying number of viewpoints. Abbreviations are: t = time, tn = time normalized (time as a function of resolution), vp = viewpoint................................................................................................................................. 67 Table 4.1 Phases of visual landscape management process (adapted from BC Ministry of Forests, 1997b) ........................................................................................................................ 72 Table 4.2 Comparison of the regional amount of Visual Magnitude per viewpoint per amount of visible landscape ................................................................................................................. 87 Table 4.3 Visual Quality Class definitions ............................................................................. 91 Table 4.4 Results of percent visible alteration comparisons between 3D simulation and Visual Magnitude using 30 m, 10 m and 3 m resolution ...................................................... 103   ix List of Figures Figure 1.1 Example harvest where visual quality is regulated. Image obtained using Google StreetView, © Google 2011 ...................................................................................................... 9 Figure 1.2 Example harvest where visual quality is not regulated ......................................... 10 Figure 1.3 Visually Sensitive Areas in British Columbia ....................................................... 11 Figure 2.1 Photo-realistic image of the landscape used to survey individuals regarding harvest design preferences, without any harvest activity. ....................................................... 23 Figure 2.2 Illustrations of the three harvest design variables that were examined in the study ................................................................................................................................................. 24 Figure 2.3 Nine examples of images used in the study with their categories of geometric primitive (p), complexity (c) and aspect ratio (a), as well as the mean ratings ...................... 26 Figure 2.4 Relationship of mean ratings and geometric primitive showing the mean and 95% confidence interval. ................................................................................................................. 30 Figure 2.5 Relationship of mean ratings with complexity ...................................................... 31 Figure 2.6 Relationship of mean ratings with aspect ratio...................................................... 32 Figure 2.7 Comparison of the three levels of complexity for each geometric primitive type showing the mean and 95% confidence interval. .................................................................... 33 Figure 2.8 Comparison of the three characteristics of shapes used in the study as rated on their importance to design preference. The table shows the mean ratings with a 95% confidence interval .................................................................................................................. 34 Figure 2.9 Example harvest design from original survey ....................................................... 41 Figure 2.10 Example harvest design converted from survey into binary representation of patch for Fragstats ................................................................................................................... 42  x Figure 2.11 Hypothetical delineation of an example harvest design, to demonstrate the possible differences in edge classification. ............................................................................. 47 Figure 3.1 Depiction of the Direct method for visibility analysis. The black line shows the projection of a view vector in the xy-plane The cells labeled 1 to 7 are used to calculate the visibility of cell 7 from pv. ...................................................................................................... 56 Figure 3.2 Depiction of XDraw method which calculates visibility by concentric rings (shown by the numbers located on the left and bottom cells. E is visible if its’ vs(E) is greater than the interpolated value of the previous ring. pv  is the observer point and pt is the target location. ................................................................................................................................... 57 Figure 3.3 Diagrammatic explanation of β. This is a 2D sideview of a single terrain cell with its center at pt. This plane shows the surface slope of the cell. ............................................... 59 Figure 3.4 Diagrammatic explanation of θ. This is a 2D sideview of the terrain grid cell, where the cell has a direction with pt at its center. ................................................................. 59 Figure 3.5 Inside Passage study area for the analysis of the HCV. The cruise route is shown in white, with the oceans in blue and a hillshade to depict the terrain. .................................. 62 Figure 3.6 Example of a larger portion of the Inside Passage in 3D. The route is shown in bold white. The thin white line and thin black line represent Figure 3.7A and B respectively. The gradient of orange represents VM values, with high values as dark orange. The output cell size is 25 m. Gray is not visible. ...................................................................................... 64 Figure 3.7 Example of Visual Magnitude demonstrated for a small area of the Inside Passage (A) and objective VM values (B). The route is shown in bold white. The orange gradient represents VM values with high values in dark orange. The output cell size is 25 m. The black outline in A is shown at a larger scale in B. Locations not visible are replaced with a hillshade. ................................................................................................................................. 65 Figure 3.8 Example of standard viewshed analysis conducted for the entire route, with hill shading to provide terrain perspective. Light orange values represent visible areas. The route is depicted in white. ................................................................................................................ 66  xi Figure 4.1 Visual Magnitude as calculated from a single viewpoint. All visible areas are shown in shade of orange, what remains is either ocean or hillshade. ................................... 78 Figure 4.2 A composite Visual Magnitude represented by 182 viewpoints along a route (purple) with the single viewpoint used in Figure 4.1 represented by a transparent white point. All visible areas are shown in shade of orange, what remains is either ocean or hillshade. ................................................................................................................................. 79 Figure 4.3 Example showing visible areas produced from the HCV not included as a visually sensitive area in the Visual Landscape Inventory. Image in A (obtained from Google Maps, © Google 2011) shows a planimetric view of a single VSU, which is shown in perspective view in B (obtained from Google Earth, © Google 2011). In C, that same VSU is shown with other VSUs for context. In C, you can see the visible area to the left of the VSU shown in all three images, where an alteration also exists. .................................................................................. 81 Figure 4.4 Visual Magnitude analysis for all freeways and highways in British Columbia. The darker the orange, the higher the VM. The highways (not shown) are within darkest orange. ..................................................................................................................................... 85 Figure 4.5 Regional comparison of Visual Magnitude from all highways and freeways in BC. Legend values represent levels of the amount of Visual Magnitude per viewpoint per area visible from the highway (black line). Regional names are available in Table 4.2. ............... 86 Figure 4.6 Example scene of Highway 97 through the Cariboo Region. Image obtained using Google Earth, © Google 2011 ................................................................................................ 88 Figure 4.7 Example scene of the Sunshine Coast Highway in the Sunshine Coast Region. Image obtained using Google Earth, © Google 2011 ............................................................. 89 Figure 4.8 Example surface and Visual Magnitude values for calculating perspective visible alteration ................................................................................................................................. 95 Figure 4.9 GIS data input information for calculating percent visible alteration ................... 96 Figure 4.10  Forested landscape in 2D showing visible harvested area ................................. 97  xii Figure 4.11 Example 3D harvest simulation ........................................................................... 98 Figure 4.12 Example 3D harvest simulation used to calculate percent alteration .................. 98 Figure 4.13 Example viewing angle created for comparison ............................................... 100 Figure 4.14 Example conversion of harvest from vector to raster ........................................ 102 Figure 4.15 Scatter plot of all 52 images comparing values calculated by the VNS 3d simulation and Visual Magnitude at 30 m, 10 m and 3 m resolution. The black line shows a perfect correlation with exactly similar values , whereas the other trendlines show the correlation with their offset values. ...................................................................................... 104 Figure 4.16 Example terrain used  for analysis of the influence of bordering cells on different viewshed results. The left shows a hillshade of the terrain. The right shows DEM values for a close-up of the hill from which the viewpoint is on top. ...................................................... 108 Figure 4.17 Comparison of (1) custom-coded Direct method with industry standard. Dark grey are areas of agreement. Black are areas only visible using ArcGIS®, and light gray is the area only visible from the comparison. ................................................................................. 109 Figure 4.18 Comparison of (2) custom-coded Direct method with industry standard. Dark grey are areas of agreement. Black are areas only visible using ArcGIS®, and light gray is the area only visible from the comparison. ................................................................................. 110 Figure 4.19 Comparison of (3) XDraw method with industry standard. Dark grey are areas of agreement. Black are areas only visible using ArcGIS®, and light gray is the area only visible from the comparison. ............................................................................................................ 111 Figure 4.20 Effects of bordering cells on high resolution viewshed analysis ...................... 113 Figure 5.1 Relationship between visual quality and sustainability with examples (adapted from Sheppard, 2000). .......................................................................................................... 125   xiii List of Abbreviations BC   British Columbia DTM   Digital Terrain Model DEM   Digital Elevation Model ENFD   Emulating Natural Forest Disturbance HCV   Human-Centered Viewshed GIS   Geographical Information System GDP   Gross Domestic Product US   United States VLI   Visual Landscape Inventory VM   Visual Magnitude VNS   Visual Nature Studio® VQC   Visual Quality Class VQO   Visual Quality Objective VRM   Visual Resource Management VSC   Visual Sensitivity Class VSU   Visual Sensitivity Unit  xiv Acknowledgements First, thank you the National Science and Engineering Research Council for the financial support of this research. First, as part of Dr. Meitner’s Discovery Grant which provided support for research assistants; and second, for supporting my Doctoral research through the Michael Smith Foreign Studies Supplement and Canada Graduate Scholarship. Second, it is with sincere appreciation and deepest gratitude that I express my thanks to so many who have sacrificed, encouraged, guided, and prayed for me throughout my life. You are the friends, colleagues, neighbors, teachers and family that have helped me become the person I am today. I never could have done this without your support. Colleagues and friends: To Dr. Mike Meitner thank you for your mentorship, friendship, time, patience, generosity and encouragement. I am grateful for your invitation and willingness to freely let me explore my interests and make mistakes along the way. As you have done for me, I hope to pass these gifts along to students of my own one day. Dr. Brad Seely for your teaching and great questions. You have allowed me to tread waters I dared not before I came to UBC. Thank you for opening your world and knowledge to me. Dr. Sarah Gergel for your enthusiasm, extremely thorough feedback and willingness to delve into the new territory of aesthetics. I am looking forward to our research endeavors together. Dr. William Evans for helping me get started in a direction of research that has had a significant influence on the trajectory of my research career.  Also, a special thank you to my mentors of the past from Pacific Lutheran University: Dr. Eli Berniker, Dr. Chung-Shing Lee and Dr. Sam Chung. I could write at length thanking you for your vision and unyielding support, but if I know you three, the mere submission of this dissertation is testament enough to all your efforts and encouragement. Also, to so many in Brazil who have opened their hearts and homes to Andrea and I. Thank you to Dr. Carlos Ribeiro for your amazing support and contagious spirit. I am looking forward to continuing our research together. To Dr. James Griffith and Bete Griffith, your home has become a symbol of love, acceptance, sanctuary and many memorable moments I will cherish until the day I die. Mr. Leo Perciano, Elke Streit and Pedro, thank you for adopting us into your family and home. And to Ginia Bontempo, for welcoming us into a community of people that make us feel like we will  xv always have a home in Viçosa. To my labmates: Julian for your energy, encouragement and engaging discussions. Oliver for the hours and hours of work you put into creating shapes, setting up and running the surveys, and for your patience. Klaus, for many fond memories in the lab, especially all your candy bars. Qin Xiaochun for your friendship and warm heart. Thank you to the Forestry staff who have supported me along the way, including: Gayle Kosh, Dan Naidu, Heather Akai, Marissa Relova, Debbie, Samantha and Lyn; and, Harry Verwoerd and Jerry Maedel for all of your incredible support. Family: To my aunts, uncles and cousins whose example of community and togetherness (amidst some pretty diverse personalities) has shaped my life and made these efforts possible. Thank you for you love and generosity. To my grandmothers, when I think of your perseverance and resilience I am humbled beyond words. Thank you for your undying support and love to me throughout my life, for all the cards and creations, I am truly blessed to be your grandson. To my in-laws, for your patience, support and prayers throughout these past several years. To my brothers and sister; for the love, devotion and faith you have in me. I take joy and find rest in knowing that whatever happens in life you will always be there for me – and I for you. To my parents, for the years and years of utter devotion, sacrifice and love to make my dreams and aspirations a reality. You have given so much to make sure that I had all I needed to become the person God intended for me to become; and for your never- ending support of my crazy adventures, enabling me to climb to the tops of the trees - and pushing me to stretch further. Finally, for helping me realize that no matter what I do, who I become, or how much I fail that I will always be loved. Yours is the example I hope to gracefully pass along. To my wife for the gift of patience, encouragement and love. Though these past six years, your devotion, support and dedication has shown me that whatever path we take, we will be on it together, supporting one another the best way we know how. And should it be along the road less traveled – as it most likely will be – that our togetherness will make all the difference. And to the One who makes all things possible. “In his hand are the depths of the earth; the heights of the mountains are his also. The sea is his, for he made it; for his hands formed the dry land.” –Psalm 95:4-5. The beauty of creation has captivated me.  xvi Dedication   I would like to dedicate this dissertation to my Grandma and Nana who will have seen this chapter of my life begin and end. To my Granddad who has been with me in spirit since the years of my childhood, and to my Papa who was with us when this work began, but will now join the celebration in spirit.  To the late Dr. Jim “Jimmy Dale” Holloway, Hollis Lasley and Stu Gardner for your lasting impressions in my life.  And finally, to Craig, Scott, Brie, Mom, Dad and Andrea.   1 Chapter 1:  Introduction In recent years there has been a paradigm shift in forestry: from the management of traditional forest products to a more holistic approach in which the complex links between forest, environment and society are recognized (Mery et al., 2005). Historically, the laws regarding forests were quite narrow, focusing almost exclusively on resources from timber (Food and Agriculture Organization of the United Nations (FAO), 2011). Yet a few decades ago these laws started regulating other non-wood forest products, including forest health, biodiversity, ecosystem function, tourism and recreation (Food and Agriculture Organization of the United Nations (FAO), 2011). This represents a shift away from traditional monoculture plantations to a more multi-objective approach to forestry, involving forests which must be managed for a range of purposes (Knight, 2000). Sustainable Forest Management addresses this shift by better accounting for the multiple roles of forests, especially when accounting for the social roles they play in society (Wang, 2004). For instance, forest are known to enhance human well-being by offering positive health and physical benefits (see Bishop and Hull, 1991). John Muir summed up these social benefits when he wrote the following: “Everybody needs beauty as well as bread, places to play in and pray in, where nature may heal and give strength to body and soul alike” (Muir, 1912). These positive benefits have been explored by numerous researchers interested in understanding the influence of nature and wilderness on human health from an empirical basis. (e.g. Kaplan, 1995; Kaplan and Talbot, 1983; Parsons, 1991; Rossman and Ulehla, 1977; Ulrich et al., 1991).One of the major outcomes of this research is that experiences in nature can lead to a reduction in stress (Hansmann et al., 2007; Kaplan, 1995; Kaplan and Talbot, 1983; Parsons, 1991).  If social benefits of forestry continue to be supported and explored, forest management practices may progress toward a more holistic approach, leaving behind the limited view of timber management as paramount. As forest management practices embrace the social benefits of forests, subsequent economic benefits might also be realized (Mather, 2001). One of these is the economic opportunity found through forest recreation, which has a direct connection with human health and well- being. Forests managed for recreational purposes have become more common in recent  2 decades (Knight, 2000), with a variety of projects being created around the world to encourage recreation and tourism (Bell and Evans, 1997; Cloke et al., 1996; Knight, 2000; Romano, 1995; Selin and Chavez, 1995). One implication to this new wave of forest management practices is that the aesthetic quality of forests will likely become increasingly important. Hints of these changes can already be found in national forest management regulations (e.g. Forestry Commission, 1989; USDA Forest Service, 1995; USDI Bureau of Land Management, 1998) and also in provincial or state forest systems (e.g. BC Ministry of Forests, 2004; Ontario Ministry of Natural Resources, 1987; Wisconsin Department of Natural Resources, 2003). Furthermore, evidence of the importance of aesthetics for forest certification or standard programs has been presented (Harshaw et al., 2005; Sheppard et al., 2004). Sheppard et al. (2004) reviewed fourteen forest certification or standard programs throughout the world for their inclusion of visual or aesthetic criteria, discovering that six of the fourteen  programs mention aesthetics as part of the program standards. Harshaw et al. (2005) reviewed eleven certification programs (many of which are included in the previously cited fourteen), finding that five of the eleven programs mention criteria or indicators for visual quality as part of the certification program, with two additional programs including visual quality more loosely. The mere inclusion of visual quality does not mean that these systems have addressed how the criteria or indicators lead to specific and measureable outcomes. In fact, these reviews point out that only one program provides an explicit and quantifiable outcome regarding visual quality (Harshaw et al., 2005; Sheppard et al., 2004). Yet, the inclusion of visual quality in forest management certification and standards, as well as in forest practice regulations, demonstrates that governments and companies are interested in the management of aesthetics. 1.1 Managing Forested Scenic Landscapes Managing forested scenic landscapes has occurred for some time, with the earliest recorded forest management plan being conducted within a national forest in 1908 in the United States (USDA Forest Service, 1995). However, scenic forest management did not play a significant role in forest practices until the mid-late 20th century (Palmer, 2008; USDA Forest Service,  3 1995). According to Hays (1987), it was not until Americans had the combination of sufficient discretionary income and an expansive highway network that afforded them opportunities to regularly explore the wide variety of the country’s natural environment. In the 1960s and1970s the timing of the available income and highway access coincided with an increased demand for natural resources, including those derived from forests (Hays, 1987). In order to meet the demand for natural resources, alternative silviculture practices were explored, and with it, an increase in clearcuts (USDA Forest Service, 1995).  The increase in clearcutting to meet the resource demand and the simultaneous exploration of scenic areas led to public outcry in 1970s (Boerner, 1986; Williams and Tolle, 2001). Eventually, due to public pressure, harvesting practices, like clearcutting, became more regulated through the National Forest Management Act of 1976. This gave the public an opportunity to impact forest management plans (Palmer, 2008). Other US Congress Acts which have influenced the management of scenic areas are: the Multiple-Use Sustained-Yield Act, which designates forests for multiple-uses and not merely for natural resource extraction (U. S. Congress, 1960); the National Environmental Policy Act, which encourages the harmony of the environment and people, including preventing impacts which will negatively impact human welfare (U. S. Congress, 1969); and, the Forest and Rangeland Renewable Resources Planning Act, which states that long-range planning of natural resources should serve the public interest (U. S. Congress, 1974). The push for the formal management of scenic landscapes in North America can be traced back to the mid 20th century when the US Chief Forester met with Dame Sylvia Crowe, who was “an eminent British  landscape architect” (Bell, 1999) and contributed to the aesthetic management of forests (Crowe et al., 1975; Crowe and Mitchell, 1988), showing how landscape design could mitigate the negative aesthetics effects of forest operations (USDA Forest Service, 1995). Soon after, the Forest Service began hiring and consulting with landscape architects to help in the management of scenic areas, including Dr. Burton Litton (USDA Forest Service, 1995) who developed one of the earlier frameworks for landscape design in forested areas (e.g. Litton, 1968, 1974). Along with Dr. Litton, the Forest Service developed several versions of The Visual Management System y(Bacon, 1979; USDA Forest Service, 1974, 1995).  4 Clearly, the public response to clearcuts has a significant influence on shaping the management of forests in the US. This influence has engendered numerous research studies which have focused on identifying the effects of clearcuts and disturbance on forests, as perceived by humans. These studies have found that people prefer uncut to clear-cut forests (Palmer, 1998, 2008; Ribe, 1999, 2005), that increased cutting activity increases negative scenic effects (Palmer et al., 1993; Palmer et al., 1995), that higher human-made contrasts in the landscape are generally less preferable (Ribe et al., 2002; Shang and Bishop, 2000), and that disruption is generally found as less beautiful, yet sometimes still acceptable (Pâquet, 1993; Pâquet and Belanger, 1997). Although the empirical evidence emerged years later, the known negative public responses still spurred the USDA Forest Service to develop a Visual Management System (USDA Forest Service, 1974) with other regions adopting similar systems (BC Ministry of Forests, 1981). Since that time, the systems, rules, and codes have been continuously adapted in both the U.S. and BC (BC Ministry of Forests, 2001b; USDA Forest Service, 1995; USDI Bureau of Land Management, 1998) and have been expanded into other state and provincial systems (e.g. Ontario Ministry of Natural Resources, 1987; Wisconsin Department of Natural Resources, 2003). The management of visual resources or scenic areas has evolved over the years, yet has continued to serve its original intent of mitigating impacts and subsequent public reactions (Gobster, 1999; Sheppard, 2000). One of the purposes of VRM is to help assess impacts of management alternatives (Bishop and Hull, 1991). This is important because visual resources provide ‘products’, to people and nearby communities. Examples of these products are: mood, mental health, physical health, confidence in land management, residential satisfaction, and prosperity (Bishop and Hull, 1991);  all of which play important roles in society. VRM was created to ensure that the ongoing provision of these products would be maintained as a result of modification, by mitigating the visual effects stemming from harvesting or other operational changes. It follows then, that since the visual resources are intended for people, a VRM system should be ‘human-centered’, where judgments and assessment of impacts are relevant to the people  5 who stand to be affected by the modifications. These judgments and assessments are most often carried out by two different approaches, the expert- and perception-based approaches (Daniel, 2001; Daniel and Vining, 1983; Zube et al., 1982). The expert-based approach has been grounded on traditions like landscape architecture and public land management practices, whereas the perception-based approach has been based on environmental perception and landscape assessment research (Daniel, 2001). The expert-based approach assumes that there are universally accepted design criteria, based on landform features (such as form, line, unity, variety), and that these features can be used to influence human perception about landscape quality (Daniel, 2001). The perception-based approach is founded on understanding the biophysical characteristics of the landscape which evoke psychological responses from individuals, and using these responses to inform how landscape features affect preferences (Daniel, 2001). An early example of this approach can be seen in Daniel and Boster’s research (1976), where the aim was to find a quantitative measure that could use individual responses to estimate a landscape’s “scenic beauty”, in order to compare landscapes and predict the consequences of alternative views. The psychophysical method, which is founded on the perception-based approach, has played a significant role in understanding the implications that different forest practices have upon individual preferences (see Ribe, 1989 for an extended review). For instance, research has shown that: distance plays an important role in determining scenic beauty (Arthur, 1977); that green-up or re-growth quickly increases perceptions of visual quality (BC Ministry of Forests, 1994a); that clearcuts significantly reduce visual quality (e.g. BC Ministry of Forests, 1996; Palmer, 1998, 2008); and that, in general, increases in retention levels positively increase scenic quality (Ribe, 2005, 2009). In the past few decades, as computers and visualization technologies have increased, many researchers have used these in combination with the psychophysical approach to try and identify how particular biophysical features influence preferences. One advantage of these technologies over previous studies using photos and real-life surveys is that the different landscapes and elements within them can be controlled. For instance, Meitner et al. (2005) used 3D visualization technology to study the impacts of forest fragmentation and retention on preferences, and found that people prefer larger and more aggregated block patterns and higher amounts of dispersed retention rather than smaller, more frequent disturbances and grouped retention.  6 1.2 Integrating Geographic Information Systems for Visual Resource Management When VRM began in the late 1960s to early 1970s, analog mediums  such as cameras, physical models, and maps constituted the bulk of the technology that landscape planners used for landscape design (B. C. Ministry of Forests, 1974; Lange, 2002; Litton, 1968), with one intriguing exception where computers were used to model geographic elements of the terrain (Travis et al., 1975). In this way, VRM was naturally bound to the limitations of technology of its era. Technology has dramatically progressed in recent years; planners can now create near photo-realistic 3D simulations of landscapes and simulate their plans. (Lewis et al., 2004; Orland et al., 2001). This technology provides a more efficient way to design harvests and explore various alternative designs quickly. Yet even with these advancements, the planning and evaluation processes still seem to be based on a discretization of the landscape into single viewpoints and distinct landscape units, an inherent reality of the original paper-based systems (see BC Ministry of Forests, 1997b; Marc, 2008 for examples of this influence). Geographical Information Systems (GIS) have been used in landscape modeling and assessment in forestry (Bishop and Hull, 1991; Bishop and Hulse, 1994; Domingo-Santos et al., 2011; Tyrvainen et al., 2006). GIS is often employed in landscape planning and harvest design, and its potential in these fields is emerging. Currently, forest planners use GIS to design harvest plans in 2D and render them in 3D (see Lewis et al., 2004). Using GIS data to drive 3D simulations enables the development of spatially accurate models (Tyrvainen et al., 2006). However, the process of iterating between planimetric (2D) and perspective (3D) to design a harvest that meets particular objectives can be cumbersome and requires large amounts of time. Tools that interoperate between both dimensions will be very valuable in aiding planners and increasing efficiency. GIS-based developments will likely lead the way in these developments as they are capable of dealing with spatial problems at large scales in both 2D and 3D. Some examples of these new developments can be found in the many areas of VRM. For instance, Dramstad et al. (2006) found that it may be possible to use 2D landscape indicators based on remote sensing imagery to help derive an indicator of visual quality in perspective  7 view. Specifically, Sang et al. (2008) explored the differences of shape and complexity in 2D and 3D, the first being measured by landscape indices, the latter based on a preference study. The comparison of the 2D measurements and the 3D ratings allowed the authors to explore potential relationships between how an individual perceives characteristics of the landscape, such as complexity, and how the same landscape is mathematically described.. In their study they found that the indices measuring complexity using 2D correlated, at times, with 3D preferences of the landscape (Sang et al., 2008). However, it is important to note that the correlations are specific for a particular landscape, and may not be applicable for others kinds of scenes and environments. The authors did suggest caution  about concluding that these kinds of indicators should be used to represent preferences more broadly (Sang et al., 2008). Ode et al. (2009) also used visualization technology coupled with GIS-based spatial metrics to identify indicators of perceived naturalness. In the study, the authors discovered that indicators such as the shape index of edges, number of woodland patches and the level of succession around different patch types were most closely related to perceived naturalness (Ode et al., 2009). Ode and Miller (2011) also used GIS-based tools to identify links between landscape preferences and spatial indicators, finding that there is a relationship between landscape preferences and complexity as measured by the metrics: Shannon evenness index, Shannon diversity Index and the Aggregation index, which focus on calculating the distribution and spatial organization of landscape characteristics (McGarigal and Marks, 1995). Spatial planning has also been addressed in the literature. Methods for automating VRM based on objective criteria have been developed (Chamberlain and Meitner, 2009). Viewshed analyses, a GIS-based visibility analysis (see De Floriani and Magillo, 2003 for a review on these kinds of analyses), is also used in VRM (e.g. Bishop, 2003; Fisher, 1996). Gimblett et al. (1996) showed how viewshed analysis could be used to model the movement of visitors through a park trail system. Fisher (1996) reported on ways in which viewsheds could be used, including discussing how viewshed can be used to conduct visual impact assessments. Germino et al. (2001) used viewshed analyses to identify the extent, relief and depth of landscape terrain area using multiple observers in order to demonstrate how viewshed modeling could be used to quantitatively model scenic quality over large regions. The ability of GIS to precisely measure landform, model visibility and conduct analyses across large  8 spatial extents, combined with new visualization technologies, provides planners with tools that enable the more effective management of visual resources. However, as posed by Lange (2002), the question is: will planners adopt the technology? 1.3 Forests and Aesthetics in British Columbia British Columbia represents a unique opportunity to use these technologies to aid in the management of forested scenic areas because it has a vast forested mountainous landscape where the management of scenic areas is not only important, but expansive. In BC, forests play a critical role in the social, economic and ecological fabric of the province.  A large portion of the economic well-being of the province comes from exports and domestic use of forest resources. In fact, wood, pulp and paper products account for over 30% of all exports from BC (Schrier, 2011) and in general, the forest sector accounted for over 4% of provincial GDP (BC Ministry of Forests and Range, 2011). With nearly 60% of British Columbia classified as forest lands, and over 14% of its land set aside as Protected Areas, forests represent a substantial ecological significance in the province, including being home to about 33% of plant and animal species in BC (BC Ministry of Forests, 2010). Forests also play a major role in tourism in the province. In 2009, forest recreation activities contributed $2.2 billion in GDP, constituting about 30% of the GDP derived from the forest sector that year (BC Ministry of Forests, 2010). There are over 45,000 individuals employed in providing forest recreation activities, some of which involve the maintenance and supervision of over 23,000 campsites and 20,000 kilometers of trails (BC Ministry of Forests, 2010). These statistics represent the significant influence of tourism on the forest sector and its importance for how forestry in BC is managed. As sustainable forest management continues to influence forest planning, and the balance of social, economic, and ecological aspects of forestry are weighed, the province must find ways to ensure that forests continue to provide for the variety of products and services needed by both humans and the environment. One of these important services rendered by forests pertains to the aesthetics of the many residential areas, scenic highways, campsites, and similar places where the scenic beauty of a landscape plays into individual experiences and sense of place. As such, the BC Forest and Range Practices Act  has identified Visual Quality as one of its eleven resource values,  9 along-side Biodiversity, Timber, Water, and Wildlife, to name a few (BC Ministry of Forests, 2004). In so doing, this Act provides protection to the aesthetic quality of BCs forests, a value shared by residents and visitors alike, through VRM (see Marc, 2008). VRM works to ensure that harvests in public lands are completed in such a way that the visual impacts from forest operations are limited. For instance, Figure 1.1 shows a harvest along the Coquihalla Highway in BC where visual quality is regulated under the BC Forest Practices Act, whereas Figure 1.2 shows a harvest just off Highway 101 in the State of Oregon, US, at a location where visual quality is not under regulation.  Figure 1.1 Example harvest where visual quality is regulated. Image obtained using Google StreetView, © Google 2011  10  Figure 1.2 Example harvest where visual quality is not regulated The program guides the management of approximately 14.6 million hectares of the forested area throughout the province (BC Ministry of Forests, 2010), including the many lake shores, scenic highways, mountains, and islands connected by numerous ferry routes throughout the province. These areas, depicted in Figure 1.3, represent a substantial space in the province, encompassing over 25% of the total forested area (or 55 million hectares) and 15% of the entire province (BC Ministry of Forests, 2010). With such a substantial spatial extent visually protected, and the influence that tourism has upon the province, BC represents a unique opportunity to explore the uses of GIS-based approaches to VRM.  11  Figure 1.3 Visually Sensitive Areas in British Columbia 1.4 Research Theme, Purpose and Goals The theme of this research is the development and application of new advancements in GIS to more effectively manage forest aesthetics, particularly pertaining to forest management in BC. New methods, using geospatial technologies, have been created to accurately depict how an individual would perceive a landscape while moving through it, whether by boat, car or on foot. Also, new empirical research has been completed showing the quantitative effects of  12 how using different forest block shapes can influence individual preferences for the harvest designs. The new methods can aid planners in finding optimal locations to place harvests and balance the economic bottom-line while simultaneously minimizing aesthetic disturbance. The empirical work builds upon this research by identifying how planners can use the shapes of these blocks to further reduce the negative visual quality outcomes of their operations (Carlson, 1977; Crowe and Mitchell, 1988; Shafer, 1967). The purpose of this research is four-fold (and in the following chapters, respectively): 1. To use 3D visualization technology to explore how the effects of harvest block shape can influence preferences for harvest design (Chapter 2); 2. To demonstrate a new visibility algorithm that can efficiently portray how an individual may experience the landform as they move through the landscape (Chapter 3); 3. To discuss the application of these new methods and insights for VRM in British Columbia (Chapter 4); and, 4. To provide a general overview about the trade-offs that exist between the management of visual quality and other values (Chapter 5). This research began by finding ways to employ technology to help managers more effectively design harvests to serve a variety of goals. Initially, the research was intended to build upon the potential for automating harvest design for VRM (Chamberlain and Meitner, 2009), extending it to include ecological and economic goals and constraints at the landscape level. As the research began, two important issues emerged. First, that automating the assessment of visual quality at the landscape level required a whole new algorithm that could quickly and efficiently calculate the percent of visible alteration. Second, that additional empirical research was needed in order to link preferences with spatial characteristics of harvest design. Until these issues were resolved, automation of design, and even landscape level assessment, was unfeasible. The first of two goals was to address a critical gap in the literature regarding the effects of harvest shape on preferences. Landscape architects have argued that more natural looking shapes, as opposed to irregular shapes, fitting within the landform are more preferred (BC  13 Ministry of Forests, 1997b; Bell, 2001; Diaz and Apostol, 1992; Ribe, 2005). However, there exists no objective or geometric standard that defines perceived shape as it pertains to harvest design. In order to objectively measure design based on the characteristics of a shape and how they influence preferences, some standard must exist. There have been numerous studies on preferences of shapes from psychology research (Attneave, 1957; Attneave and Arnoult, 1956; Day, 1967), but not as it pertains specifically to forestry. So, in order to spatially assess what defines a “natural-looking” harvest on the landscape, or to automate their creation, a spatial metric must be defined and validated against individual preferences. The hypothesis was that there exists a mathematical expression of organically-perceived shapes and that the expression may lie within metrics often used by landscape ecologists. The second goal of this research was to develop an efficient algorithm that could measure the amount of visible alteration of harvest designs in perspective view. If this worked, it may be possible to measure the percent of visible alteration from any view and for any harvest plan, depending upon the resolution of the analysis and type of silvicultue practice. Furthermore, the outcome of this analysis could be used to help planners quickly identify areas on the landscape that may be more sensitive to alterations, reducing the amount of time it takes to make designs in scenic areas. Since forests provide a variety of products and services to humans and the environment, it is also wise to consider the interplay between aesthetics and ecology. Past studies have shown that technology and spatial models can be used to help planners better manage forests for multi-objective purposes (Seely et al., 2004). So, a final goal of this dissertation, was to begin exploring the trade-offs between management of visual resources and other forest management objectives. The seeds of this exploration may serve to inform the further development of a decision-support tool which was envisioned at the outset of this research endeavor. 1.5 Research Structure The first and last chapter of this dissertation provide context for the main body of research presented. The initial chapter opened with an argument about why the management of scenic areas is important in forestry, for both social and economic reasons. A review of VRM was  14 presented, along with how new geospatial technologies are influencing the management of these scenic places. The concluding chapter summarizes the main body of this research, and gives an overview of how geospatial technologies can be used to more effectively manage aesthetics in forested landscapes. In Chapter 2, ratings from individual interviews using 3D renderings of photo-realistic landscapes is shown, which quantitatively identifies how different shape characteristics influence preferences. Section 2.5 is an added extension of this research. In this section, a few spatial metrics are compared against the mean ratings from the user study to see if there are any possible connections between the indicators and people’s aesthetic preferences of forest landscape. With the exception of section 2.5, this chapter has been submitted for peer- review in an international journal. In Chapter 3, an efficient viewshed algorithm is presented, which provides important information about how the landform would be perceived by an individual moving through the landscape. Unlike the standard viewshed which provides only binary information, the outcome of this algorithm uses a combination of trigonometry and the anatomy of the human eye, to produce a viewshed that gives a range of visibility for areas on the landscape. Several applications of this research are presented in Chapter 4, as well as some lessons learned about viewshed analysis in general, which are important considerations for planners. Chapter 4 presents several applications of the new algorithm presented in Chapter 3. These sections were not included in Chapter 3 as they pertain to issues primarily relevant to BC s. The focus of this chapter is how the algorithm can be used to address problems and limitations of the current VRM system used in British Columbia. The chapter is more methodological in its approach, demonstrating how the algorithm can be used by planners to accomplish particular objectives for forest management. Furthermore, some arguments are made about the degree to which the representation of a single viewpoint as a proxy for a landscape experience can influence visual quality assessments. Chapter 5 takes a step back from the methodological and technical advancements and looks at VRM as one component of multi-objective forestry. In this brief synthesis several trade- offs with regards to aesthetic, economic and ecological goals are presented. Special  15 consideration to the ecological trade-offs is given. This chapter provides context to planners dealing with operations design in visually sensitive areas. Its focus is on presenting the relationship between visual quality and ecological integrity pertaining to emerging silviculture practices in British Columbia.  16 Chapter 2:  Quantifying the Effects of Harvest Block Design on Aesthetic Preferences This chapter presents new findings on the effects of different of harvest design shapes onpreferences. The results are based ona user survey to quantitatively define the differences between several different kinds of shapes and how the elements of these shapes impact preferences. 2.1 Introduction Forest landscapes are constantly in a state of flux as a result of both human and natural factors. These changes, such as clearcuts, can dramatically impact aesthetics in places where the scenic quality affects social, cultural and economic benefits derived from the forest (Palmer, 2008; Palmer et al., 1995).  In order to be more responsive to these changes, landscape planners can use GIS and visualization technology for landscape management (Lewis et al., 2004) in order to help reduce the visual impact of these harvests. This is especially true when it comes to the realities of commercial forest harvesting, where operational planners are often tasked with the challenge of achieving a multitude of competing goals (Knight, 2000). In British Columbia, landscape/forest planners have a keen interest in the protection of aesthetics in places called visually sensitive areas. These areas are typically visible from communities, travel corridors or other public use areas and are often seen by a large number of people (BC Ministry of Forests, 1997b).  They are areas where the public has come to expect a high degree of scenic quality and regularly possess landscape characteristics that draw attention because of their natural beauty. In the 1960’s and 1970’s the development of Visual Resource Management (VRM) was begun in the United States in order to deal with public backlash as a result of timber harvest practices (USDA Forest Service, 1974) and is still in use today (BC Ministry of Forests, 2001b; USDI Bureau of Land Management, 1998). It has evolved over the years, but even with these changes it can be argued that it has served its original purpose of mitigating the visual impacts and subsequent public reactions (Gobster, 1999; Sheppard, 2000).  17 VRM draws from principles in landscape architecture and environmental psychology and attempts to connect those principles to operational forest management, as is evident in different VRM systems (BC Ministry of Forests, 1994c, d; USDA Forest Service, 1995). VRM adopts findings from psychophysical research, which can elucidate how particular biophysical elements of the landscape evoke positive or negative emotional responses (Daniel, 2001), and the more traditional expert-based approach, which is based on landscape elements like pattern, line and form (Bell, 2004), to ensure the management of visual quality. Several methods have been used by researchers who aim to understand how people perceive the landscape (Taylor et al., 1987; Zube et al., 1982). One of the more ubiquitous is the psychophysical method, which enables researchers to quantitatively identify the role of certain landscape characteristics upon an individual’s perception (e.g. Daniel, 1990; Daniel and Boster, 1976; Ribe, 1989, 2005, 2009). These methods have impacted the way VRM is conducted as they provide the essential connection between potential landscape change and perceived visual impact. Numerous studies have worked to identify specific elements of landscape modification and their relationship to individual preferences within forestry. For instance, the amount of visibly altered landscape seems to be a good predictor of public acceptability (BC Ministry of Forests, 1996). Shang and Bishop (2000) found that it was not just the size of the modification, but the contrast weighted visual size, which is a function of the amount of modification and the tonal contrast of different landscape patterns on the landscape. Dispersed harvesting patterns have been shown to improve scenic beauty ratings (Ribe, 2009), specifically, in the foreground and middle ground distances (Ribe, 2005).  Also, the rate of green-up can influence preferences, with visually effective green-up nearing recovery to original preferences within ten years (BC Ministry of Forests, 1994a). On the contrary, clear-cuts are known to garner negative aesthetic judgments  (e.g. BC Ministry of Forests, 1994a; BC Ministry of Forests, 1996; Lindhagen, 1996; Palmer, 2008; Pâquet and Belanger, 1997; Ribe et al., 2002). Another key finding that emerged from perception-based research is that for landscape assessments, middleground distances (500 m – 5000 m) may be considered the most critical and sensitive (Hull and Buhyoff, 1983; Litton, 1979; McCool et al., 1986; Pâquet, 1993).  18 Ribe (1989) provides a comprehensive review of preference studies where the influence of particular forest landscape elements are examined for their relationship to preferences. In recently published work, authors suggest that it may be possible to identify landscape indicators and relate them to preferences (e.g. Germino et al., 2001; Giles and Trani, 1999; Gulinck et al., 2001; Hagerhall et al., 2004; Jessel, 2006; Ode and Miller, 2011; Palmer, 2004; Sang et al., 2008). Specifically, there have been recent efforts to identify visual landscape indicators. For instance, Tveit et al. (2006) developed a theoretical framework that includes nine visual concepts: disturbance, visual scale, complexity, naturalness, coherence, historicity, ephemera, stewardship, and imageability; describing each with regard to their physical landscape characteristics. Ode et al. (2008) builds upon this framework by focusing specifically on the application of these visual indicators, but cautioned that they need to be tested for their applicability for measuring visual quality. Recent research attempts to objectively measure visual quality using spatial indicators. Sang et al. (2008) explored the differences of shape and complexity for harvest patches as measured in 2D and 3D to see if an individual’s experience of the 3D landscape could be associated with landscape indicators measured in 2D. In their study they found that the indices measuring 2D complexity correlated most closely with 3D preferences of the landscape, but were careful about concluding that these indicators should be used to represent preferences more broadly (Sang et al., 2008). In a continuation of this research, Ode and Miller (2011) identified links between landscape preferences and spatial indicators, finding that there is a relationship between landscape preferences and complexity as measured by the metrics: Shannon evenness index, Shannon diversity index and the Aggregation index, which focus on calculating the distribution and spatial organization of landscape characteristics (McGarigal and Marks, 1995) As Ode et al. (2009) argue, since perceived naturalness has a shown to be a well-documented effect on landscape preference, identifying measureable indicators of naturalness may allow for a way to more objectively monitor and assess landscape change. In the study, the authors discovered that indicators such as the shape index of edges, number of woodland patches and the level of succession around different patch types were most closely related to perceived  19 naturalness (Ode et al., 2009). These recent efforts, demonstrate the potential for, and argue that landscape indicators might be used as a way to measure landscape character and how they relate to psychological responses (Ode et al., 2010). This new research builds upon past work that identified landscape elements and how they would be perceived, by finding ways to measure landscape characteristics using spatial metrics. Given many of the advancements in geographic information system and image analysis, it may be possible to use landscape indicators to measure spatial elements of design relevant to landscape planning in forestry. Doing so would enable planners to more objectively evaluate the effects of their designs on the landscape. The research builds on this idea by increasing the understanding of how harvest block shape influences preferences, and to see if shape can be described by spatial landscape indicators in order to identify if any of these metrics could be used as visual quality indicator for harvest blocks. Two examples of these research gaps are: harvesting along lines of force or along ridgelines. For instance, the designation of lines of force, which is often depicted by concave and convex landforms such as valleys and ridgelines, are important for Integrated Visual Design (BC Ministry of Forests, 1994c, 1997b, 2001a; Marc, 2008). Yet, the relationship of using lines of force to influence aesthetic preferences has not been quantitatively defined. As for ridgelines, empirical research that examines the effect of harvesting across ridgelines is almost non-existent. An exception is the work of Hammitt et al. (1994) where they found that the greater number of visible and forested mountain ridgelines increased visual preference. Ridgeline effects have been generally discussed in the landscape architecture tradition, where researchers counsel against harvesting in areas where dissimilar materials merge, such as along ridgelines, skylines and mountain tops, because these edges are visually vulnerable to changes(Litton, 1974). Theories about landscape aesthetics are not always based on quantitative empirical research connecting those theories to human responses (Ode et al., 2010). However, even though there remain gaps in our knowledge, decades of experience in applying VRM systems to forest operations suggest design does play an important role in how people perceive landscape change. With better visual landscape design, the amount of timber harvested does not need to be an inverse relationship with  visual quality (Picard and Sheppard, 2002a, b). The goal here is to not to merely hide the harvesting operation, but to focus on using landscape design to  20 demonstrate visual stewardship (see Sheppard, 2000). So, perhaps the idea of using landscape indicators could be adapted from landscape ecology and used to created indicators for harvest design. The goal would be to identify spatial aspects of design that could influence preferences for a design and subsequently, the landscape. The shape of a harvest block is one obvious candidate for a design indicator. Landscape designers within forestry have taken note of the influence of block design, but limited research has attempted to quantitatively identify the effects of varying shape criteria. For instance, square-shaped harvests are dominant in Oregon’s national forests (Ribe et al., 2002). In the same region a recent study was conducted to identify the effects of varying shape, pattern and retention for ecosystem management. This provided an opportunity to explore the ramifications of these characteristics from a visual quality perspective. Ribe (2005) used images from these harvests and explored how different design attributes would affect perceived scenic beauty. The results showed that the shape of the design had little effect on individual preferences compared to the effect of retention amounts (Ribe, 2005). Yet, this study cannot answer how particular aspects of shape alone can influence preferences, because other variables, such as retention amount were included in the study. Understanding how shape alone is important information because for a given harvest amount, planners can use different kinds of shapes in their designs (BC Ministry of Forests, 1994d), which can effect preferences. With such little research on the influence of block shape design, perhaps there is an opportunity to explore the impacts of shape design more thoroughly as it pertains to individual preferences. Human shape perception has been studied in other fields in the past. Michels and Zusne (1965) provide a review of many of the earlier studies on shape perception in psychology, identifying several  mathematical and physiological themes that describe shape and complexity; including: linearity of line contours, area, rotation, number of sides, angles, dispersion of points and sides, symmetry, elongation and others. Attneave and Arnoult (1956) give extensive detail in describing specific mathematical descriptions of shape and pattern and some issues in understanding shape complexity, but also state that from a physiological perspective many of these kinds of descriptions may have little or no value because they are context dependent, although none were explicitly mentioned. For instance,  21 the context in Attneave and Arnoult (1956) was a simple interface showing drawn shapes on a blank background, this context is quite different than how the shape of a harvest is perceived in a natural landscape. Attneave (1957) was one of the earlier studies attempting to identify how people understand shape complexity. This study revealed that the number of turns in a shape was the greatest predictor of complexity and that curved shapes were judged no more complex than angular shapes. Day (1967) also analyzed the influence of shape complexity on preferences and found that humans prefer a medium level of complexity more than either low or high levels. In this particular study, this inverted-U relationship may be due to an extremely high and low level of complexity for the drawn shapes, as has been notied in (Kaplan et al., 1972). Given such limited research pertaining to the shape of harvest designs, this research investigates how the shape of a harvest block can influence public preferences. We acknowledge that landscape indicators may be useful in measuring visual quality, but argue that it is important to first identify how different shape characteristics can influence preferences before we use indicators to measure shape. This is an important step to helping researchers identify psychologically-relevant mathematical descriptors of shape for harvests. For the purposes of this research, the term shape consists of a combination of three characteristics: geometric primitiveness, complexity and aspect ratio. Geometric primitive pertains to the simple shape, such as a square or circle. These kinds of shapes relates to the different kinds of harvest blocks often visible on a landscape. Complexity has been shown to influence preferences for shape (Attneave, 1957; Day, 1967) and more generally on landscapes (Ode et al., 2010; Ode and Miller, 2011). In this study, complexity was increased by increasing the number of edges and lengths of the edges. Aspect ratio was selected because it provided a way to control for the possible effects of an elongated landform. Aspect ratio is the ratio of the height and width of a block.  This chapter expands the notion that more organic shapes are preferred to geometric shapes by attempting to quantify the differences between the two. There are, and for the foreseeable future will continue to be, harvest regimes where partial-cutting or variable retention will remain either economically or operationally impractical. For these situations, shape may take on an important role in mitigating visual impacts on the landscape.  22 2.2 Materials and Methods To investigate the influence of different harvest shapes on preference ratings, we systematically developed a set of possible harvest designs for a single landscape. These designs were then simulated by the creation of 3D near photo-realistic images. These images were used to survey individuals about preferences for each of the different harvest designs. The use of 3D photo-realistic software, allowed for the control of three distinct characteristics of shape (geometric primitive, complexity and aspect ratio) while eliminating many other possible elements that could have influenced preference ratings. We needed to ensure that as many landscape and design characteristics as possible were controlled. To accomplish this, we identified a single landscape unit in British Columbia, which was classified as a visually sensitive area according to the Visual Landscape Inventory (BC Integrated Land Management Bureau, 2011). The simulated 3D landscape consisted of relatively smooth terrain with a moderately dense second-growth single species forest. There are few unique or distinctive features on the landscape; slope was mostly constant, there were no surrounding mountains or valleys, there were no obvious or strong lines of force, and the visual absorption capacity of the underlying terrain would be considered relatively low. The total area of the land base that was visible and set aside for possible harvest was 62ha (roughly 1km across and half a km deep), which allowed for a variety different harvest to be tested. Figure 2.1, illustrates the landscape with full forest cover.  23  Figure 2.1 Photo-realistic image of the landscape used to survey individuals regarding harvest design preferences, without any harvest activity. In total, 52 near photo-realistic images were generated, each depicting a potential design varying along the three variables. While we were able to balance the number of images across the shape elements, the survey design did not lend itself well to a fully balanced design.  We found that at low levels of complexity it was impossible to vary the designs sufficiently while keeping true to the geometric primitives.  Therefore we chose to reduce the low complexity designs to 12 while the medium and high levels of complexity designs were represented by 20 images each.  We also found aspect ratio to have a similar limitation. At high and medium levels of complexity it was difficult to provide distinctly unique designs in the high aspect ratio cell and therefore limited the number of designs to 12 for high aspect ratio while creating 20 designs in the low and medium levels of aspect ratio. Regarding the simulated designs, three variables were altered: geometric primitive, complexity and aspect ratio. Geometric primitive refers to a specific atomic shape; circle,  24 triangle, trapezoid and square. Complexity refers to the irregularity of the shape (three levels from high to low), and is connected by the number of perceived edges, although the number of edges were not individually controlled within each design. Aspect ratio refers to the elongation of a specific shape (three levels from high to low). Table 2.1 provides the study design matrix depicting the number of harvest designs derived for each of the three elements. Figure 2.2 illustrates the three different variables that were used to alter the harvest designs. Table 2.2 provides details regarding the landscape variables that were controlled for or altered, including the three design variables. Finally, Figure 2.3 provides nine examples taken from the study with their associated classifications for reference. Table 2.1 Study design matrix for harvest designs Complexity Aspect Ratio Geometric Primitive Circle Square Trapezoid Triangle Low Low 1 1 1 1 Low Med 1 1 1 1 Low High 1 1 1 1 Med Low 2 2 2 2 Med Med 2 2 2 2 Med High 1 1 1 1 High Low 2 2 2 2 High Med 2 2 2 2 High High 1 1 1 1 Total 52   Figure 2.2 Illustrations of the three harvest design variables that were examined in the study  25 Table 2.2 The list of variables used in the development of the harvest designs. Element of Design Constant Variable Description Landscape Y Only one landscape was used. The foreground cover was modified from its original in order to draw attention to the hillside where the harvest exists. Viewpoint Y Distance to foreground: 500 m; distance to background: 4km. Forest Cover Y Mix of Douglas Fir, Spruce and Hemlock Location of Shape Y The location of harvests on the landscape we kept closest to the center of the landscape as possible Perspective Area Altered Y The following are the ranges of measured perspective visible alteration (%): Avg (14.63), SD (0.76), Min (13.87), Max (15.38) Shape N Shape was altered across four basic geometries: square, circle, trapezoid and triangle. Complexity N Complexity was altered along three levels: low (simple shape), medium and high based on the author’s qualitative judgements. Hard edges were used for all deviations from the simple geometric shape (exc. circle). Aspect ratio N Three levels of aspect ratio were used: low, medium and high (ranging from nearly 1:1 to 12:1). A high level of aspect ratio meant the shape was elongated from left to right, following the landform. Note that high aspect ratio shapes could not fit on the landscape vertically while simultaneously maintaining the perspective area altered.   26  Figure 2.3 Nine examples of images used in the study with their categories of geometric primitive (p), complexity (c) and aspect ratio (a), as well as the mean ratings In order to create the images used in this study the following software was used: geographic information systems (ArcGIS®) for harvest designing, Visual Nature Studio® (VNS) with a custom set of tree images for rendering and Adobe Photoshop for quantifying the percentage of visible alteration, as these tools often used for the purposes landscape planning and harvest design by forest companies in British Columbia. The process of creating the designs required an intensive amount of iteration, from design to rendering to quantifying and then repeating  27 the steps we took helped to ensure that the specific characteristics of the harvest designs fit within the parameters we were controlling. An important challenge to note is the development of shapes requiring precisely straight or smooth lines. Following common practice, designs were created in 2D (planimetric view) using GIS software. An attempt was made to develop them using the 3D interface, but without real-time vegetation or precise terrain representation it was easier to develop in 2D. This required the simultaneous consideration of the landscape surface and the surface as it would be perceived from the viewpoint. After a design was created or altered, it would be rendered in VNS using a high contrast of land to vegetation for the visually sensitive area to check for two important elements. First, the amount of the altered vegetation was quantified by pixel amounts in Adobe Photoshop. Second, the edges were checked and changed in 2D to ensure that the integrity of the three design elements was consistent with the desired shape. This process revealed an inefficient method to precisely control for complexity, including the straightness of a line or curve, and the overall shape. This gap can significantly impact the speed and precision with which designers can quickly sketch out a harvest design and measure its effects in 3D. It would be far more effective to create designs if software could provide a 2D view for planning, while simultaneously showing how the design would affect a view in 3D where changes to vegetation were revealed in real-time. The survey consisted of five unique steps. Step one provided a description of the survey. Step two showed a photo (unrelated to the rendered images) and the rating interface which required the user to select a rating and click next before they could continue to the stage three to familiarize them with the rating task. In step three, nine preview images were simultaneously displayed to the user using a 3 x 3 matrix. The purpose of these preview images was to represent the full range of variability of our harvest designs to allow the user to anchor the end points of the rating scale (Brown and Daniel, 1990). In this stage, users were asked to carefully look at each image and identify how they would rate each on a scale of 1 (least preferred) to 10 (most preferred) following a similar scale used in Daniel and Boster (1976). Stage four consisted of rating each of the 52 images. Before the rating commenced, users were again reminded to rate the harvest design for the preference and to use the full range of the 10 point scale in their evaluations. Stage 5 consisted of follow-up  28 questions related to the ratings and the user-preferred design characteristics. The follow-up portion consisted of both open and closed-ended questions. Demographic data was collected along with the individual’s area of study. The survey was conducted using a custom user interface designed to guide the user through each of the five stages. The interface was designed for a standard 19” LCD monitor, which allowed the images to be displayed at a resolution of 962 x 674, in nearly full screen with space provided below for the rating scale. The order of each image was randomized for each user. The duration of response time for each image was also collected to determine any abnormal behavior with regard to the time it took for a respondent to rate an image (either far too little or too much). The participants were forty one students at the University of British Columbia. Advertisements, offering a $10 cash incentive for participation, were placed throughout campus in public places and individual departments. Emails were also sent out to departments around the university. All respondents were given an equal opportunity to participate.  In total there were 16 males and 25 females, ranging in age from 19 to 36. 28 individuals identified their study area as having an environmental emphasis and 11 identified their study area as having a design influence. One individual was removed from the analysis due to a poor range of ratings as this subject showed little discrimination among the stimuli. This left us with a total sample of 40 individuals. Several statistical analyses were produced. The degree of internal reliability between groups was produced using a Cronbach’s Alpha (Cronbach, 1951; Cronbach and Shavelson, 2004). Second, a Univariate General Linear Model using demographic data as dependent variables was produced to ensure that there were no statistically significant differences between demographic groups based on the mean ratings of the images. Third, a Univariate General Linear Model using the three shape characteristics as dependent variables, as well as their interactions, was produced to determine the statistical significance and effect size for each characteristic and their interactions with regard to the mean ratings. 2.3 Results We tested for effects of our demographic variables and found that neither age,  gender or area/program of study had any influence on the mean ratings. As well, whether respondents  29 identified their study area as having a design or environmental emphasis bore no effect on the mean ratings of images.  These variables were therefore not considered further in our analysis.  The group to group internal reliability coefficient (Cronbach’s Alpha) was quite high at .954 indicating a high degree of confidence that another similarly designed study would yield essentially the same results presented here. An ANOVA was conducted that revealed four statistically significant effects related to the preference for harvest designs (see Table 2.3). Table 2.3 ANOVA results of harvest attributes on preference ratings for the 52 harvest designs and 40 subjects surveyed Variable df Mean Square F-ratio   p Part. Eta2 Geometric Primitive 3 13.955 48.745 .000* .901 Complexity 2 15.532 54.251 .000* .871 Aspect Ratio 2 1.343 4.692 .025* .370 Geometric Primitive x Complex 6 .858 2.996 .037* .529 Geometric Primitive x Aspect Ratio 6 .477 1.665 .194 .384 Complex x Aspect Ratio 4 .185 .646 .638 .139 Geometric Primitive x Complex x Aspect Ratio 12 .267 .934 .539 .412 Error 16 .286 * Statistically significant results at the P = 0.05 level   30 The largest effect was related to geometric primitive: F(3, 52) = 48.745, p < .000, Eta2 = .901. Figure 2.4 illustrates the properties of this attribute and what individual’s design preferences are. The results show that people prefer circular primitives over non-circular ones. Yet, it is also interesting to note that there is some middle-ground between square and circle-type primitives. The results reveal that trapezoids are preferred more than square primitives, and further, that triangle primitives are more preferred than trapezoids.  Figure 2.4 Relationship of mean ratings and geometric primitive showing the mean and 95% confidence interval.  31 The other large effect is due to complexity: F(2, 52) = 54.251, p < .000, Eta2 = .871. Figure 2.5 illustrates how complexity relates to individual preferences. Note that a low level of complexity is preferred less than moderate complexity (about 45% less). However, the difference in mean ratings from moderate complexity to the high complexity is just under 2%, suggesting that most gain in preference for harvest designs can be realized at moderate levels of complexity.  Figure 2.5 Relationship of mean ratings with complexity  32 Aspect ratio was also significant, but with a considerably smaller effect size: F(2, 52) = 4.692, p < .025, Eta2 = .370. Figure 2.6 shows the mean ratings of each of the different aspect ratio groupings. Note that the overall difference between the groupings is quite small compared to the differences of the previous two characteristics.  Figure 2.6 Relationship of mean ratings with aspect ratio  33 There was also a significant interaction of geometric primitive by complexity: F(6, 52) = 2.996, p < .037, Eta2 = .529 seen in Figure 2.7.  Figure 2.7 Comparison of the three levels of complexity for each geometric primitive type showing the mean and 95% confidence interval. This interaction can be broken down into three observations.  First, for harvest designs based on squares, increasing complexity was linearly related to proportional increases in preference.  This improvement in preference was small but constant. Second, when evaluating circular harvest designs, increased complexity also showed continual improvement in stated preference, the relationship of complexity and preference was more logarithmic. For circles, a large preference increase was realized quickly with the jump to medium levels of complexity, but only a small increase in preference was found when complexity increased from medium to high levels. Third, trapezoid and triangle harvest  34 designs also showed large improvements in stated preference with increasing complexity from low to medium levels. For these primitives, increasing complexity from medium to high levels actually decreased stated preference. Further analysis was conducted to investigate if the results from the image ratings lined up with those from the follow-up questionnaire (see Appendix A). When asked to directly rate the importance of each of the three characteristics of design presented in this survey, participants judged complexity to be the most important design factor of the three while rating both shape and aspect ratio as less important (see Figure 2.8).  Figure 2.8 Comparison of the three characteristics of shapes used in the study as rated on their importance to design preference. The table shows the mean ratings with a 95% confidence interval  35 2.4 Discussion It is important to note that shape, as discussed in the literature, is not necessarily synonymous with geometric primitive, but rather may have more to do with complexity. For instance, it is possible to have a shape appear perceptually like a square primitive, where a side can be made of a large number of tiny undulating edges that mimics an overall appearance of a single edge of the square. This edge may be perceived as a single edge, but would be measured as complex using spatial indices. This point calls into question how shape should be specifically defined, whether through particular metrics or elements.  For instance, we found that designs based on circle primitives with curved edges were preferred compared to those derived from squares primitives with right angles.  Harvest designs based on trapezoids and triangles had similar preference ratings on average and were fell between ratings for circle-based designs on the high-end and square-based designs on the low end of the rating scale.  Note that in these cases, designs are made up of edges and primitives, two components of shape. 2.4.1 The Influence of Geometric Primitive This study represents the first known effort to distinguish the effects of geometric primitive as it pertains to preferences of harvest design. Interestingly enough, the results indicate that geometric primitive had the largest effect on preference for harvest design of the three variables tested, specifically that circular shapes were more preferred than any others. This finding supports previous discussions in the literature where researchers have argued that edges of shapes influence perceptions of naturalness (Ode et al., 2009; Ode et al., 2008), which is an important component of landscape character (Tveit, 2009). Bell (1999) also suggests that straight edges may not be perceived as natural compared to varied or undulating edges. Even the BC Ministry of Forests provides incentives in their Effectiveness Evaluation for better shape that fits within the landscape (Marc, 2008),  as it is argued that better shape design can increase visual absorption capability (BC Ministry of Forests, 1997b) and improve the perceived visual impact (BC Ministry of Forests, 1994d). Although shape was not found to be a large predictor of preference in Ribe (2005), the study also considered the effects of retention amount, in which the retention amount was found to have the most  36 significant influence on preference. Our study did not include retention amounts and kept the amount of visible alteration consistent in an effort to focus solely on the elements of shape. When considered simultaneously with higher retention levels, the amount of retention may indeed be more important to the protection of visual aesthetics than shape itself, but that in the case of low levels of retention shape may prove to be a strong predictor as well.  Yet, more research is needed to determine if there an interaction between level of retention and the underlying geometric primitive. 2.4.2 The Influence of Complexity Overall complexity of the harvest designs was also highly significant as a main effect.  Low levels of complexity were clearly not preferred while increasing complexity to moderate or high levels brought about substantial improvements in rated preference. This finding is similar to that found by Day (1967) where people preferred a medium level of complexity to either low or high levels of complexity for randomly constructed polygons. Unlike Day (1967), we found that people still preferred a high level of complexity. However, it is unclear if these results are directly comparable because Day (1967) used abstract shapes on a black and white surface, and our study used shapes within the context of a forested landscape. In Day (1967), the author provides an analysis for two rating components: interestingness and pleasingness. Yet, our question specifically asked users to rate preference for harvest design. It is not known if these distinctions affect the trend we found regarding complexity versus the trend found in Day (1967), but we hypothesize that our results should be similar to the pleasingness dimension. Also of concern are the competing definitions of complexity. Day’s (1967) definition was based on counting the number of sides in a polygon while our definition was a little more loosely defined, but included the number of sides as a consideration. It is easy to count the number of sides in a polygon when each side is straight line from one vertex to the next, but in our study, harvest designs often have rounded edges. In the research by to Day (1967) complexity was measured by the number of vertices, which is also easy to determine for straight lines, but far more difficult for curved lines. This opens up the question of how complexity should be defined as it pertains to human perception within forested landscapes. Given some of the recent research dealing with perceived  37 landscape complexity (Ode et al., 2010; Sang et al., 2008), it may be possible to begin describing complexity as it pertains to landscapes by using metrics often used in landscape ecology such as edge-to-interior ratio, shape and fractal dimension (see McGarigal and Marks, 1995). Yet, again, caution is advised in this matter, as these metrics may not be psychologically relevant. 2.4.3 The Influence of Aspect Ratio Aspect ratio also had an effect on preferences, but the effect sizewas smaller than the two other variables. In general, as the aspect ratio increased, so did the preference for the design. Untortunately we were unable to collect information about how varying aspect ratios would impact preference ratings for other types of landscapes. This study was based only on a single landscape consisting of a relatively high aspect ratio (similar to the aspect ratio of the high-level designs). If another, other  aspect ratio-landscapes would have been used, it may have changed the statistical outcome of this variable. Given the landscape used in this survey it is not surprising that an increase in the aspect ratio of harvests increased the overall rating, as the two might have been perceived as more similar. This effect of aspect ratio on harvest design preferences has not been investigated by empirical research in the literature, so further studies are needed to clarify the influence of aspect ratio on other landscapes where the aspect ratios of the landscapes vary as well. Another discovery was found when comparing the rated importance of the aspect ratio collected from the follow-up questionnaire withthe influence of aspect ratio based on the ratings of the designs. In the questionnaire, individuals were asked to rate each of the three variables in regards to its importance for influencing preference. This question forced users to assume that each of the variables were an element of harvest design. However, the relative impact each variable has upon preferences was left up to the user to decide and rate. The results from this questionnaire provide a quantitative assessment as to the degree of influence each variable has on preferences, relative to the other. The closest statistical equivalent is a comparison of Eta2 for each of the variables. Based on Eta2 for each of these characteristics, we would have anticipated that aspect ratio would have been rated much less on its contributing importance to preferred designs relative to complexity and geometric primitive.  38 So, while individuals rated aspect ratio just as important as complexity and geometric primitive in the questionnaire, the results based on the rated harvest designs suggest otherwise. This discrepancy demonstrates the need for additional research in order to understand the influence of aspect ratio on preference, perhaps by using more diverse landscapes with similar images to examine the importance of aspect ratio. 2.4.4 The Influence of the Interaction between Geometric Primitive and Complexity The interaction of geometric primitive and complexity presents another new discovery, as the relationship between complexity and rated preferences for the four primitives was not consistent across all shapes. Of particular interest is the relationship between preferences and complexity for trapezoid and triangles. We suggest that as complexity of these shapes increases, the perception of hard angles (e.g. right angles) also increases. Therefore, at the highest levels of complexity, trapezoid and triangle harvest designs appeared to be more square-like. At high levels of complexity, the perception of these primitives approached that of square primitives; leading individuals to rate them similar to the high complexity square harvests. 2.4.5 Implications for Forest Management This study provides results that can be linked directly to forest management. Although we were not surprised to discover that geometric primitive and complexity are the two most important characteristics of shape design, we were intrigued by the extent to which complexity played a role. Figure 2.5 shows a non-linear increase in mean ratings across all designs. Operationally, this suggests that an increase in complexity will increase individual preferences, except to the point where non-circular shapes begin to look blocky or square- like or when complexity is taken to absurd levels as is indicated in other studies (Day, 1967). For operational planning this means that the negative visual effects of harvesting cannot be drastically reduced by simply adding a high amount of complexity to the cut block. Yet, increases in preference can be made by adding only medium level of complexity to an atomic primitive. This addition can be more easily integrated into harvest design, resulting in a low operational cost for a high preference gain. Whereas, harvest blocks with higher levels of complexity can be more difficult to plan and may only increase preferences slightly. In areas  39 where the visual integrity of the landscape should be preserved, circular shapes should be the preferred method of harvesting, and should incorporate a moderate to high degree of shape complexity. The results of this study uncover three additional issues for future research that would be applicable to VRM. First, the interaction of shape and percent of visible alteration needs to be explored. This study constrained the amount of visible alteration to within a small range, 13.87%-15.38% of the visible landscape. This represents only one modification class as per BC Ministry of Forests (2001b). Yet to be determined is the extent to which shape can be used to potentially decrease the visual impacts of a harvest while maintaining timber supply. Further studies should explore the effects of shape at lower amount of visible alteration (e.g. 2% - 8%). By varying shape and the amount of visible alteration, it may be possible to find the degree to which shape can influence preferences across different visible alteration amounts. For instance, could a 20% alteration using more natural and complex shapes be more preferred than a blocky shape at 10% visible alteration? This kind of information would be highly useful in practical and operational applications. Second, in situations where retention within harvests exists, it would be helpful to know the influence the shapes within the harvest. Another useful investigation would be the interaction of  how internal retention block shapes interact with shape of the external perimeter to see if there was a way to use design for each of the aspects to improve aesthetics.. Finally, it would be beneficial to have an objective description of complexity as it relates to the perception of harvest shape. There may be some mathematical formulae, or combination of different spatial indices that can provide designers with an objective indicator of shape that can then be correlated with individual preference. The answers to these questions may influence forest operations in visually sensitive issues, and may help to provide a more objective way to measure landscape modifications.  40 2.5 Using Shape Indicators to Measure Aesthetics1 One of the early goals of this dissertation work was to explore the possibility of using a mathematical indicator to predict preferences for harvest shape design. It was hypothesized that indicators often used for pattern analysis in landscape ecology may be able to provide this connection. There was reason to believe that these indicators might be a suitable option given recent research (see Ode et al., 2009; Ode and Miller, 2011; Palmer, 2004). This section is an extension of the work produced for the survey that was not included in the original submission of this chapter for peer-review. The spatial analysis conducted for this section was produced by Fragstats, a well-known and often used program for landscape spatial analysis in ecology (McGarigal and Marks, 1995). This software package provides a variety of shape metrics to quantify the pattern of a single patch. The patch metrics that measure shape are: the Perimeter-Area ratio, Shape Index, Fractal Dimension, Linearity Index, Related Circumscribing Circle and Contiguity Index.. . This mini-experiment,  was created to investigate the possibility of using one or more of these metrics to measure harvest shape to see if these measurements could be correlated with the mean ratings from the 52 images used in the study. If a good correlation existed, then it would provide additional evidence for the possibility of using objective measures of shape to predict rated preferences for harvests of these kinds.  1 This section is an extension to the chapter and was not part of the article submitted for review. Therefore, within this section, a separate analysis, discussion and conclusion are provided as they pertain to the investigation into the potential use of indicators to describe the perception of shape.  41 Data Preparation Before running the analysis,  all original 52 images from the survey needed to be converted to a binary reproduction  where the patch was given one value and the remainder of the image was given another. Figure 2.9 shows an example harvest, and Figure 2.10 shows the conversion of that harvest into binary format.  Figure 2.9 Example harvest design from original survey  42  Figure 2.10 Example harvest design converted from survey into binary representation of patch for Fragstats To derive the binary equivalent, VNS was calibrated to render only the ground, sky and trees (where the latter was a single color). This rendering was then modified in Adobe Photoshop to reclassify the entire landscape: the ground within the harvest area as white, and all other values as gray. The binary format simplifies the setup and configuration process in Fragstats and ensures that the output from the metrics only consist of the patch itself. As discovered in the original study, both geometric primitive and complexity were found to be statistically significant in their prediction of preferred design, so those shape metrics that could potentially describe these patterns were selected. Of the aforementioned patch shape metrics, all were explored for their correlation with the mean ratings from the study. However, the Linearity Index was not used in the analysis because it was not available as an option in the most recent Fragstats software. The following is the list of the metrics  43 investigated and a brief description of what they measure (summarized from McGarigal and Marks, 1995): • Perimeter-Area Ratio (PARA): describes shape complexity by taking the number of edges cells and dividing it by the number of non-edge cells in the shape; • Shape Index (SHAPE): describes complexity by dividing the number of perimeter cells by the minimum number of perimeter cells necessary to create the highest perimeter-area ratio for the number of cells given (as close as possible to a square); • Fractal Dimension Index (FRAC): describes complexity by measuring the fractal dimension, this is a scale-independent metric; • Related Circumscribing Circle (CIRCLE): Like the Shape Index, this metric can help to identify patches that might be both elongated and narrow (or linear). It is not necessarily a description of complexity. • Contiguity (CONTIG): Describes the connectedness of cells within the patch to one another. The parameters for Fragstats included the full resolution of the 3D renders (1280 x 960 pixels), a 1 m resolution and the explicit statement of the background value as the gray color used to represent all cells which are not part of the patch. An 8-neighbor rule was used for the analysis. Upon completion of the analysis using the patch metrics, correlations between each of the metrics and the mean ratings for all images were produced. The results are given in Table 2.4 below. Table 2.4 Correlation of mean ratings and spatial metrics of 52 rendered scenes   Metric PARA SHAPE FRAC CIRCLE CONTIG Correlation coefficient with mean ratings of 52 images .4192 .4010 .4161 .3398 -.3901  The results suggest a moderate to low correlation coefficient, one being negative.. From an operational point-of-view it is likely managers would need more convincing evidence that the  44 correlation coefficients were high enough to warrant using these shape metrics in VRM for determining the aesthetic quality of a harvest’s shape design. Aside from the shape analysis, the results from the survey demonstrate that, depending upon the geometric primitive, mean ratings associated with complexity are quite different between primitives. For instance, adding complexity to circle and square shapes increased the mean ratings. However, for triangle and trapezoid, increased complexity actually resulted in decreased ratings for high levels of complexity. So, by isolating the geometric primitive and focusing on complexity, perhaps the correlation coefficients would also show different outcomes. Table 2.5 shows the correlation coefficients for three measures of complexity (perimeter-area ratio, fractal dimension and shape index) with the mean ratings separated by geometric primitive. Table 2.5 Correlation coefficients of complexity metrics and mean ratings categorized by geometric primitive Shape Complexity Metrics PARA SHAPE FRAC CIRCLE CONTIG Correlation Coefficient of mean ratings by geometric primitive Square .8445 .8275 .8350 .5511 -.8419 Trapezoid .2068 .2214 .2300 .0251 -.2246 Triangle .4634 .4830 .5093 .5093 -.4563 Circle .7185 .7254 .8092 .7593 -.7136  The data in Table 2.5 demonstrates something quite different than in Table 2.4. In Table 2.5 the correlation coefficients for the first four metrics are quite strong for all designs emulating circular primitives. For square primitives the same is true, except that the correlation strength with regards to the CIRCLE metric is lower than the others. A similar coefficient for the CIRLCE metric is observed with regards to the triangle primitive, whereas for the trapezoid primitives the correlation using this metric is almost zero. It is important to note that the CIRCLE provides a higher value as a patch is elongated. So, for circular-based primitives, one can say that as the designs became elongated the mean ratings also increased. However, this effect is not as consistent for triangle and square shapes. For trapezoid shapes there seems to be no correlation between the elongation of the design and preferences for the design. In fact, for trapezoid primitives there seem to be no strong correlation between mean  45 ratings and any of the metrics used. However, for triangle primitives there is a stronger correlation between mean ratings and the first four metrics provided, but the correlations are not high. Finally, for CONTIG, a relationship between mean ratings and this metric seem to mirror similar results of the first four metrics, but are inversely related. In general, circle and square primitives have high negative correlations, triangle has a medium correlation and trapezoid a low. Given these results, one may not conclude that any particular metric is most viable for identifying the complexity of shapes on the landscape, unless the shapes are either square with straight edges or curved with round edges. In these cases, it may be possible to use these metrics. In particular PARA, SHAPE or FRAC metrics seem most appropriate because their mathematical description is more focused on identifying the overall shape, rather than identifying how close the shape confirms to a circle (CIRCLE) or how spatially connected it is (CONTIG) – neither of which may be important to how an individual determines their preference for a harvest design These indicators were originally designed for use by landscape ecologists, so perhaps a few lessons can be learned from the history and use of these metrics within that field. Landscape ecologists are primarily concerned with the relationship between the landscape pattern and the ecosystem process (Cale et al., 1989; Turner, 1989). Metrics are often used to analyze the landscape in order to help ecologists understand what has caused the pattern and what process might be affected as a result of the pattern (e.g. Cumming and Vervier, 2002; Fahrig, 2003; Riitters et al., 2000). However, many times these metrics are often misused. Li and Wu (2004) suggest that such misuse may be the result of conceptual flaws in pattern analysis, inherent limitations of indices, and the improper use of them. The scale and grain (resolution) of the images are known to influence the analysis (Turner et al., 1989), with some indicators being particularly sensitive to resolution and spatial extent (Riitters et al., 2000; Saura, 2004). It is also important to ensure that the scale is sensitive to the process being studied (Lord and Norton, 1990; Vogt et al., 2007).As such, it is crucial to understand how  these problems in landscape ecology might translate to similar problems of using metrics to  assess visual quality.  46 Regardless of whether the analysis is conducted for visual quality or landscape ecology it is both are still, fundamentally, image analysis. Whereas landscape ecologists are concerned with the biological process, landscape designers are dealing with the process of perception. So, problems relating to resolution, scale, and extent which are relevant in landscape ecology will also be important elements to consider in landscape design insofar as problems relating to pattern analysis or landscape classification are relevant to the process of human perception of visual quality.. One issue that landscape designers must carefully consider is the effect of misclassification of elements in landscape design. Numerous researchers and landscape designers have discussed the role of edge, edge density, edge length, etc. on landscape preferences (e.g. Dramstad et al., 2006; Hagerhall et al., 2004; Hammitt et al., 1994; Litton, 1974; Ode et al., 2008; Palmer, 2004; Zube et al., 1982), so it may be crucial to appropriately classify different landscape elements that create the perception of an edge, in order for the spatial analysis to provide relevant results. No studies on the effects of misclassification of edges have been conducted in the landscape design domain, but Langford et al. (2006) has demonstrated that image misclassification can cause errors in indices, with higher classification error occurring along boundaries of different classes for applications in landscape ecology. Therefore, planners and researchers who use photos or develop photo- realistic models to capture a scene and analyze landscape visual quality, must ensure that their classification method deals with problems related to the misclassification of elements in the scene. These problems are also not only an issue at the landscape level, but also at the patch level. Haines-Young and Chopping (1996) state that in landscape ecology, measuring patch shape should be approached with caution as it one of the more difficult characteristics to quantify accurately. They also state that perimeter-area ratios are known to be insensitive to changes in morphology (Haines-Young and Chopping, 1996). So how do these issues pertain to the indices used in the measurement of the 52 harvests used in this section? An important characteristic of the shape metrics used in this section relate to the number of edge cells. Of the three metrics used to measure complexity, fractal dimension, shape index, and perimeter-area ratio all rely on a count of the number of edge cells. The analysis used in this exploration represents only one of many ways in which the harvest edges could have  47 been configured. So, if this research were to continue further, additional edge classifications would need to be developed. For instance, the three possible classifications provided in Figure 2.11 show three possible classification schemes delineated in gray. Each would result in a different outcome.  Figure 2.11 Hypothetical delineation of an example harvest design, to demonstrate the possible differences in edge classification. Although the differences may not be as substantial for designs that contain relatively high perimeter to area ratios, like the one shown in the example, these differences in classification could have major implications for designs with a higher proportion of edge and with a smaller areas. This is one reason why caution should be advised on the use of these metrics as a way to objectively measure shape design. In addition to misclassification of edge, another reason for caution is that adding additional edge, like that of internal retention blocks might effect the correlations with square and circle shapes. If the use of landscape indicators progresses in landscape planning, it will be vitally important to ensure that what is being measured relates to how people perceive the landscape. The same is true for investigating the use of an indicator to measure the design  48 quality of shape. Just as landscape ecologists argue that a metric and scale of assessment must be appropriate for the ecological process of interest, the metric and classification of a harvest must be appropriate to the process of human perception. 2.6 Conclusions The purpose of this study was to understand how different characteristics of shape may influence individual preferences for harvest designs.  The use of 3D photo-realistic software, allowed for the control of three distinct characteristics of shape while eliminating many other possible elements that could have influenced preference ratings. The results from this study offer advice for future harvest design in visually sensitive areas. Landscape architects have argued that more natural looking shapes, as opposed to irregular shapes, that fit within the landform are more preferred by people (BC Ministry of Forests, 1997b; Bell, 2001; Diaz and Apostol, 1992; Ribe, 2005). However, shape has not been well defined in studies pertaining to landscape management, and the literature offers no measurable indicators which express the relative difference of impact for different kinds of shapes on preferences. This study reveals new discoveries about how geometric primitive, complexity and aspect ratio influence preferences for harvest designs. For instance, geometric primitive and complexity are nearly equally important as an influencing factor in perceptual ratings. Also, shapes that were more circular with smooth undulating edges were far more preferred than more blocky harvest designs. Individuals preferred at least moderate levels of complexity, while ratings for atomic shapes were always at the low end of the scale. We suspect that the more primitive the shape the more it seemed out of place in a natural forested landscape, causing a sense of imbalance in the scene. Another discovery was the interaction of geometric primitive with complexity. Of the designs in this study, the more curved and more complex a primitive was, the higher its mean rating. This merely confirms what has been practiced for years in design (Carlson, 1977; Shafer, 1967), but it also provides a way to quantify these differences. Yet, for triangular and trapezoidal primitives an increase in complexity, did not, on average, provide an increase in mean ratings. We suspect this outcome is a reflection of too much complexity which reduces the ability of individuals to distinguish between these primitives and that of the square based  49 designs. The data suggests that individuals do prefer triangles and trapezoids more than squares, but the characteristics that define these primitives become blurred as a high level of complexity is added. These findings provide a way to better understand exactly how shape of a harvest design can be used to influence preferences. Additional research is necessary to verify if these outcomes are consistent across a range of landscapes. From this study we know that a highly complex curved-edged primitive is rated nearly three times as high as the simple geometric square primitive. It is likely that this relationship is not the same for the amount of visible alteration each shape affords, but there may be some relationship between the two. Although there are many elements of harvest design, this study shows that shape can impact the overall aesthetics of the landscape. In scenic areas where forestry operations are active, a harvest design has an influence on an individual’s preferences. This study demonstrates how using shapes with softer and rounder edges can mitigate the overall aesthetic impact by providing a quantifiable difference of how different shapes, complexity and aspect ratio can influence an individual’s preference.  50 Chapter 3:  The Human-Centered Viewshed: An Efficient Algorithm to Evaluate Landscape Visual Magnitude This chapter presents the Human-Centered Viewshed (HCV), which is a geographic information system (GIS)- based method for determining the visible areas and the average degree of visibility for any location in the landscape  given a viewpoint.. In this chapter Visual Magnitude is used to describe the visible space that a particular area on the landscape occupies within an individual’s field of view. Specifically the mathematical representation of the surface of the earth based on its 3D slope and distance to the viewer. The objective of this research is to demonstrate how the algorithm can be used to simulate what an individual might see as they move through the landscape. 3.1 Introduction Visibility or Viewshed Analysis (also referred to herein as viewshed) are one of the more commonly used functions in GIS systems (Davidson et al., 1993). There are applications including landscape management and assessment (Germino et al., 2001; O Sullivan and Turner, 2001; Palmer, 2004; Smardon et al., 1986), for the placement of wind turbines (Möller, 2006), for telecommunications (Sawada et al., 2006) and urban tourism (Wilson et al., 2008), amongst an innumerable number of other applications. The point of using a viewshed in these fields is to prioritize important or significant areas on the landscape that are publicly exposed – whether for possible critique or praise. One of the problems with viewsheds in use today is that they are usually based on binary output, where the binary representation simply shows if a location is visible or if it is not (Fisher, 1996). For applications which are based on human perception, it is not enough to merely consider what an individual will see in the landscape at one time, as is the case with common viewsheds, but also to demonstrate the character of the landform they will see as they move through the landscape. This is particularly important is in areas where the effects of development are of visual concern to people, for instance in places of landscape management, forestry, highway design and tourism. In these fields, it is common for planners to develop designs in planimetric or 2D and render their designs using 3D software to capture  51 the visual effects. This rendering process can quickly become tedious when there are many changes to the plan or when multiple viewpoints must be considered. It is also common for planners to use viewsheds to help derive these 2D plans that will eventually be modeled and ultimately carried out in 3D real-world situations. Since viewsheds are not capable of providing planners with more than just binary representations of visibility, the results may be misleading. In fact, some even suggest that any type of visibility computation is alone insufficient for landscape planning (Ervin and Steinitz, 2003). Even so, viewsheds are often used as a starting point for landscape design as they are readily available in common geographic information systems (GIS) software. One of the setbacks to these available analyses is that they are insufficient to adequately describe the landform as an individual might see it, let alone how the landform is seen as a function of movements in space. Typically, viewsheds only provide information which specifies if a single location (whether as a representative triangle or cell) is visible; it does not provide away to calculate the degree of visibility. The purpose of this chapter is to demonstrate a new viewshed that provides a range of visibility that is based on human perceptions of landform. The algorithm combines an efficient visibility algorithm based on the viewshed analysis, XDraw (Franklin et al., 1994; Izraelevitz, 2003) and a variation of Visual Magnitude (Travis et al., 1975) . Although we recognize some of the problems mentioned in Ervin and Steinitz (2003), this method demonstrates a superior starting point to the common viewsheds available to most planners. The algorithm uses concepts based on early work by Travis, et al. (1975) and Iverson (1985) where each proposed methods for measuring how visible a given object was from a viewpoint and combines these with an efficient visibility analysis (Franklin and Ray, 1994; Franklin et al., 1994; Izraelevitz, 2003; Wang et al., 1996). The primary contribution of this work is that it demonstrates an efficient way to produce a standard raster-based 2D output that supplies relevant 3D information based on a mathematical representation of human vision. Furthermore, this algorithm is based on the critical component of human perception and the limitations of the human eye, the result is a measure of visibility that ranges between not visible to fully occupying an individual’s field of view. The combination of these two elements provide the kind of spatial information  52 important for planners dealing with issues pertaining to human perception and the visual effects of development of landscape change. In order to demonstrate these contributions, an example depicting a typical cruiseliner route through the Inside Passage of British Columbia is presented. This example shows how the method can be used to help planners identify sensitive areas of the landform, where negative visual impacts may be more likely to be seen by a group of tourists passing through the landscape. 3.1.1 Visibility Analysis and Algorithms De Floriani and Magillo (2003) provide a survey of many of the types of algorithms in production, including analysis based on raster grids, irregular triangulated networks and regular triangulated networks. De Floriani et al. (1999) also presents the wide variety of the applications for which viewsheds have been used. There are two predominate types of viewsheds: those produced using Triangulated Irregular Networks (TINs) and those with Digital Terrain Models (DTM) or raster grids (Maloy and Dean, 2001). One advantage with TINs is that they typically require less space in memory to compute and may be more accurate than raster-based grids (Evans et al., 2001). Evans, et al. (2001) demonstrates that in regards to memory and error (accurate approximation of the surface), at a low spatial resolution the TIN is a better approximation of the surface, whereas at a higher resolution the grid-based approach contains less error. In either case, algorithms that use DTM data to produce viewsheds remain the more prominent method in landscape management and are the focus of this article. With DTM viewsheds, the output from an analysis tends to be of binary nature; either visible or not visible (Burrough et al., 1998). Calls have been made to develop ‘fuzzy’ viewsheds (e.g. Fisher, 1991, 1992, 1993, 1996), which provide a probabilistic or likeliness of visibility rather than a binary result. As a result, some alternative methods have been developed (e.g. Franklin and Ray, 1994; Franklin et al., 1994). However, these algorithms are limited in the literature and most tend to focus on binary representations. Other research in viewshed analysis have been conducted, such as: viewsheds developed to find the fewest number of observers necessary to yield  yield the highest amount of visible landscape (Franklin and Vogt, 2004) which could be used for the optimal placement of viewpoints (Kim et al., 2004)  53 such as for telecommunication towers. Other examples include: an efficient algorithm, which at the time was 3.5 times faster at conducting a viewshed analysis for datasets that were larger than the available RAM (Magalhães et al., 2007); also, methods with variations from the typical Direct method (see Izraelevitz (2003)), such as a sweep based algorithm which used a similar approach than the Direct method, but with increased efficiencies (van Kreveld, 1996); a partial visibility algorithm that provides an amount of visibility for a given cell (Sorensen and Lanter, 1993), and as well an algorithm that approximates viewshed by focusing on specific terrain features (Rana, 2003). While progress in efficiency has been made, viewsheds remain focused on simple outcomes emphasizing which cells are likely visible. 3.1.2 Visual Magnitude Using viewshed to identifying what is visible is a necessary component of landscape management, but due to its standard binary approach is limited in its ability to convey crucial information about the landform as seen in perspective view. Visibility algorithms can determine if a cell is visible, but are unable to determine the degree to which a cell is visible. This problem likely stems from the natural development of GIS as a geographical-based view of space, rather than a human-based view of space (Llobera, 2003). To address these shortfalls, cumulative viewsheds (sometimes referred to as times seen) have been produced by combining viewshed results from different viewpoints (e.g. Fisher et al., 1997; Travis et al., 1975; Wheatley, 1995). This is merely a function of adding the results from numerous binary viewsheds produced from several viewpoints. Using this technique, each cell is comparable by the number of times each location is seen, but even though they provide a way to compare between cells, they still lack key landform information. In applications where viewsheds are intended to provide locations to maximize visibility (Sawada et al., 2006), understanding the perspective view is not terribly important. However, when viewsheds are used to understand what people will see on the landscape, such as in forestry or highway design, it is crucial that they provide information that pertain to human vision. Visual Magnitude (VM) focuses on providing this key information. It does so by representing certain characteristics of an object and an approximation of its amount of visibility within an  54 individual’s field of view. In the case of DTMs, VM provides a degree of visibility for a particular cell relative to the other cells in the DTM. The concept has existed for some time in visual design. Travis, et al. (1975) developed the computer program, VIEWIT, which could calculate these values. Iverson (1985) documented VM more thoroughly, showing how these calculations could be produced and the rationality driving them. In the original VIEWIT program, VM was referred to as visual perception sensitivity, and based on the report provided by Travis, et al. (1975), visual perception sensitivity is identified as a combination of: 1) distance, 2) aspect relative to the observer, and 3) times seen. The descriptions and methods have been maintained in (Grêt-Regamey et al., 2008; Grêt- Regamey et al., 2007), with others using the concepts as well (Bishop, 2003; Bishop et al., 2000; Caldwell et al., 2003; Chamberlain and Meitner, 2009; Shang and Bishop, 2000). Visual exposure (Llobera, 2003) and apparency (Fairhurst, 2010) are also based on using the relative aspect of the surface to the observer and its distance. Although VM has been used for measuring impacts to viewscapes and for assessing aesthetics, these applications have focused their efforts on small areas and from single or few observer locations. In order to begin understanding the influence of landscape development and the potential visual impacts as visible by someone moving through a larger landscape, it is necessary to develop an efficient method that can accurately represent the landform as seen from multiple viewpoints. The algorithm proposed, combines an efficient viewshed algorithm with an adaptation of the original concept of VM that more precisely calculated the degree of visibility the landform. 3.2 Methods 3.2.1 Notation and Assumptions Before diving into the methods, we clarify our notation and assumptions. First, assume x and y represent the Cartesian coordinates on a plane with z as the elevation above this plane. Second, assume that in these calculations, the curvature of the earth was accounted for when calculating the z value.  55 The notation is as follows: • pv  Viewer’s position in 3D space • pt  Center point on the target DTM cell • n  Normal vector of the cell containing pt • v   View vector from pt to pv • β  Angle between the projection of v and n in the vertical xz-plane • θ  Angle between the projection of v and n in the vertical yz-plane • u  A cell of the DTM • vs(u)  Angle that the vector from pv to u makes with the xy-plane • d  The cell diameter of the discrete data • VM(x,y) Visual Magnitude Continuous Surface; (x,y) is single cell value 3.2.2 Visibility and Viewshed The visibility analysis used in this paper is based on an approximate viewshed using XDraw (Franklin et al., 1994; Izraelevitz, 2003). XDraw is an extremely efficient viewshed algorithm that calculates values for a DTM cell only once. The typical method of calculation, or the Direct method, is far less efficient, with decreasing efficiency as the search radius increases. Izraelevitz (2003) provides a good comparison of the differences between the two methods.  56 Since the earlier references provide a substantial overview of the differences between the Direct method and XDraw, only a brief description will be provided here. The Direct method identifies all possible view vectors in the xy-plane to the most distant cells within the search radius. Then, it calculates the visibility for each cell along each view vector. Figure 3.1 depicts an example view vector. Starting from the viewer’s position, pv, and working toward pt, the projection of v crosses seven cells. The penalty is that to calculate visibility for all cells crossed by all view vectors, the visibility for cells nearest to the pv will be calculated numerous times. Along each viewing direction from pv in the xy-plane, the angle above the xy-plane of the view vector to each cell along vs is calculated. So, in Figure 3.1, if vs(6) > vs(5), vs(4), …, vs(1), then cell 6 is visible.  Figure 3.1 Depiction of the Direct method for visibility analysis. The black line shows the projection of a view vector in the xy-plane The cells labeled 1 to 7 are used to calculate the visibility of cell 7 from pv. XDraw calculates each cell’s visibility only once by iterating through concentric rings, beginning at cells nearest the viewpoint and extending outward. It does this by calculating vs(u) and comparing it to the maximum interpolated angle above the plane that all inner rings make in this view direction. This eliminates redundant visibility calculations for cells near pv. Figure 3.2 shows an example of how vs(E)  is compared with the previous ring. In  57 the case of vs(E) , the projection of v crosses ܥܦതതതത, at the point 0.6*vs(C)+0.4*vs(D). Therefore, E  is deemed not visible if vs(E) is less than 0.6*vs(C)+0.4*vs(D).  Further, XDraw maintains the maximum interpolated vertical angle in every viewing direction from pv and updates it when considering the new ring. Upon completion of all rings, a final binary viewshed is produced. This method is known to provide an increased efficiency compared to the Direct method as is further detailed and depicted in Izraelevitz (2003).  Figure 3.2 Depiction of XDraw method which calculates visibility by concentric rings (shown by the numbers located on the left and bottom cells. E is visible if its’ vs(E) is greater than the interpolated value of the previous ring. pv  is the observer point and pt is the target location. At this point, the viewshed methods have only distinguished between cells that are visible and those that are not, however in reality some visible cells are more apparent than others due to their proximity and angle relative to the viewer. This is where Visual Magnitude can be used simultaneously along with the viewshed to calculate the influence of landform on visibility. 3.2.3 Visual Magnitude: The Human-Centered Dimension Visual magnitude is calculated from a single viewpoint to a single location on the earth by incorporating the viewing angle relative to the surface slope, as well as, the distance of the  58 viewer to the location on the earth’s surface. The technique is demonstrated in Iverson (1985), Grêt-Regamy, et al. (2007), and Travis et al. (1975). Llobera (2003) and Domingo- Santos et al. (2011) refer to a similar concept called visual exposure, where, in the latter reference visual exposure is calculated to include atmospheric corrections and the influence of background color as well. The function of viewing angle relative to the earth’s surface slope can be presented as α, where this is the angle between the viewing vector (v) and the normal vector (n) to the surface plane of the DTM cell. This angle satisfies: cos(α) = v · n where we assume that v and n are unit length vectors. However, in our work, it is convenient to express the normal to the DTM cell’s surface as two angles: the angle β  between the projection of v  and n in the vertical xz-plane (the plane for which the slope gradient or degree of slope is determined) and the angle θ  between their projection in the vertical yz- plane (the plane for which the orientation, or cardinality is determined). Therefore, we may then obtain cos(α) using the equation: cos(α) = cos(β) * cos(θ) To calculate β  and θ , the cell’s surface angle in the vertical xz-plane, as well as the angle in the vertical yz-plane are calculated using a 3x3 kernel (Burrough et al., 1998; Shi et al., 2007). The method from Shi et al. (2007), which calculates a cell’s tilt on both planes, was used instead of Burrough et al. (1998) as it has been shown to be a more accurate representation of the landscape surface. A number of methods could have been used for deriving these angles, such as deriving n using the average of the cross products of two adjacent sides of the cell with the other pair of adjacent sides (Corripio, 2003). Although this method is slightly more mathematically accurate than the 3x3 kernel, the differences are extremely subtle and nearly non-existent for higher gradients slopes, which are of particular importance for VM. Considering these minor differences, we opted to use the kernel method due to its use in conventional GIS software. Figure 3.3 and Figure 3.4 show the angles used to calculate VM. The first figure assumes that the slope of the surface has been determined by using the 3x3 kernel. From there the cos(β)  is use that t there Pract of the when orien Figur pt. Th Figur a dire The f |࢜|. N dimin the ca autho distan arc m d to describ he orientatio the cos(θ) i ically, what  surface is p  vs is perpen tation of the e 3.3 Diagram is plane show e 3.4 Diagram ction with pt a inal element aturally, as ishes. Sever lculation (B rs providing ce and state inute of visu e the degree n or aspect s used to de this provide arallel to vs dicular to th  surface (wh matic explana s the surface s matic explana t its center.  in the calcu  the distance al papers do ishop, 2003  functions. s that the vi al angle. Iv  of visibility of the cell h scribe the d s is a value . Likewise, t e slope and en both v an tion of β. Thi lope of the ce tion of θ. Thi lation of VM  away from cumenting ; Iverson, 19 First, in Iver sual limitati erson (1985  in the verti as been dete egree of visi of 0, or no v he maximum when the vi d n are poin s is a 2D sidev ll. s is a 2D sidev  relates to  an object in VM have al 85; Llobera son (1985) t on of a pers ) draws a re cal xz-plane rmined by u bility in the isibility, if e  visibility ewing orien ting directl iew of a sing iew of the ter the Euclidea creases, the l referred to , 2003; Tra he author d on with 20/2 ference from . The latter sing the 3x3  vertical yz- ither the slo from these d tation is fac y at one ano le terrain cell rain grid cell n distance f  ability to se  distance as vis et al., 19 iscusses the 0 vision is l  Maertens ( figure assum  kernel. Fro plane. pe or orient imension oc ing the ther).  with its cente , where the ce rom pv to pt e the object an influence 75), with tw effect of ess than a o 1884) wher 59 es m ation curs  r at  ll has  or   in o ne ein  60 the author states that no object is perceivable when its distance is 3450 times its size. In modern terms, we understand this as being the limits of the central area of the eye’s fovea. For most intents and purposes, this equates to the smallest visual angle of 1 arcminute, or .0167° (see Strenström (1964)). Measuring the perceptual size of an object with a known size and distance can be done by adopting the concept of angular diameter, commonly used in astronomy. The formula for angular diameter is: ܽ݊݃ݑ݈ܽݎ ݀݅ܽ݉݁ݐ݁ݎ = 2 ቆtanିଵ ቆ12 (݀/|࢜|)ቇቇ The reliability of this formula is well documented and used in physics and astronomy for measuring celestial bodies (Giovannini, 2008), but it should also be relevant to the limits of the human eye or optics, in order for it to be used in the calculation of Visual Magnitude. Let us use the example from Iverson (1985) in which he suggested that an object of 1.5 ft2 at one mile (5280 ft) is essentially the maximum distance that this object could be seen. It stands to reason that if the limit of human vision is roughly .0167° of angular diameter, then the angular distance of this example should approach this limit, and in fact it does: 2 ቆtanିଵ ቆ12 (1.5/5280)ቇቇ =  .016° So, in Iverson (1985), the author argues that angular diameter squared should be the function for distance. According to Iverson (1985), this function served as the distance factor in the program VIEWIT(Travis et al., 1975), a program written for the land-use planning by a group of individuals from the U.S. Department of Agriculture. More recently another alternative has been suggested by Grêt-Regamy, et al. (2007) where the authors factored in distance by: d2 / |࢜|2. This method represents the modern approach of calculating the effect of distance in computer graphics, such that as an object moves away from the viewer the area the object occupies in the viewer’s field decreases in proportion to the square of the distance. In fact, the result of these formulae are nearly exactly the same, with differences only occurring when the distance is less than five times the size of the object. To ensure that these calculations were valid, the authors compared the size of known  61 objects on a landscape (counted by the number of pixels from a digital photograph) with the sum of the VM values calculated for all visible areas in a GIS (Grêt-Regamey et al., 2007). The result was a high correlation (R2 of .81) between the calculated values between pixel counts and VM values. Although we acknowledge the original formula presented in Iverson (1985), we have chosen to use the function given by Grêt-Regamy, et al. (2007) in this paper for two primary reasons: it is a well-documented case of validation, and it is a common formulae used in computer graphics. We further acknowledge that either of these formulas may serve well to provide an accurate VM result, but since we are generally concerned with landscape level distances at greater than five times an objects size, the use of the distance squared seemed most practical. Therefore, we chose to calculate Visual Magnitude of a single cell from a single pt as: ܸܯ(ݔ, ݕ) = (cos β)(cos θ)( ݀ ଶ |࢜|ଶ) Where |࢜|, d, β and θ are all non-negative. 3.2.4 Details of the Study Area The study area of this research is located in the waters between Vancouver Island and the West coast of mainland British Columbia, called the Inside Passage (Figure 3.5), much of which falls within the visual protection of the BC Ministry of Forests. The Passage is a well- known tourist route for cruise liners and ferries. It consists of wide and narrow passes surrounded by numerous islands, mountains, dense forests, rock and incredible scenic vistas. The terrain is quite mountainous with several visible peaks; the highest just under 1900 m. The area of the study encompasses roughly 65 km x 75 km; the route itself roughly 85 km long. The Digital Terrain Model (DTM) and ocean boundaries were obtained from province of BC (BC Integrated Land Management Bureau, 2011) and the cruise route was manually drawn based on available online maps.  62  Figure 3.5 Inside Passage study area for the analysis of the HCV. The cruise route is shown in white, with the oceans in blue and a hillshade to depict the terrain. 3.2.5 The Process The study aims to demonstrate how the combination of XDraw and VM can be used to construct a viewshed that produces a probability surface of the visual space that any one cell will occupy within an individual’s view at any point along a given route. To accomplish this, several tests were conducted to compare the processing times between the Direct method and XDraw with VM simultaneously calculated. The tests compared four different spatial resolutions ranging from 100 m to 12 m. First, DTMs were generated for each resolution. Then viewpoints were designated along the ferry route, separated by the ground distance of each resolution , resulting in nearly 2500 individual viewpoints for the 25 m test and  Nearly  63 one-quarter of those viewpoints for the 100 m test. The viewpoint heights were kept the same throughout the study locations and were based on an estimate height of the top viewing deck of a cruiseliner. To function, the algorithm requires the following inputs: • Digital Terrain Model • Viewpoints, specifically: o An x,y location on the surface o An offset height above the surface o A weighting of significance The viewpoint weights are intended to allow for situations where a user may want to assign a higher weight to one viewpoint versus another. For instance, along a highway, each viewpoint may be given a weighting of 1, but a scenic stop along the highway may be assigned an increased weight so its significance in the analysis is weighted according to its significance. In this analysis, a weighting of 1 was assigned to all viewpoints. The process is performed by looping through all viewpoints, calculating VM for each visible location on the DTM from a single viewpoint. Simultaneously, the weighted sum of these values are then added together to provide the final result. The result provides a VM value for a single cell on the terrain, weighted by its assigned significance and its potential visibility. The following the formula for calculating VM for a single cell: ܸܯ(ݔ, ݕ) = ෍(cos β)(cos θ) ቆ ݀ ଶ |࢜|ଶቇ ൬ ݓ௜ ∑ ݓ൰௜ ௝ ௜ୀଵ  where j is the number of viewpoints, and w is the viewpoint weights. 3.3 Results The output from the algorithm is a raster of the entire extent of the study areas ranging in values from 0% to 100%, either as an objective or stretched (min/max) value. Obtaining the maximum possible objective value would infer that a single cell is facing the individual and encompasses the entire viewer’s visual area – for all viewpoints. In reality this is not plausible, so for all outputs shown herein, the objective values have been normalized o more effectively communicate the data. The cells with the highest normalized values are  64 represented in the darkest shade and the lowest values by the lightest shade of color. Any value of zero is removed from the figures as it represents a cell which is never visible from any of the viewpoints. What remains is a greyscale hillshade of the DTM. 3.3.1 Outputs Two depictions of the VM result are shown in Figure 3.6 (a 3D representation) and Figure 3.7 (a 2D representation). Figure 3.6 provides a picture of the terrain, the route, and shows the VM map overlaid on the surface. The steep slopes along the shore bank continuing up to the mountain tops become very apparent in the image. The black outline designates the areas shown in Figure 3.7B. And the thin white line represents the area depicted in Figure 3.7A. With the 3D example, the relationship of VM with regard to the landscape characteristics like steep banks and mountain slopes becomes easier to see with steeper areas facing the route identified by a dark orange color. The same is true for lower lying or flat terrain shown in light orange. For all figures, the output is shown with 25 m cell resolution. The proximity of the route (white) to the terrain spans between 500-800 m. VM is represented by the gradient of orange, with the darker levels showing higher visibility.  Figure 3.6 Example of a larger portion of the Inside Passage in 3D. The route is shown in bold white. The thin white line and thin black line represent Figure 3.7A and B respectively. The gradient of orange represents VM values, with high values as dark orange. The output cell size is 25 m. Gray is not visible.  65  Figure 3.7 Example of Visual Magnitude demonstrated for a small area of the Inside Passage (A) and objective VM values (B). The route is shown in bold white. The orange gradient represents VM values with high values in dark orange. The output cell size is 25 m. The black outline in A is shown at a larger scale in B. Locations not visible are replaced with a hillshade. Note that the output in both figures includes an analysis from all viewpoints along the route for the entire Inside Passage, not just for those areas depicted in the figures. This can be illustrated by a very the light orange color on the backside (or topmost) part of the large island in Figure 3.7, showing that VM is not low. This influence is due to the viewpoints that extend along the route and to the west. Remember that the VM calculation provides objective values ranging from 0 to 100%. Also that 100% would require a space on the landscape to completely occupy an individual’s view for every viewpoint. When the analysis is performed over large areas with viewpoints well- spaced, the objective values may end up being so miniscule that they are too small to store within 32-bits (the default output). So, it is helpful to use stretched values, with the maximum value obtained for all viewpoints being 100% and the minimum (or not visible at all, being 0). So the gradient that you see is based on the stretched objective values; the colors themselves are also stretched to provide a clear representation of the data. Figure 3.7B shows  66 the objective values for each cell on the terrain for a subsample of the area shown in Figure 3.7A. These are the stretched objective values. Finally, Figure 3.8 is provided to demonstrate the differences between VM and that standard binary viewshed. In this figure, any area visible is represented in a light orange shading, and areas not visible are shown in gray. Unlike the VM output shown in Figure 3.6, there is no way of understanding the relative difference in the degree of visibility for any cell, except whether a cell is visible or not.  Figure 3.8 Example of standard viewshed analysis conducted for the entire route, with hill shading to provide terrain perspective. Light orange values represent visible areas. The route is depicted in white. 3.3.2 Computational Examples Efficiency tests were conducted for a range of resolutions using both the Direct method and XDraw method for the viewshed. Four DTMs were tested, ranging from 100 m to 12 m (by halving the resolution), encompassing the same area, with the number of viewpoints varying by the resolution factor. Of the four, the minimum number of viewpoints is 768 (for 100 m) and the maximum number is 6434 (for 12 m). The tests were conducted using a PC with an Intel i7 processor, 3.20 Ghz, 24GB of RAM running a 64-bit operating system. One of the processor cores was completely devoted to the calculations (by threading) so that no additional computer functions were operating on that core. Table 3.1 provides details on the range of resolutions, number of viewpoints and compares the efficiency of the algorithms.  67 Table 3.1 Comparison of results for XDraw and Direct methods. Results are based on four different spatial resolutions (12 m – 100 m) and a varying number of viewpoints. Abbreviations are: t = time, tn = time normalized (time as a function of resolution), vp = viewpoint. xDraw Method Direct Method Resolution (meters) Dimensions # vp Total t t / vp (sec) tn / vp (sec) Total t t / vp (sec) tn / vp (sec) 12 4890 x 5742 6434 24.28 hrs 13.59 0.21 42.02 hrs 23.51 0.37 25 2445 x 2871 3086 2.86 hrs 3.34 0.21 4.99 hrs 5.83 0.36 50 1223 x 1436 1540 21.42 min 0.83 0.21 38.13 min 1.49 0.37 100 612 x 719 768 2.74 min 0.21 0.21 4.91 min 0.38 0.38  The results suggests a noticeable efficiency increase by using the XDraw implementation (roughly 45%). The increase fits with the theoretical running time (n) of XDraw (n) compared with the Direct method of (n log n). This efficiency enables time savings for an analysis which requires a large area to analyzed or for an analysis with a large number of viewpoints. This is especially useful when needing to identify the amount of space any cell might occupy within an individual’s view throughout an entire route. In fields where the landscape is visually important and experienced through time, this is an essential planning tool. 3.4 Discussion Visual Magnitude provides a unique way to understand the landscape as it may be seen by people. Unlike a standard viewshed analysis, which provides a binary output of visibility (visible or not), VM indicates a degree of visibility as seen from the perspective view. In this case the view from along a cruise route. Simply put, VM can provide a way to identify areas on the terrain that are more apparent, obvious or potentially attention grabbing. The output shown in Figure 3.6 demonstrates how VM provides a unique perspective about the  68 landscape that is not part of binary viewshed available standard GIS software (shown in Figure 3.8). VM can be explained by an illustration of a candle in a dark room. The standard viewshed would identify all those areas in the room that reflect any light, while VM would consider the amount of light reflected. However, VM should not be confused with techniques where luminance is concerned. VM has nothing to do with measurements of light, or relief shading from light (e.g. Zakšek et al., 2011), it strictly measures the landform as it would be visible to an individual. VM differentiates from the standard viewshed because it provides a way to measure the amount of visibility of any one cell, rather than just if a cell is visible or not. The examples shown in Figure 3.6 and Figure 3.7 demonstrate how higher VM values are at areas which are steep facing and near the route, suggesting a higher degree of visibility compared to low-lying terrain. Using these characteristics, VM is able to produce an analysis which clearly identifies those areas on the terrain that garner more attention (based solely on terrain characteristics aside from ground cover, foliage, etc.). So, unlike binary forms of a 2D raster-based viewshed (Fisher, 1996), VM includes distance, slope and aspect relative to the viewer and associates the calculation to an individual’s limits of perception (Iverson, 1985). Although VM has been used in landscape management before, there are potential validation issues relating to the proper and most ideal method for calculating the influence of distance. Two formulas have been proposed (Grêt-Regamey et al., 2007; Iverson, 1985), with the most recent providing a well-documented validation case. However, without a more systematic approach for validating these formulae with human vision, we may not understand what the most appropriate function should be. One approach for validating the effectiveness of the formula is to use the method offered in Grêt-Regamey et al. (2007), where the authors used pixels counts of known objects as captured by a camera to verify if the number of pixels correlated with the VM values calculated for that object. Even though they found a high correlation using this method, the authors wonder what may improve the method.  69 One thought is that the method for conducting the correlation is slightly and inherently flawed. The capacity of the human eye to perceive an object in the near-ground versus an object in the distance is far greater than a camera’s resolution with a fixed lens. That is, the number of pixels representing the object near the camera is more accurate than the number of pixels representing an object of a large distance away from the camera. In addition, the further an object is away from the focal distance, one would notice a decrease in accuracy for pixels counts of an object because the pixels that represent and object would begin to blur with pixels from an adjacent object. One way to assess this would be to compare correlations with different images captured at different resolutions. It may be that the higher the resolution, the higher correlation. Alternatively 3D simulations could be used because these simulations can eliminate much of the edge blurring problems cause with cameras. Nevertheless, one cannot disregard the high correlation between the authors’ pixels counts of an object and the calculated VM, suggesting that the author’s formula is an appropriate account of the effect of distance in their study (Grêt-Regamey et al., 2007). As the demand for an analyses covering large areas becomes more common (e.g. Germino et al., 2001) or the demand for analyses with high resolution data such as LiDAR increased, an efficient algorithm becomes increasingly necessary. In addition, when these calculations are to be completed using numerous viewpoints across viewsheds landscapes, the importance is further magnified. Consistent with the results found in Franklin et al. (1994), this algorithm provides an increase in efficiency over the Direct method. 3.5 Conclusion Arguably, many fields would benefit from a viewshed which expands the notion of visibility beyond a binary representation of visibility, such as landscape design, forestry, highway design, the tourism sectors and even ground-based military application, as current viewshed methods lack important 3D information about the landform. The HCV presents a much- needed improvement over the standard viewshed analysis, both in terms of efficiency and in providing a relative degree of visibility between areas on the landscape. In landscape design or forestry, viewsheds are used to identify visible areas along highway corridors or where harvest activities might be seen (Fisher, 1996), this algorithm will enhance a planners ability  70 to better understand not just what is visible, but the degree of visibility relative to other locations in an individual’s perspective. In highway design, it can help determine important scenic stops, or aid the management of roadside obstructions which influence aesthetics for nearby and distant landscapes. For military applications, like many viewsheds before (e.g. Fisher, 1992; Franklin and Ray, 1994; Franklin et al., 1994; Goodchild and Kemp, 1990), this algorithm is well suited to more precisely identify locations where troop movements are less likely to be spotted by ground-based troops of the opposing force or to detect locations of increased prominence. The HCV combines a measure of Visual Magnitude with the efficiency of XDraw to create a way to quantify how the landform may be ‘seen’ by an individual as they move through the landscape. As with any binary solution, this algorithm is still suited for all applications where a standard viewshed is often used. This method represents a shift away from geographically- based GIS tools toward a human-centered approach, providing vital information important to making better decisions with people as the focus.  71 Chapter 4:  Geospatial Techniques and Considerations for the Management of Visually Sensitive Areas in British Columbia Landscape aesthetics play an important role in forestry. In BC, the Ministry of Forests has identified aesthetic areas in the province which are called “visually sensitive areas”. They are designated to provide some level of visual protection against human-induced disturbances. The BC Ministry of Forests outlines how forestry companies can manage these sensitive areas (Marc, 2008), and requires that modifications from forest activities be limited to specific visual quality requirements set by the Ministry (BC Ministry of Forests, 2004). A major part of this management deals with visual quality assessments, completed by the Ministry to ensure that the integrity of scenic areas is being upheld. With geospatial tools, the management and assessment processes can be made easier and more objective. This chapter provides a demonstration of how geospatial technologies, specifically Visual Magnitude, can be used to automate assessment and improve management of these visually sensitive areas by helping landscape planners be more efficient in designing their harvests. In addition, some considerations about viewshed analysis and the importance of viewpoints for the assessment are also discussed. The BC Ministry of Forests has established a process for ensuring that visually sensitive areas are identified and maintained throughout the province. This process consists of six phases, from the development of the Visual Landscape Inventory (VLI) to the Monitoring of these visually important areas. The phases are presented in Table 4.1.  72 Table 4.1 Phases of visual landscape management process (adapted from BC Ministry of Forests, 1997b) Phase Outcomes 1 Visual Landscape Inventory a. Delineation and classification of the provincial landbase into visually sensitive/not visually sensitive areas b. Delineation of visually sensitive areas into visual sensitivity units (VSUs) and their classification into visual sensitivity classes (VSCs) 2 Recommended Visual Quality Classes a. Assessment of implications and options b. Recommendations, including recommended Visual Quality Classes c. Modeling of current management practices for timber supply reviews 3 Approved Visual Quality Objectives a. Identification of scenic areas b. Establishment of VQOs c. Approval of operational plans 4 Visual simulations and design solutions a. Visual simulations b. Visual impact assessments c. Visual landscape design solutions 5 Achieved visual conditions a. Achieved visual conditions 6 Program audits a. Monitoring/inspections of field activities  This chapter focuses on addressing three ways in which geospatial technology can be used to increase the effectiveness of the visual landscape management. These three demonstrations span several the phases of the process. The first suggestion shows how the HCV can be used at a provincial scale to help identify visually sensitive areas, and in the delineation of visually sensitive units (Phase 1). The second suggestion shows how the HCV might be useful for estimating the amount of visible alteration of a harvest plan, an important component of the management and assessment of harvests in visually sensitive areas. This would enable planners to help meet the Visual Quality Objectives identified in Phase 3, more efficiently design harvests in Phase 4, and increase the likelihood of achieving the visual conditions in Phase 5. The third suggestion focuses on one of the systematic problems inherent in the visual landscape management process. Currently the process relies heavily on the analysis of visual quality from a limited number of viewpoints. This final suggestion provides commentary and considerations for planners to take note of when using viewpoints, and also offers suggestions for how the process could be changed.  73 4.1 The Visual Landscape Inventory The BC Ministry of Forests maintains an inventory of locations throughout the province, depicted earlier in Figure 1.3, that have been identified as visually sensitive areas. These occupy a substantial amount of forested area, approximately 14.6 million hectares throughout the province and over 25% of all forested lands in BC (BC Ministry of Forests, 2010). Visually sensitive areas represent places where the public may be interested in protecting aesthetics. These areas have been deemed important because of their proximity to communities, well-traveled destinations, and for tourism and recreational purposes. One of the earlier descriptions of visually sensitive areas can be found in Amir and Sarig (1977), where they mention six landscape types that are associated with these areas: cliff areas, steep rocky slopes, exposed plateaus, prominent peaks, valleys and pine forest areas, with the most sensitive as the cliffs, valleys and forest. More recently, visually sensitive areas are also called “high stakes” places, such as areas that are waterfront, ‘scenic areas’ and historic districts (Kearney et al., 2008). The B.C. Ministry of Forests defines visually sensitive areas as places identified by unique areas or corridors in the province, which “...could give rise to concern if their visual appearance were altered by forest practices or other resource development activities” (BC Ministry of Forests, 1997b). These tend to be places seen by numerous viewers and have some inherent scenic value (BC Ministry of Forests, 1997b). More specifically, these areas are “viewsheds that are visible from communities, public use areas, and travel corridors – including roadways and waterways – and any other viewpoint so identified through referral or planning processes” (BC Ministry of Forests, 1997a). These ‘other’ viewpoints may not be located in well-traveled places and may not even be readily accessible, but for particular reasons may be identified as sensitive. Clearly, the definition of these areas is quite broad and allows for subjective interpretation. In BC, identifying a visually sensitive area (refers to Phase 1a in Table 4.1) is done by distinguishing at least one of thirteen social or biophysical criterion (BC Ministry of Forests, 1997b), which are: • “Areas visible from communities, public use areas, or travel corridors; • Areas seen by a large number of viewers;  74 • Areas where public expectation for scenic quality is well above average (viewshed around back country lodge, tourism destination, highway rest stop, area adjacent to a Forest Service trail/site); • Areas containing regional or local topographic features that are valued by the public; • Areas that possess inherent visual or scenic value; • Areas identified as visually sensitive or scenic through referral or planning processes (e.g. Commission on Resources and the Environment direction; land and resource management plans; local resource use plans); • Areas identified by previous Visual Landscape Inventories; • Areas of proposed new highway routes or changes to highway alignment; • Areas visible from important high elevation viewpoints; • Areas identified by tourism operators or by MSBTC [Ministry of Small Business, Tourism and Culture] as important for tourism; • Areas adjacent to high-use Forest Service roads which lead to popular recreation areas; • Areas around important recreation features that attract the public; • Other”  Using these criteria as a guide, these areas are delineated on a map, using a scale of roughly 1:250,000 to broadly identify potential locations. This delineation is completed by Ministry experts, with input from other organizations or stakeholders where appropriate. These criteria and the process of identification have been criticized as being too subjective, and therefore vague (Dakin, 2003), based on subjectivity. For instance, according to Dakin (2003) “areas seen by a large number of viewers” lacks an objective and measureable description. Yet, whether measureable or not, these 13 criterion serve as the backbone for the Ministry’s entire spatial database of important scenic areas. Figure 1.3 (see page 8) shows these areas as per the current Visual Landscape Inventory (VLI) data base (BC Integrated Land Management Bureau, 2011). The other component of the VLI is the discretization of visually sensitive areas into separate units (see item 1b in Table 4.1) called Visually Sensitive Units (VSU). This component works to more objectively identify a landscape’s visual condition, capability to absorb further  75 disturbance, biophysical properties, viewing condition and numerous other aspects important to visual quality (see BC Ministry of Forests, 1997b, Appendix 2).  These areas are often delineated on a paper-based map at a larger scale of roughly 1:50,000. The process of delineation combines both objective and subjective criteria. The Visual Landscape Inventory Manual (BC Ministry of Forests, 1997b) provides fifteen procedural steps for the proper delineation of these units. Of these there are a few that guide the spatial delineation of the boundaries of these units. These can be summarized into the following: • Generate a visibility analysis using viewpoints and travel corridors. • Delineate VSU boundaries based on landform attributes. These units may include panorama or distinct and separate areas; more specifically: o Record the upper boundary of the VSU, typically identified using the top of the contours along a map. If working in a flat area, this may be the boundary of an area furthest away from an observer. o Identify the lower bounds (or flat topography) closest to an observer. Contours may again serve as a guide. Terraces or hollows or other non-visible areas could be identified and mapped separately so they are not included in the VSU. • Flat of low-level topography which are screened behind vegetation on either side of a road, highway, or along water shorelines should be identified separately with an approximate width of 200 m (depending upon vegetation cover). These become their separate VSUs. Many of the other steps involve procedural criteria such as how to take photographs or capture the scene using a camera or video recorder, and as well, to label important landscape features that may draw the attention of observers. Aside from the additional procedural steps, the summarized bullets demonstrate how the Ministry attempts to objectively delineate VSUs. In the past, these procedures listed above were produced using paper maps and GIS. With modern GIS technology, each of these items can be more easily delineated. A viewshed could be produced using road segments. Upper and lower bounds could be identified through ridgeline and valley identification methods  76 often used in hydrographical modeling techniques and the identification of low-level topography can be identified by a combination of slope maps and viewshed analysis. Although these methods will not be discussed, they do represent ways with which GIS can aid in the delineation of VSU by providing precise and objective information. However, even with the objective methods for delineating these areas, GIS is still reliant on the identification of viewpoints, which build the foundation for the analysis. This is true of the first and last bullet points above. A viewshed analysis, necessary for both identifying areas visible and screened on the landscape is contingent upon the location of an observer. In a geographical area where the views are predominately experienced from a car or boat, it may not be appropriate to represent these experiences by using a single viewpoint. So, even if the delineation of VSU is based on spatial analysis, there still exists subjectivity in the analysis due to the influence of the selected location of the viewpoints. Furthermore, the mere identification of visible areas using a binary form of viewshed analysis may not best represent how an individual would actually ‘see’ the landscape. 4.1.1 Using Visual Magnitude for Visual Landscape Inventory One of the significant contributions of the HCV was its efficiency at providing information about a landform in a 2D surface for large areas represented by numerous viewpoints. Unlike the typical assessments in which only a few viewpoints are used for conducting a binary visibility analysis, this algorithm facilitates the exploration of Visual Magnitude to enrich the understanding of a user experience and to model a more continuous approach for landscape planning. The HCV may offer a more precise and objective approach to the VLI process, through a better representation of how the landform would be experienced by individuals passing through the landscape. Three examples are presented in this section in order to demonstrate how the HCV could improve the VLI process. The first shows an example of a static versus route-based analysis using a small portion of a route from within the Inside Passage of BC. The second is a demonstration of how using a route-based analysis can reveal additional visually sensitive areas. The third is a large-scale example showing Visual Magnitude values at the provincial- level. The first two examples are meant to demonstrate the differences of a route and single  77 viewpoint analysis, and to provide context to the provincial-level analysis so that the significance of the contribution of the HCV for large-scale analysis is demonstrated. The first example is taken from a hypothetical ferry route in the Inside Passage. The distance of the path to the land varies between 500 m and 800 m away from the shoreline of Vancouver Island. Visual Magnitude is calculated for the single viewpoint and for the entire path. The latter is represented by a composite output of 182 unique points. The resolution of the data is built on a 10 m elevation model based on BC TRIM data (BC Integrated Land Management Bureau, 2011). The viewpoints are located at 20 m above the water surface to represent the view from a deck of a cruise liner. The 2D figures show Visual Magnitude as an orange gradient; the darker the orange, the higher the Visual Magnitude. Figure 4.1 depicts the Visual Magnitude analysis using a single viewpoint, shown as an orange point in the ocean. There are several high Visual Magnitude values. The most obvious exists in three distinct clusters of high values along the banks of the shore and slightly inland. These values suggest that the banks of the shore are relatively steep. The largest clustering of high values is grouped in the terrain nearest to the viewpoint. This is partially due to distance, but most likely because the terrain maintains a steep incline that faces the viewer as it extends into the mainland. The second clustering shows two narrow paths of dark orange. These two paths are also steep terrain with surfaces that face the viewer. In between these areas is possibly some kind of topological depression where the cells are either flat or turning away from the viewer. The striping may be a result of a water outlet. The third major clustering along the shores represents a peak. The remaining high or middle values exist in the distance, and are predominantly sharp peaks or faces where the aspect is directly facing the viewer.  78  Figure 4.1 Visual Magnitude as calculated from a single viewpoint. All visible areas are shown in shade of orange, what remains is either ocean or hillshade. In Figure 4.2 Visual Magnitude was analyzed for 182 unique viewpoints along a route, intersecting the single viewpoint used for the analysis shown in Figure 4.1. From an operational perspective this gives planners a far better picture of those areas on the landscape that will be less noticeable throughout the route. The output is a composite for all 182 viewpoints, all assigned equal weights. Similar clusters are found in both figures and there is a small increase in that amount of visible areas. However, the distribution of Visual Magnitude throughout these ‘seen’ areas is quite different as shown by the prevalence of mid-level Visual Magnitude values in the latter figure. Since the composite value is taking a weighted average of all Visual Magnitude values for each of the 182 viewpoints along the route, this results in a flattening of the spatial distribution of low- and high-level Visual Magnitude cells.  79  Figure 4.2 A composite Visual Magnitude represented by 182 viewpoints along a route (purple) with the single viewpoint used in Figure 4.1 represented by a transparent white point. All visible areas are shown in shade of orange, what remains is either ocean or hillshade. The comparison between Figure 4.1 and Figure 4.2 demonstrates how a weighted average Visual Magnitude analysis from multiple viewpoints can provide more accurate information about the landform versus a single viewpoint. In this next example, this concept is carried further to demonstrate how using multiple viewpoints can provide a more accurate delineation of visually sensitive areas. The example, shown in Figure 4.3, depicts a portion of the landscape near Pemberton, BC.  80 In Figure 4.3, three different images are provided: A shows a planimetric view of the area, along with a single VSU; B shows the same area and single VSU from a single location along Highway 99; C shows the outcome based on an analysis from the HCV with all VSUs in the area based on the highway road network of BC (BC Integrated Land Management Bureau, 2011).  The view shown in B is approximately located by the black dot in C. Note that in B, not all of the VSO is visible, nor can the boundaries of the VSU be seen from this view. The bold white line depicts the boundary of the VSUs like those shown in the other two images.  81  Figure 4.3 Example showing visible areas produced from the HCV not included as a visually sensitive area in the Visual Landscape Inventory. Image in A (obtained from Google Maps, © Google 2011) shows a planimetric view of a single VSU, which is shown in perspective view in B (obtained from Google Earth, © Google 2011). In C, that same VSU is shown with other VSUs for context. In C, you can see the visible area to the left of the VSU shown in all three images, where an alteration also exists.  82 A few observations can be made about these images. First, in A, one can see a clearcut harvest located toward the west end of the VSU. That harvest continues along the north and west of the mountain, extending beyond the VSU. In B, that same harvest is visible to the right of the image in the upper portion of the mountain side. Second, C shows that there is no VSU that exists where the harvest extends beyond the VSU located in A and B. Third, that the area where the harvest extends is visible from along the highway, as shown in C by the light orange colored cells. These examples suggests that when the VSUs for this areas were delineated, that a portion of the landscape which is visible from along the highway was likely not correctly identified in the delineation of visually sensitive areas. For this example, a binary viewshed could have been used to help identify additional areas that should be considered within a particular VSU. However, the addition of Visual Magnitude might also be helpful for delineating VSUs because they also include the perceptual effects of slope and orientation, which may provide a more natural separation for these units. It is unknown exactly how the visually sensitive areas in this region were delineated. However, based on the earlier summarized bullet points about the spatial delineation of VSUs, it can be inferred that the bounds were drawn by using contour lines on a paper map, a viewshed from a few selected viewpoints, or a combination of the two. In Figure 4.3 B, the VSU does seem to follow the ridgeline at the top of the mountain, but from the viewpoint shown, the VSU bounds to the right of the image which extends down along the mountain misses portions of the visible landscape where the harvest exists in the background. The assumptions about how the VSUs were defined suggest that if a few viewpoints were used to provide a visibility analysis for the area that those few viewpoints failed to capture the total possible visible area from along the highway. Figure 4.3 C shows a more accurate assessment of the visibility of the landscape as seen from the highway. The analysis in this figure used a 125 m resolution with viewpoints selected at every 125 m along the highway. For the purpose of identifying visually sensitive areas, the particular Visual Magnitude values are irrelevant. Rather, the mere demonstration of places visible (any shade or orange) from along the highway, represents the most valuable outcome  83 of this analysis. Had the original VLI process included the multiple views from along the highway, it is likely that the large area extending west from the VSU of interest would have been included as a visually sensitive area. This would have opened up the possibility of identifying a VSU particular for that area. Furthermore, the analysis shown in C also demonstrates an increased precision as to the delineation of the VSUs. In C, there exists a VSU to the south of the one shown in A and B. Notice how the visible area actually extends beyond the bounds of that VSU in all three directions shown in the map. This example highlights how the HCV (or even other viewshed analyses) can be used to more accurately depict visible areas from along a route. This depiction is one of the first steps in identifying visually sensitive areas. As Figure 4.3 shows, there are areas around the province which may be visible from along the highway, but not included as a visually sensitive area. If the HCV were to be used to help derive visually sensitive areas, it is likely that additional land in BC would fall under the umbrella of visual protection. If this were so, the harvested area that extends beyond the VSU shown in Figure 4.3 A and B would have needed to meet government regulated visual requirements. 4.1.2 Provincial Level Analysis and Regional Comparison In this next application, the HCV was put through an extensive test to reveal an extremely large-scale result to demonstrate the potential of using the HCV over large areas with a number of viewpoints. A single analysis was conducted for all highways and freeways in British Columbia to demonstrate the potential of the algorithm to allow for a large-scale analysis. This demonstration does not preclude the importance of lakes, waterways, trails and other routes important to residents and visitors to BC, and in the future, with appropriate datasets, these kinds of analyses can be completed. The analysis was conducted using only the terrain information, without any vegetation heights considered. A surface model was created from TRIM data collected from the Land and Resource Data Warehouse (BC Integrated Land Management Bureau, 2011). The roads used in the analysis are all highways and freeways selected from the provincial road network (BC Integrated Land Management Bureau, 2011). The analysis was conducted using 125 m resolution which resulted in an elevation model of 12,307 x 11,054 cells and 113,402 unique viewpoints. The viewpoints  84 represent sample points located nearly every 125 m along all freeways and highways (both major and minor). The total time for analysis took just under 44 hours to complete using two i7 Intel processing cores, each operating at 3.2Ghz with the maximum amount of available RAM (~22GB), running a native 64-bit process. The result of this analysis is shown in Figure 4.4. The top pane of the figure provides a coarse overview of values along the selected road networks (the darker the orange the higher the VM). At this visual scale it appears that there are large expanses of land that are not seen from any of the highways or freeways. The two inset maps at the bottom of Figure 4.4  show areas in Southwestern Mainland BC. The leftmost inset shows the Sea-to-Sky Highway, Lillooet Highway and part of Highway 1 through Hope, BC. The rightmost figure shows Vancouver and West Vancouver. The bottom two inset maps make it easier to see the how of the landform might be experienced from along these highways. When driving through areas where there is a high amount of Visual Magnitude along the highways, one would likely experience numerous steep mountains that were very close to the highway. It should be noted that Visual Magnitude is not necessarily a descriptor for feature prominence, but from these maps is seems that prominent features, such as steep slopes facing the highway, are depicted by dark orange. There also seem to be a number of these features along the Lillooet Highway. Even at a 125 m resolution, the steep northern edge of Stanley Park is noticeable.  85  Figure 4.4 Visual Magnitude analysis for all freeways and highways in British Columbia. The darker the orange, the higher the VM. The highways (not shown) are within darkest orange. At this scale, it may be difficult to interpret what the top pane in Figure 4.4 is showing. Therefore an additional analysis was conducted to compare the relative Visual Magnitude values for all freeways and highways throughout the provincial regional districts. This analysis compares the amount of Visual Magnitude per region, per viewpoint and is shown in Figure 4.5 and Table 4.2.  86  Figure 4.5 Regional comparison of Visual Magnitude from all highways and freeways in BC. Legend values represent levels of the amount of Visual Magnitude per viewpoint per area visible from the highway (black line). Regional names are available in Table 4.2.     87 Table 4.2 Comparison of the regional amount of Visual Magnitude per viewpoint per amount of visible landscape Regional District # of Views Absolute VM^ VM Factor* 1 Northern Rockies 6826 1.98 3 2 Alberni-Clayoquot 1183 1.64 2 3 Bulkley-Nechako 3847 5.37 7 4 Capital 1084 6.76 9 5 Cariboo 7782 0.73 1 6 Central Coast 946 17.78 24 7 Central Kootenay 7851 2.04 3 8 Central Okanagan 1033 11.60 16 9 Columbia Shuswap 7376 4.72 6 10 Comox Valley 1236 2.88 4 11 Cowichan Valley 942 2.02 3 12 East Kootenay 5858 8.97 12 13 Fraser-Fort George 6527 12.06 17 14 Fraser Valley 4113 13.07 18 15 Greater Vancouver 2433 10.76 15 16 Kitimat-Stikine 8913 8.64 12 17 Kootenay Boundary 3225 1.18 2 18 Mount Waddington 1853 9.11 13 19 Nanaimo 1707 2.13 3 20 North Okanagan 1816 8.99 12 21 Okanagan-Similkameen 3470 6.81 9 22 Peace River 9073 1.45 2 23 Powell River 212 8.83 12 24 Skeena-Queen Charlotte 1975 11.05 15 25 Squamish-Lillooet 2534 12.42 17 26 Stikine 4422 6.69 9 27 Strathcona 1780 5.88 8 28 Sunshine Coast 254 40.93 56 29 Thompson-Nicola 12659 9.89 14 ^Absolute Visual Magnitude values calculated per viewpoint per the area visible from all highways and freeways in the province, in millionths *The factor is based on the inverse of Absolute VA, using the lowest amount of Visual Magnitude per viewpoint per area visible within the Cariboo region as the base.  This analysis provides a little more practical information about what Visual Magnitude describes at the provincial scale. Although Visual Magnitude is not a measure of feature prominence, like tall mountain peaks, when compared in this way, it seems that the best way to describe the outcome is landform prominence. A couple examples are provided. Based on  88 the VM Factor in Table 4.2, the Cariboo region has the lowest measure of Visual Magnitude per viewpoint per visible area. A depiction of this region can be found in Figure 4.6.  Figure 4.6 Example scene of Highway 97 through the Cariboo Region. Image obtained using Google Earth, © Google 2011 Notice how the landform is relatively open near the highway. An open landscape with relatively few hills, except in the distance, does constitute a low level of Visual Magnitude.  89 In contrast, Figure 4.7 from the Sunshine Coast shows the highway cutting through hills very near the road. In this case the Visual Magnitude values would be quite high.  Figure 4.7 Example scene of the Sunshine Coast Highway in the Sunshine Coast Region. Image obtained using Google Earth, © Google 2011 Until recently, viewshed analyses at these scales were virtually impossible. With the algorithms like the HCV and the availability of high-powered computing, these kinds of analyses can allow for comparisons across wide tracks of land. Even though a single  90 provincial analysis like this will not be used for operational planning, it does present an opportunity to demonstrate the potential for using numerous viewpoints to compare areas in the province. In this way it relatively safe to say that a highway drive along the Sunshine Coast, Central Coast or Fraser Valley is likely to have vast expanses of highway where the proximity of the road is very near hills and mountains. This method could be used in the VLI process for easily identifying all visible locations, and simultaneously providing biophysical measures (slope and aspect), as well as observer context (distance) for an entire landscape or highway trip. This would bring an additional level of objectivity to the delineation of VSU and might help to more precisely distinguish where a unit should begin and end. If an analysis like this were to aid in the VLI process, additional consideration should include various other perspectives, such as views from ferry routes, hiking trails, lakes, etc. Potentially these can provide analysts with new information that may not have been considered in the past. The important consideration here is to recognize that calculating Visual Magnitude along a route or multiple routes, and not just a small portion of a landscape, can ensure that planners are given useful information about how a landscape may be experienced from a variety of views. It can also be used to better represent the continuous experience of seeing the landform from the highway. 4.2 The Management and Assessment of Harvests in Visually Sensitive Areas After Visual Sensitivity Units (VSU) have been defined, the remaining landscape management processes focus primarily on dealing with the management and assessment of visual quality at the unit scale. These processes are depicted as Phases 2 – 6 in Table 4.1. In the second Phase, the Ministry designates each VSU with a recommended Visual Quality Class (VQC). These classes help to distinguish the recommended amount of visible alteration and generally, specify particular design elements that should be considered when designing a harvest plan. These classes are provided in Table 4.3.  91 Table 4.3 Visual Quality Class definitions VQ C Basic Class Definition2 % Alter. Ranges Pr es er va tio n Alteration is • very small in scale, and • designed to be indistinguishable from the pre-harvest landscape; 0 Re te nt io n Alteration is • is difficult to see, • is small in scale, and • has a design that mimics natural occurrences; 0-1.5 Pa rt ia l Re te nt io n Alteration is • is easy to see, • is small to moderate in scale, and • has a design that appears natural and is not angular or geometric; 1.6-7.0 M od ifi ca tio n Alteration is very easy to see and is either • large in scale with a design that is natural in its appearance, or • small to moderate in scale but with a design that has some angular characteristics; 7.1-18.0 M ax im um  M od . Alteration is extremely easy to see and one or both of the following apply: • the alteration is very large in scale; • the alteration is angular and geometric. 18.1-30.0  Phase 3 deals more with the regulatory controls of the management system. The difference between a recommended VQC developed in Phase 2 and the VQOs approved in Phase 3 is that the latter is a legally binding objective, set by the Ministry of Forests and mandatory for companies(BC Ministry of Forests, 2004). Since the creation of the Visual Landscape Inventory Procedures and Standard Manual (BC Ministry of Forests, 1997b), where VQOs were originally designated for many VSUs, new regulations have come into play that represent a slight procedural shift in how VQOs are established. The new regulations under  2 Alterations are described as a modification to the landscape resulting from the presence of cutblocks or roads, such that when assessed from a viewpoint that is representative of significant public viewing opportunities  92 the Forest and Range Practices Act (BC Ministry of Forests, 2004) specify that VQOs from the old system be grandparented into current regulations, but that any VSUs that do not have VQOs assigned now fall under the jurisdiction of regional or district anagers (BC Ministry of Forests, 2004; Marc, 2008). Operationally, VQOs and VQCs are quite similar in their conception. VQCs are broad recommendations which suggest the amount of alteration and suggested shape design of harvest blocks (natural vs. angular) that would be appropriate for a harvest within this landscape (BC Ministry of Forests, 1997b, 1998). VQOs identify the same criteria, as well as other potential operational restrictions, but are primarily different because they are legally binding as a Government Actions Regulation, whereby a Regional District Manager has been given the jurisdiction to enforce VQOs in their regions (BC Ministry of Forests, 2004). So it is the VQO that companies must manage for, or face possible financial penalties. In some forested areas, these VQOs may be perceived as a limitation to harvesting. In some cases visually sensitive areas are avoided by licensees altogether (Picard and Sheppard, 2002a). This is likely due to the high costs of meeting the VQOs or possibly the potential for these front country areas to elicit a negative public response. The Ministry of Forests is aware of these issues and acknowledges that good design is important for future logging activities in these areas (BC Ministry of Forests, 1994d), and provides materials and support to help companies in the planning process (Marc, 2008). 4.2.1 Effectiveness Evaluation One of these support documents is the Protocol for Visual Quality Effectiveness Evaluation (Marc, 2008). This document covers concepts from previous phases identified in Table 4.1, but is more heavily aimed at helping planners with designing harvests (Phase 4) and in assessing the harvests (Phase 5). From the design standpoint, the evaluation clarifies the kinds of criteria that will be assessed. In effect, the document becomes a “cheat sheet” for planners who can anticipate how harvests will be assessed, and thus, plan accordingly. The Effectiveness Evaluation is based on principles of  VRM and empirical research in forest aesthetics (Marc, 2008). In Marc (2008) two overarching design principles greatly influence  93 individual perceptions of forest operations in visually sensitive areas: 1) large alterations with visible roads or sidecast are less preferable, and 2) significant tree retention and alterations which are designed to fit the landscape are greatly preferred. The Effectiveness Evaluation focuses on these effects at distances of 1-8km from observer locations, as these have shown to be the more important distances for aesthetics in forestry (Hull and Buhyoff, 1983; Litton, 1979; McCool et al., 1986; Pâquet, 1993). The procedures are based on measuring conditions for silviculture practices such as: clearcut, aggregate retention (patch-retention) and partial cut alterations, which constitute the majority of alteration types in BC’s visually sensitive areas. Dispersed retention is also discussed in the evaluation and if used, can often increase the visual quality of a harvest. The actual evaluation of harvesting activities consists of a comparison between the VQO of a VSU set by the Ministry and the visual quality of the VSU, post-harvest (Marc, 2008). The VQO (which is based on VQCs shown in Table 4.3) is established before harvesting activities, whereas the evaluation of the harvest is completed after operations are finished. Upon the assessment of a post-harvest operation, the VQO of the pre- and post-harvest scenarios are compared. Differences between the two represent the degree of a plan’s effectiveness at meeting the VQO (Marc, 2008). It is crucial for companies operating in visually sensitive areas to follow these procedures, as failure to meet the established VQOs can lead to financial implications, and also holds the potential to impact their reputation for forest stewardship (Marc, 2011). Arguably, one of the most significant elements of a VQO is the amount of visible alteration that is allowed. These percentages are largely based on the recommended VQCs found in Table 4.3. Evidence has shown that the percent of alteration is one of the most influential characteristic of landscape modification, as it pertains to preferences (e.g. BC Ministry of Forests, 2001b; Shang and Bishop, 2000). Since the amount of visible alteration is such an important design element, several methods have been proposed for using geospatial technologies to automate the calculation of these impacts (Domingo-Santos et al., 2011; Fairhurst, 2010; Grêt-Regamey et al., 2007; Iverson, 1985; Shang and Bishop, 2000; Travis et al., 1975). In a continuation of these prior efforts, the next section delves into the  94 feasibility of using the HCV for automating the calculation of the percent visible alteration of a harvest and used 3D photo-realistic simulations to validate the algorithms effectiveness. 4.2.2 Automating the Calculation of the Perspective Amount of Alteration In BC, each Visually Sensitive Unit (VSU) has an established Visual Quality Objective (VQO), within which an allotted amount of the landscape can be visibly altered. As part of the Effectiveness Evaluation (Marc, 2008), the amount of alteration is used to determine the visual quality effects resulting from a harvest. Since these requirements carry with them possible penalties or fines, planners have found ways to use technology to simulate their proposed plans in order to assess their potential visual ramifications. Often, the technology employed is a suite of 3D software called Visual Nature Studio® (VNS), which can generate photo-realistic models of their plans before they are approved for harvesting. However, the process of developing a plan and modeling it in 3D requires significant efforts. Typically, planners use GIS to develop 2D maps of their plans. Then, this spatial information is passed to VNS and the landscape is rendered. From this rendering, the planners then use additional software to calculate the number of pixels where modification is visible and divide that by the number of pixels of the entire landscape. This area becomes the amount of visible. Theoretically, Visual Magnitude can provide a way to calculate the perspective visible alteration of a proposed harvest without any rendering in 3D. Again, Visual Magnitude calculates the effects of distance, and the difference of the viewing perspective from the observer to a target surface of the earth. These calculations provide an objective quantitative value for the surface (represented as discrete units) that ranges from invisible (0%) to completely visible and entirely occupying an individual’s view (100%). These numbers provide a way to quickly assess the amount of alteration that has occurred from an individual viewpoint. How the Calculation is Completed using Visual Magnitude Consider a simplified example using the Visual Magnitude values as shown in Figure 4.8, where The cells shown in this figure represent the total area of interest, or the VSU. Assume that all cells are visible. In this figure, let the viewpoint be located to the top right of the grid,  95 and outside of the grid. Shaded cells represent a harvested area. The darkly shaded cells represent the visible cells, while the lighter shaded bordering cells are those that are not visible (screened by the trees in white cells in front). Cells with any shading represent a harvested area. The Visual Magnitude values are calculated using the surface, and visibility is calculated using tree heights. Assume that the total area within the grid is visible; this means that the harvested area represents 52% (13 of 49 cells) of the total area. In order to calculate the perspective visible alteration, the Visual Magnitude values for the visible harvested area would be summed (2.54) and divided by the total Visual Magnitude values for the whole grid (13.14). This reveals a total of 19.39% visible alteration. 0.12 0.08 0.08 0.11 0.09 0.04 0.01 0.2 0.15 0.15 0.16 0.15 0.1 0.01 0.32 0.28 0.24 0.21 0.17 0.12 0.07 0.37 0.34 0.33 0.34 0.32 0.29 0.03 0.31 0.29 0.35 0.42 0.46 0.45 0.44 0.26 0.28 0.34 0.42 0.48 0.45 0.42 0.31 0.32 0.36 0.45 0.51 0.48 0.46 Figure 4.8 Example surface and Visual Magnitude values for calculating perspective visible alteration In order to calculate this value using GIS, several data must exist (refer to Figure 4.9): • Viewpoint location (shown in black) • Harvest (white outline) • Digital Surface Model (underlying terrain shown as hillshade) • Forest cover heights (opaque green at 20 m, brown at 0 m) • Visually Sensitive Unit or viewing angle (white overlay)  96  Figure 4.9 GIS data input information for calculating percent visible alteration The data given in Figure 4.9 is drawn from one of the harvest shapes used for the survey in Chapter 2. The data is used here in order to draw comparisons between the Visual Magnitude calculations and the rendered amount of percent visible alteration based on the shape itself. The process for calculating the perspective amount of visible alteration with these data is completed through several steps. Figure 4.10  depicts an example forested area in 2D to illustrate the process. First, the Visual Magnitude values are calculated for the landform using the viewpoint and terrain. Then, tree heights are added to the DTM to simulate forest cover. The harvest is then simulated by subtracting the same tree heights from within the harvest boundaries. What remains is the landscape with forest cover, except where the harvest exists. Then, a viewshed is conducted to find the visible harvested cells (those cells in the harvest that are visible). Finally, the sum of the Visual Magnitude values of these cells are divided by the sum of the Visual Magnitude values of the entire visible portion of the landscape within the VSU.  Figur 4.2.2 When have simul pixel Figur to tha divid e 4.10  Foreste .2 How t  forest plan a way to cal ating their p s within the e 4.9. Yet, w t shown in F ed by the tot d landscape i he Calculat ners are in t culate the p lan using 3D harvestable hat is used igure 4.12. al black and n 2D showing ion is Comp he process o otential amo  photo-real area. Figure to calculate From this fi  gray cells i  visible harve leted Using f developing unt of visibl istic softwa  4.11 is the the percent gure, all the n the image sted area  3D Simula  proposed h e alteration re and then c 3D simulati visible alter  black cells . tions arvest plan . Typically, alculating t on of the har ation is a ren are counted s, they must this means he number o vest depicte dering simi  and then 97   f d in lar  98  Figure 4.11 Example 3D harvest simulation  Figure 4.12 Example 3D harvest simulation used to calculate percent alteration For this particular image, the total percent of visible alteration is 15.02%. It should be noted that the process of counting the number of pixels may be a bit subjective, that is to say, that no one planner simulates the harvests and counts the number of pixels exactly in the same way. In all the literature available by the Visual Quality Branch within the BC Ministry of  99 Forests, there is no mention of precisely how these values should be calculated. So, the onus is left to the planners to make sure that what they simulate – and eventually harvest – stays within the range specified by the VQO that will later be assessed by a Forest District Manager. The way in which the cells were calculated for the harvest depicted in Figure 4.11 was by replacing trees with blocks of color, and the ground cover with a single color tone. Then, this image is reclassified into two colors to ensure that pixels counts are accurate, as shown in Figure 4.12. However, where subjectivity creeps into the process is both in the rendering stage, where blocks of color replace trees, and in the reclassification process. Blocks of color do not provide the same tree structure as an actual tree and the reclassification may exclude or include those cells along the edge of the harvest which could potentially fall within either the harvest or the forest cover. This is because during the rendering process, the edges of the harvest are influenced by shadows and light reflections. In the case of the percent calculation found here, classification was conducted several times to ensure that the percent visible calculation was accurate. The range of percentage calculation conducted for this image was found to fall within .25%, a small margin of error. Validating Visual Magnitude: a Comparison with 3D Simulations If Visual Magnitude is to be used for calculating percent visible alteration, it must provide an accurate and reliable assessment. In this section, the percent visible alteration was calculated for 52 harvests on a landscape using both the 3D simulations and Visual Magnitude methods. All the data depicted in Figure 4.9 were included in the analysis, but the VSU had to be substituted for the viewing angle. It is difficult to see from the 3D simulation in Figure 4.11, but the extent of the VSU delineated actually extends beyond the left and right edges of the simulation. Therefore, a careful delineation was done to produce the viewing angle. This required knowing the target location (lat/long/elevation), the viewpoint location (lat/long/elevation) and the viewing angle of the camera rendered in the simulation. With these identified, a boundary was hand-drawn on the 2D GIS map shown in the yellow line in Figure 4.13. The analysis was conducted using the original data resolution of 30 m.  100  Figure 4.13 Example viewing angle created for comparison The average percent visible alteration from the 3D simulations resulted in 14.63% with a standard deviation of .76%, while the average using Visual Magnitude was 23.59% with a standard deviation of 2.92%. The average differences of these two outcomes are substantially large. Also, the differences in standard deviation between the two methods suggest that the calculations using Visual Magnitude vary substantially more than the 3D simulations. Perhaps these differences are a result of a systematic error between the methods where the Visual Magnitude values were all measured much higher than the values form the 3D simulations. To test this hypothesis a correlation between the percent visible alteration from the 3D simulation and that derived from using Visual Magnitude was produced, resulting in a coefficient of .186. Initially, this would seem to suggest that the Visual Magnitude method may not be an adequate surrogate for the 3D simulation method; however, there are two possible explanations for this difference. The first is due to the influence of resolution. Resolution can  101 play an important role in the calculation because of edge effects. For example, look at the yellow line in Figure 4.13 and in Figure 4.14 that represents the viewing angle taken from the 3D simulation. The raster equivalent, shown by an opaque white overlay (the original VSU is shown as the large opaque white border in Figure 4.13), demonstrates one possible cause of this difference. If the error was only caused by the difference in the amount of area included in the VSU, then the difference in the visible alteration amounts between these two methods would have been reflected merely as proportional differences. That is, for each harvest rendered in 3D, the proportion of visible alteration using Visual Magnitude would have most likely been systematically higher across all harvests. However, this would have resulted in a higher correlation coefficient between these two methods, so it is most likely that the differences are not due to the conversion of the vector delineation of the VSU to the raster model. If the differences in the amount of visible alteration from the Visual Magnitude values and the 3D simulation are not a result of the conversion of the vector VSU to raster, then most likely reason for the disparity is from conversion of the vector harvest to a raster equivalent. This is likely because the 3D simulation uses a vector input to delineate the harvest, whereas the Visual Magnitude calculation uses a raster. An example of the conversion is shown in Figure 4.14 where the harvest vector is delineated in orange and the raster in opaque blue. As in many cases, the conversion from vector to raster data will result in some error. In the case of the VSU, this error would have been consistent throughout all percent visible calculations. However, with harvests, the amount of error from the conversion of vector to a raster format would differ from one harvest shape to another. This conversion error may account for some of the observed differences between the two approaches.  102  Figure 4.14 Example conversion of harvest from vector to raster Another possible reason for the differences is the way in which the terrain is considered. The values for Visual Magnitude were produced using 30 m resolution to interpolate the surface, whereas the 3D simulation uses an entirely different interpolation process. Visual Nature Studio® (from which the 3D simulations were created) interpolates the terrain data using fractal depth (or multiple subdivisions of a cell). These subdivisions are created by taking the original cell size and halving it for each dimension up to 7 dimensions so the equivalent cell size becomes 27 of the original input (Hauldren, 2008). For instance, what started as a cell size of 30 m became 15 m, then 7.5 m, etc., until the original cell size was interpolated to near .25 m. This process is meant to imitate a more realistic terrain for 3D rendering purposes(Hauldren, 2008). Aside from the resolution effects, the second potential source of error lies in the way VNS simulates the scene using the viewing angle and how that angle is drawn in GIS. In GIS, the angle was drawn based on standard trigonometry. Since the angle and viewing distance to the  103 target were known it was easy to pinpoint the exact latitude and longitude of the extent of the viewing angle. However, it is possible that the actual extent used in VNS varies slightly from the stated viewing angle. More research needs to be conducted in order to verify exactly how VNS controls its viewing angle extent. It was hypothesized that if the resolution was higher, it might be possible to increase the accuracy of the percent visible alteration amounts calculated from Visual Magnitude. Further tests were conducted using two additional resolutions of 10 m and 3 m. However, since the terrain elevation was unavailable at this resolution at the time of analysis, the terrain values had to be generated through the process of resampling using bilinear interpolation. This approach likely differs from that employed in VNS and thus represents an additional source of error. The results of the calculations are shown in Table 4.4. Table 4.4 Results of percent visible alteration comparisons between 3D simulation and Visual Magnitude using 30 m, 10 m and 3 m resolution  VNS Visual Magnitude  3D Simulation 30 m 10 m 3 m Correlation Coefficient .1858 .7924 .8303 Average Alteration (%) 14.63 23.59 17.81 17.53 Standard Deviation (%) 0.76 2.92 1.08 1.10  The use of increased resolution for the analysis of Visual Magnitude led to a substantial increase in the correlation between the two approaches, with correlation coefficient increasing to 0.79 and 0.83 at the 10 m and 3 m resolutions, respectively. While the correlation is greatly improved by increasing the resolution, the Visual Magnitude method still tends to overestimate perspective alteration relative to the VNS approach, but the relative difference is much smaller. More research is needed in order to determine the precise cause of these differences. However, so long as the proposed method produces a reliable correlation with the 3D simulations Visual magnitude can be used to approximate the percent of visible alteration within a VSU. The full set of all 52 values are plotted in Figure 4.15,  104 depicting this trend. The differences in the 30 m, and 10 m and 3 m resolutions are quite clear.  Figure 4.15 Scatter plot of all 52 images comparing values calculated by the VNS 3d simulation and Visual Magnitude at 30 m, 10 m and 3 m resolution. The black line shows a perfect correlation with exactly similar values , whereas the other trendlines show the correlation with their offset values. These outcomes suggest that given an appropriate resolution, it may be possible to calculate the percent of visible alteration directly using Visual Magnitude as the method is further refined. This becomes increasingly important when trying to calculate perspective alteration over large areas and from numerous viewpoints (e.g. simulating a highway drive). This kind R² = 0.0345 R² = 0.6279 R² = 0.6894 10% 14% 18% 22% 26% 30% 10% 14% 18% 22% 26% 30% Ca lc ul at ed  U si ng  V is ua l M ag ni tu de Calculated Using VNS 3D Rendering VNS 3D vs Visual Magnitude 30m 10m 3m  105 of analysis is cost and time prohibitive if using 3D simulations. Yet, since there is such a high correlation coefficient and similar deviations when increased resolution is used, it may be possible to render a random sample of locations from along a highway, calculate the percent visible alteration and then find the average differences between these values and the calculated values from Visual Magnitude. These differences could be used to correct the Visual Magnitude values for all remaining locations along the highway to derive an accurate portrayal of how the harvest may be seen throughout the route. Unlike many of the previous attempts as automating this calculation, the HCV can efficiently calculate the amount of alteration for a large number of viewpoints at extremely fast speeds. This means that the percent visible alteration could be calculated over an entire route, providing information about the effects as seen from a single location, but also averaged throughout an entire route. To validate the algorithm’s method as a viable option for calculating the percentage of visible alteration, further research must be conducted. This means developing many other 3D simulations from different landforms and with different amounts of percent visible alteration. It also means validating the calculations using real-world examples of pictures taken from a variety of landscapes as the 3D rendering method is also subject to inaccuracies when compared to actual outcomes on the ground. 4.3 Considerations and Problems of Using Single Viewpoint Evaluations As has been discussed in the previous sections, the VLI process, including the Effectiveness Evaluation, was created to provide a more objective way to assess the impacts to visual quality (BC Ministry of Forests, 1997b; Marc, 2008). To accomplish this, a number of design criteria area considered (this paragraph summarizes the main points in Marc, 2008 regarding viewpoints and evaluations). These include criteria such as: the distance of the harvest, the fit with the natural character of the landscape, the consideration of lines of force, and, as well, measures of the amount of the harvest that is visible (BC Ministry of Forests, 1994b, 2001a; Marc, 2008). Each of these criteria is evaluated from several viewpoints. Each viewpoint is carefully selected before an assessment is completed. The selection can include a number of locations such as: a recreation site, stretch of highway, rest stop, scenic pull-out, or a tourist- related commercial enterprise (e.g. lodge). For each viewpoint, a number of data are  106 collected, including: coordinates, direction, distance to harvest, a weighting of importance, etc. It is from each of these viewpoints that an evaluation is executed. When all the evaluations have been completed, a composite evaluation is made. A significant portion of the Effectiveness Evaluation is devoted to the process of selecting and rating the importance of a viewpoint’s perspective (Marc, 2008). The Ministry provides a set of viewpoints which may be selected for the assessment (available for download as Visual Landscape Inventory data from BC Integrated Land Management Bureau, 2011), or other viewpoints may be suggested or chosen if they are determined to be better representations than those provided by the Ministry (Marc, 2008). Again, the identification of these locations is often associated with scenic locations, places where there is likely a high flow of people, or locations where the greatest amount of the landscape can be seen. Once a viewpoint is selected, it is assigned a rating of importance on a scale of 1-5, with 1 representing a short glimpse and 5 a long-term sustained view like that from a lodge or commercial tourist location (Marc, 2008). These viewpoints form the foundation for which all evaluation criteria are based. The process of selecting and weighting a viewpoint places a significant amount of value on the degree to which that viewpoint represents the possible views from within a similar location. The reliance on a single viewpoint to capture the range of possible views brings about an interesting observation. As was shown in Figure 4.3, the representation of multiple points from along the highway provides a more accurate way to show all possible visible terrain. Using a few viewpoints to attempt to capture this full visible area would be an insufficient representation of how an individual would experience a landscape or site as they move through it. The VLI process relies on identifying and weighting viewpoints for their relative significance as a location from which assessments will be conducted (Marc, 2008). The process of identifying a viewpoint is largely based on a combination of quantitative and qualitative data. Although the method for selecting viewpoints will not be detailed here, a few key issues arising from the selection process will be identified. Implications for using a single (or few)  107 viewpoints will be discussed, as well as potential shortfalls of viewsheds and possible alternatives. 4.3.1 Concerns with Viewshed Analysis Viewshed algorithms have a long history, and have been used in numerous applications, such as landscape management and assessment (Fisher, 1996; Germino et al., 2001; O Sullivan and Turner, 2001; Palmer, 2004; Smardon et al., 1986), for the placement of wind turbines (Möller, 2006), for telecommunications (Sawada et al., 2006) and urban tourism (Wilson et al., 2008). They are even used in the delineation of Visual Sensitivity Units (as discussed in section 4.1). The development of the HCV revealed an important observation about viewshed analyses that may have implications for the accurate assessment of harvest designs and delineation of visually sensitive areas. The observation confirms previous authors’ findings that the cells most near a viewpoint can influence a viewshed result (Fisher, 1991, 1992). This section demonstrates inconsistencies from different viewshed methods, also showing how a viewshed analysis can be effected by the cells most near a viewpoint; something planners must be aware of when using viewsheds for planning purposes. This analysis and discussion depict what authors have said earlier about the potential errors stemming from the effects of cells most near the viewpoint (Fisher, 1991, 1992). The development of the HCV focused on comparing two viewshed algorithms, XDraw and the Direct method. As the development progressed, and the two methods were recreated in a custom software solution, they needed to be validated against a reliable pre-existing viewshed analysis. In the commercial world, ESRI’s ArcGIS® represents the lion’s share of GIS software used around the world. It is also one of the oldest and most-tested systems on the market. So, the viewshed analysis in ArcGIS® served as a reliable analysis to compare the Direct method in the custom software. In the remainder of this section, two versions of the Direct method and one versions of XDraw method are compared to the viewshed output from ArcGIS® 10. The comparisons depict the visibility of an individual sitting on a bluff looking out over a surrounding valley and toward steep mountain slopes surrounding the value. Only the surface of the earth is considered, the curvature of the earth and light refractivity are the same for all analyses. The  108 precise location is irrelevant as this is only meant to demonstrate the discrepancies between the different viewshed methods. The terrain for this analysis is shown in Figure 4.16. In this figure A shows the surrounding terrain and B depicts the elevation model directly near the viewpoint. In B, the grayscale representation of the elevation values has been stretched such that black represents at most 720 m elevation and which 750 m or above.  Figure 4.16 Example terrain used  for analysis of the influence of bordering cells on different viewshed results. The left shows a hillshade of the terrain. The right shows DEM values for a close-up of the hill from which the viewpoint is on top. Three comparisons between the standard viewshed available in ArcGIS® 10 were conducted using this terrain and viewpoint at a cell size of 30 m. The three comparisons are as follows: 1. Direct Method starting with all bordering cells 2. Direct Method starting at bordering cells, excluding the orthogonal 3. XDraw method The first comparison is a recreation of the Direct method and is found in Figure 4.17. In the figure, the black cells represent the standard viewshed only, the dark grey represent cells visible in both methods, and the light gray the alternate viewshed. Notice that in A, some  109 differences occur along the inner perimeter of the valley. Although not substantially large, the difference between the two methods is 4.55% within this view. In B, some additional differences are noted. There are several locations where the result differs (shown by either black or light gray), but overall the results are fairly similar (dark gray).  Figure 4.17 Comparison of (1) custom-coded Direct method with industry standard. Dark grey are areas of agreement. Black are areas only visible using ArcGIS®, and light gray is the area only visible from the comparison.  110 The second comparison is also a recreation of the Direct method, but differs from the first method in that does not consider the cells directly bordering the viewpoint in the orthoganal directions. The comparison is found in Figure 4.18. In the figure, the black cells represent the standard viewshed only, the dark grey represent cells visible in both methods, and the light gray the alternate viewshed. Notice that in A, some differences occur along the inner perimeter of the valley. The difference between these two methods is 4.50%, slightly better than in the previous comparison. In B, the differences are more numerous than with the first comparison, with the recreated method identifying more cells visible than the first recreation (as shown in light gray).  Figure 4.18 Comparison of (2) custom-coded Direct method with industry standard. Dark grey are areas of agreement. Black are areas only visible using ArcGIS®, and light gray is the area only visible from the comparison. The last comparison shows the differences between the XDraw and standard veiwshed and is depicted in Figure 4.19. In the figure, the black cells represent the standard viewshed only, the dark grey represent cells visible in both methods, and the light gray the alternate viewshed. Notice that in A, there are still some differences occuring along the inner perimeter of the valley, but fewer than in any of the previous two methods. Upon closer  111 inspection, notice that there is a large patch of light gray roughly at 15 degrees and near the edge of the map where the result from the XDraw method differs from the standard analysis. Overall this method reveals a difference of 6.14%, the largest of any of the methods.  In B, there seem to be a similar number of differences to the first comparison, but they are spatially different. Also note that the cells directly bordering the viewpoint are not identified as being visible in XDraw. This is due to constraints within the algorithm, but for all intents and purposes, those cells are assumed to be visible. From this depiction alone, it would be difficult to say if this method is more accurate than the Direct method, but given the overall comparison in A, seems like a better fit with the industry standard than the Direct method.  Figure 4.19 Comparison of (3) XDraw method with industry standard. Dark grey are areas of agreement. Black are areas only visible using ArcGIS®, and light gray is the area only visible from the comparison. Two observations from this analysis can be noted. First, that neither of the algorithms produces a result exactly the same as the base analysis. It was anticipated that ArcGIS® uses the Direct method, but neither of the recreated Direct method viewsheds demonstrate that effect. Without knowing the precise algorithm that ArcGIS®  uses, it is time prohibitive to attempt to recreate the exact analysis, and irrelevant in order to demonstrate how easy it is to have different results, even if the analyses are theoretically similar.  Second, that if ArcGIS®  112 uses the Direct method, it likely using a slightly different algorithm. It would be interesting to find exactly how this viewshed is created. A follow-up hypothesis is that ArcGIS® is using the Direct method, but likely not including all 8 cells directly bordering the viewpoint into the analysis. If this is the case, then the potential impacts regarding the use or disuse of the first cell must be discussed. Without validation information comparing the algorithms to actual landscape analyses, no conclusive statement can be made regarding the accuracy and precision of any one algorithm. Yet, what is certain is that the cells most near the viewpoint can dramatically influence a viewshed result (Fisher, 1991, 1992). Consider a cell resolution of 30 m. If the viewshed analysis is not including the cells most near the viewpoint, then it is removing the potential implications of roughly 45 m around each orthogonal direction and over 63 m at the diagonals. These effects are not only problematic at 30 m resolution. Even with higher resolutions, this can still have an effect. Take the example in Figure 4.20, where the viewer is standing on a landscape and looking downward. The first of the illustrations show that if the bordering cell is included in the viewshed, then the cell 6 m away is not visible. However, as the second illustration shows, if the first cell is not included in the analysis, then the cell 6 m away is considered visible. Which method is more correct? When the resolution is set to 1 m, it makes no difference whether or not the neighboring cell is included in this case as either option would still provides the same result with the cell 6 m away not being visible. Of course, at this resolution the height of the viewer also influences the result, but it is not hard to imagine that if the viewer was located on a mountain peak looking down along the mountain side, there the result would be quite different if the terrain data was based on 1 m resolution, 30 m resolution or 100 m resolution.  Figur These the ac exam But, f needs inclu One p be to the si of on terrai migh large resolu base e 4.20 Effects  kinds of re curacy and ple suggests or cases wh  to be condu ding or remo otential sol include an a mplest case e or more ce n data. Obvi t reveal how area. Anoth tion. If the data is 100 m of bordering solution and error of a gi  that as reso ere the unde cted to con ving the fir ution to dea nalysis usin , the distribu lls away fro ously, this d  some small er solution w problem is d , it may be cells on high r  methodolo ven viewshe lution incre rlying data sider how al st cell effect ling with the g a spatial d tion could b m the viewp oes not rem  variations i ould be to iscovering w  appropriate esolution view gical effects d calculatio ases, the acc is of low res gorithms mi s.  interaction istribution a e modeled u oint locatio ove the inhe n viewpoint model visib hat the vie  to remove t  shed analysi  can have si n (Fisher, 1 uracy of the olution, it is ght deal wit  of resolutio s mentioned sing a Gaus n, dependin rent error in  location co ility based o wer can see he first cell. s gnificant im 991, 1992).  analysis m  clear that m h the potent n and first-c  in the prev sian curve w g upon the r  the method uld affect an n the proble  from a mou  Yet, if the r plications fo This small ay also incre ore researc ial impacts ell effects m ious section ith a distan esolution of ology, but i  analysis fo m and ntain and th esolution is 113 r ase. h of ay . In ce  the t r a e  3 m  114 it may not be appropriate to remove the first cell. The key is to understand how the data may influence the analysis and to determine if these influences may be problematic. As viewsheds, visibility and similar analyses continue to be used in landscape planning, greater precision and accuracy becomes crucial to ensuring that the outputs of these algorithms are relevant to the problem being addressed.  This may mean that subtle changes in these algorithms become necessary or that at the very least a greater transparency in the assumptions of a particular method is warranted.  It may also be likely that with the accessibility of higher resolution data, some of these problems may become less of an issue, but they should not be written off without further investigation and validation. 4.3.2 Significance and Subsequent Weighting of a Viewpoint Another concern with the representation of single viewpoints in the landscape inventory process is the degree to which they influence an assessment or landscape plan. Viewpoints are often associated with publicly designated scenic stops along highways, at places where the public gather (parks, lodges, etc.), or where large numbers of tourists gather (boat dock, scenic stop on trails, etc.) (BC Ministry of Forests, 1997b). However, as argued in earlier chapters, these points may not be the most accurate proxy for an experience. Yet, whether dealing with a single, few, or numerous viewpoints, the significance of each must be identified in order to appropriately measure the visual quality of landscape modification. The BC Ministry of Forests uses a five-point importance scale to weight the significance of viewpoints (Marc, 2008). This five-point scale is as follows: 1. glimpse view, less than 10 seconds; 2. sustained side view; 3. sustained focal view or traveling toward the alteration for more than 1 minute; 4. viewpoint at a rest stop, campsite, or other static short-term view location; 5. viewpoint at the location of a community, commercial tourist-related enterprise, or other static long-term view site. While the importance scale may be crucial to the current assessment procedures, the ability to simultaneously quantify the percent alteration from multiple viewpoints opens new  115 possibilities. The current importance scale calls into question the accuracy with which the value assigned to a viewpoint is representative of the importance of that viewpoint for measuring visual quality. Using GIS, it may be possible to use more quantitative approaches to weighting a viewpoint. For instance, both frequency of views and time spent viewing could be included in the weighting process. Traffic counts, for example, may be used to give a measureable number of visits. The speed limits associated with a road or highway could be used to calculate the duration of time an individual may spend along a portion of the road represented by a viewpoint. Coupled with Visual Magnitude, this could reduce the subjectivity of selecting a static viewpoint to represent the experience along a highway corridor. Trail-use information could also be applied in similar ways. Again, the frequency of visits to a hiking trail or scenic stop along the trail could be used. The duration of time spent along a trail could be calculated by using hiking guidebooks or similar resources that estimate the distance and time of travel along a route. The total time could then be divided into increments along the route and used as weights for a Visual Magnitude Analysis. This same logic could be expanded to include, campsite visits, or tourist visits to community centers. It would also be possible to include into the weighting function, a sensitivity analysis which further proportions weights along a route by the percent of time or likelihood an individual would be viewing the landscape. This function would suggest that scenic viewpoints be given a higher weighting than that of places along a difficult hiking trail where hikers are often focused on the ground under their feet more than the views along the route. This kind of data can be easily modeled and coupled with the Visual Magnitude algorithm to provide a quick overview of likely places where landscape modification may result in the highest frequency of exposure. 4.3.3 Representing a Viewpoint as a Spatial Distribution Another assumption that has not garnered much attention in the literature is the degree to which a viewpoint reflects the combined viewing experiences of all individuals within an area of interest. For example, assume that many people visit a particular lake for recreational purposes. It is almost certain that where people choose to setup their chairs, blankets, etc.  116 may cause them to see the landscape differently than other visitors. If that is the case then it calls into question whether or not it is appropriate to identify a single precise location to represent the whole range of possible locations from which individuals see the landscape. Again, the limitation of selecting a static location becomes apparent. Perhaps, a viewpoint does not need to be a single location, but could be represented by a number of possible viewpoints within a general location, such as a lake shore. In this case, it may be more appropriate to model the likelihood of an individual seeing the landscape from a location. This idea have been presented elsewhere in the application of viewsheds for quantifying visibility using different observer heights from within the same and different locations (Sander and Manson, 2007). As with Sander and Manson (2007), instead of using a single point for visibility, a spatial distribution of the likelihood of possible locations could be used. This would be similar approach to Fisher (1991, 1992, 1995), whereby he modified the DEM using a statistical distribution to alter the elevations of cells to determine error. However, in the approach suggested here, rather than altering the DEM, shift in location (x, y, and z) and weighting of a viewpoint would be based on a spatial distribution. Continuing with the shore example, the most likely locations would be those near paths from parking lots to the shore, and possibly near park facilities such as restrooms or garbage cans. Likeliness of location experience could then be some function of distance from these areas. An assessment of this kind will ensure that locations within an area, represented by a single location, can be included as part of the analysis, but will be weighted less important relative to the higher traffic areas.  It is also possible that where observational data is available it would not be necessary to predict popular locations but directly input them instead; weighted by visitor counts for example. Using a spatial distribution to weight viewpoints, which ultimately reflect a single location still carries the problem of needing to visit multiple viewpoints for assessment. Rather than taking pictures from a single location or rendering a 3D image from a single location, each view within the spatial distribution would need to be considered and weighted. Using these processes becomes time prohibitive. Yet, with the Visual Magnitude algorithm, these kinds of weights can easily be applied to hundreds or thousands of individual viewpoints within an  117 area, as is done by the HCV. The result would enable a designer to use the composite value of all viewpoints and identify a single best location from the distribution to use as the viewpoint for a more detailed analysis, such as a 3D render or site visit. 4.4 Conclusions The BC Ministry of Forests has made it clear that landscape aesthetics play an important role in forestry. Maybe this is because forest recreation activities contribute about 30% of the GDP derived from the forest sector in 2009 (BC Ministry of Forests, 2010), or because aesthetics are important for intrinsic reasons; in either case the Ministry has gone to lengths to ensure its protection. BC has had a long tradition of managing visual resources (see BC Ministry of Forests, 1981) and has continually demonstrated interest in finding ways to better relate land-use management plans in visually sensitive areas to individual preferences (e.g. BC Ministry of Forests, 1994a, 1996, 2001a, 2003). The BC Ministry of Forests provides guidelines and procedures for helping companies manage these sensitive areas (BC Ministry of Forests, 1997b; Marc, 2008), and has provided training to foresters as well (Marc, 2010). The procedures have attempted to make the management and assessment of visual resources more objective. This chapter has shown how geospatial technologies can further improve current procedures by helping to identify visually sensitive areas, delineating VSUs, and calculating the amount of visible alteration using the HCV. This chapter has also provided a review about some inherent problems with using a limited number of viewpoints to represent the potential observer experience throughout a landscape, and alternative approaches that could be used for making more accurate assessments of the effects of harvest plans.  118 Chapter 5:  Trade-offs and Limitations for Visual Resource Management This space between ecology and aesthetics has numerous potential conflicts (Sheppard, 2000). From a purely ecological point of view, aesthetics can seem irrelevant. From an aesthetics view, the ecology of a stand is not of primary importance. For forest planners, both aesthetics and the ecological health of a stand are critical components of planning, especially in areas of visual sensitivity. These three perspectives are addressed throughout this chapter; the interactions, limitations and opportunities for win-win solutions are also presented. 5.1 Important Aesthetic and Ecological Considerations for Forest Management The dichotomy between aesthetics and ecologists has been presented in Kimmins (1999), where he states that the perceptions of sustainability and good forest stewardship by the average citizen, will at times be different from the ecologist. This is not news to foresters. Sheppard (2000) suggests that from personal experience, professional foresters have not found the idea of managing visual resources to be very important, except perhaps for political purposes. The protection of visual quality in our forests is often seen as a constraint to foresters, as there is generally a reduction in the amount of timber that can be harvested within visually sensitive areas, compared to those outside of a sensitive area (see BC Ministry of Forests, 2003; Picard and Sheppard, 2002a; Picard and Sheppard, 2002b). It is important to note that VRM has primarily focused on restricting the ‘seen’ effects of harvesting, relegating operations to the backcountry in an attempt to avoid the public’s ire (Gobster, 1999; Sheppard et al., 2004), but that even if clearcuts are invisible, it does not necessarily make them acceptable (Ribe and Matteson, 2002). It may also be that a growing emphasis on ecology and forest health has decreased our focus on aesthetic quality. Economics, social welfare and ecological elements all play a role in ensuring that forests provide for the needs of people, wildlife and the environment. In areas of multiple-use, dealing with these trade-offs between ecology, aesthetics and operations, is important and requires some knowledge about the interactions of these elements in order to make wise management decisions. Numerous responses have emerged to address the arguments from recent and longer-term industrial pressure to harvest in visually sensitive areas. For instance, regarding the argument  119 for increased salvage of beetle wood in scenic areas, the BC Ministry of Forests has enabled some flexibility in areas drastically impacted, but have stated that under most circumstances it is possible for the VQOs to be maintained (Marc, 2007). The longer-term issues have also been dealt with to some degree. Regarding timber supply limitations, using alternative silvicultural practices has shown the potential to increase the available timber within a VQO (Picard and Sheppard, 2002a, b) and a report conducted in the Robson Valley TSA showed that partial cutting could increase timber supply by 36% in the area (Schuetz, 1998). Even though management of visual resources may not be as high on the list of competing goals as more traditional values in the minds of foresters, there are a number of social reasons that management of these resources are vitally imperative. For instance, modifications which affect aesthetics, such as clear cuts, may elicit negative emotional responses (Bliss, 2000; Parsons et al., 1997). It is well known that these responses, caused by these kinds of stimuli, can impact an individual’s opinion about the management of the environment (Newhouse, 1990), and that these emotional responses can leave lasting impacts. From an intrinsic perspective, beautiful landscapes are a source of public information and pleasure (BC Ministry of Forests, 1996; Lucas, 1991). Areas that are naturalistic, wooded and have scenic roadside views have been linked to quality of life and associated health benefits (Kaplan et al., 1998; Ulrich, 1986). Natural landscapes can reduce stress (Ulrich, 1981) and provide opportunities for better mental and physical health (Bishop and Hull, 1991). There have also been numerous studies that demonstrate an economic benefit from the presence of scenic areas. Urban forests, suburban forest and vegetation preserves have been shown to increase property values (Devitt, 1988; Kim and Johnson, 2002; Seila and Anderson, 1984; Thorsnes, 2002; Tyrväinen, 1997). Kim and Johnson (2002) also showed that at the time of purchasing, prices of homes in Corvallis, OR were lower if clear-cuts were visible. Iverson (1997) also showed that property values were negatively affected by landscape changes. More generally, scenic byways have been shown to generate revenues from non-residents in scenic areas (see Sheppard et al., 2004). Even beyond the documented literature, it is not difficult to argue that aesthetics are important to people, that scenic areas are important for tourism and recreation, and that beautiful places may influence nearby property values.  120 As stated earlier, VRM has largely served its purpose of mitigating public ire. In some cases, this has had a repercussion of simply limiting modification in the front country. But what is the reaction when significant changes to the landscape are not controllable? For instance, in recent years the mountain pine beetle attack has dramatically altered the landscape in BC, with much of the attack existing along scenic corridors, recreation areas and in nearby communities (BC Ministry of Forests and Range, 2010). The effects of these attacks do, generally, negatively impact visual quality ratings from the middle ground view (Sheppard and Picard, 2006). A few studies have looked into individual preferences for managing pests or fires within national parks (e.g. McFarlane et al., 2006; Müller and Job, 2009), but these may not directly address the impacts to areas outside of parks (McFarlane et al., 2006). These kinds of impacts stir up debate with respect to the balance of values (ecological, economic, aesthetic, or others values) in forest management. As the move toward ecosystem management has progressed, various attempts to bridge the gap between ecology and aesthetics in forest operations and planning, have been made. This includes calls for an ‘ecological aesthetic’ (Gobster, 1995, 1999). However, as Sheppard (2000) points out, the connection between ecology and aesthetics has not been well addressed. Yet, there is a clear argument that ecological- and aesthetic-based management regimes cannot always be aligned (Parsons and Daniel, 2002), but that they can augment one another to serve their own purposes (Gobster et al., 2007). 5.1.1 Variable Retention: Ecological and Aesthetic Implications on Forestry One of the recent trends in forestry has been with the practice of variable retention (Franklin et al., 1997). From an operational perspective, variable retention is intended to retain structural elements of the forest by preserving some of the existing forest (Franklin et al., 1997). Generally, this requires that at a minimum of 10% of the trees within a stand be maintained (Sougavinski and Doyon, 2002). From a visual quality perspective, retention of trees, especially if distributed, does seem to increase acceptability of harvests when compared to clearcuts (Meitner et al., 2005; Ribe, 2005; Ribe, 2006; Ribe, 2009), but only if the level of retention exceeds 15% (Aubry et al., 2004; Aubry et al., 2009), with broad acceptability when retention approaches 40%-50% (Aubry et al., 2009; Ribe, 2005). This is  121 why the BC Ministry of Forests Protocol for Visual Quality Effectiveness Evaluation (Marc, 2008) encourages retention within blocks. So, perhaps there is a win-win solution if variable retention can be coupled with a harvest that can distribute trees throughout the cutblock in order to increase its acceptability. Variable retention has been suggested as a means to reducing the visual impacts in visually sensitive areas with highly restricted VQOs (BC Ministry of Forests, 2002), and it has been adopted by companies in areas where visual integrity is important (e.g. see Picard and Sheppard, 2002a). Yet, after a decade of research on the ecological implications of variable retention, there are some interesting data that planners must consider before implementing variable retention. First, it is important to clarify that variable retention can mean the trees retained within a stand can either be retained in aggregate fashion (clumped/grouped) or dispersed throughout the harvest, each having unique costs/benefits ecologically (Franklin et al., 1997), operationally and visually. For instance, a dispersed pattern of retention has a lower probability of retaining diverse tree sizes and conditions, is limited in keeping undisturbed forest floor understory, and is likely more expensive to create. A summary of these differences can be found in (Sougavinski and Doyon, 2002). Additional research has been completed and provides interesting implications for dispersed variable retention and the health of the retained portion of the forest within a harvested stand. For instance, a 25% dispersed retention of pre-harvest overstory can reduce the early growth of montane conifer seedlings, likely because of a reduction in available light (Mitchell, 2001). Although this may be beneficial for reducing some of the negative aesthetic impacts, and maintaining some ecological structure, it may reduce the potential for timber production. Other studies have verified that the reduction of light is steep as retention level increases (Aubry et al., 2009; Halpern et al., 2005). Reductions in growth were not limited only to the first years after dispersed harvesting, rather it continued after seven years, where both amabilis fir and western hemlock showed a reduction in height growth in dispersed retention patterns of 25% (Mitchell et al., 2004). Although dispersed retention can be ecologically beneficial in certain circumstances, there is more for planners to consider than just the effects of seedling growth. In dispersed retention,  122 windthrow can also reduce the number of established standing trees, depending upon the specific type of retention (Beese and Bryant, 1999). So, not only is there a potential for effecting the growth of seedlings, but now there is also a potential for reducing more valuable timber within a stand. Furthermore, damage to tree boles was higher with dispersed compared to aggregate retention (Aubry et al., 2009). In the coastal forests of BC, root rot, hemlock dwarf mistletoe, and decay pathogens are cause for concern, but can be managed through more careful planning (Beese et al., 2003). Part of this planning process involves identifying stands where group (aggregate) retention can be used to limit these effects, but in areas of higher risk, dispersed retention may not be a viable solution (Beese et al., 2003). In general, tree mortality was low for both types of variable retention at 40%, but at 15% the dispersed treatment have significantly higher mortality, near 7% (Aubry et al., 2009). Reductions in timber production have been realized in retention harvests (North et al., 1996), including varying retention levels and rotation lengths (Hansen et al., 1995). Franklin et al. (1997) observes that dispersed retention may have a larger negative impact than aggregate in terms of stand growth ultimately leading to a reduction in wood yields, which is also observed in Aubry et al. (2004). Also, in order to maintain a greater variety of structural elements, and subsequently ecosystem components, of the stand aggregate retention is the preferred type (Franklin et al., 1997). The effects of disease and mortality beyond the harvested timber are not only important to account for in operations because of the potential reduction in revenue, but also because these can further the negative aesthetics effects. For this reason, harvest designs must allow for the death of uncut trees on the landscape, so that the aesthetics of the harvest are not greatly influenced by the death rate occurring from windthrow, disease and other forms of mortality. From an ecosystem perspective, with ecosystem types that are adapted to large disturbances, introducing a disturbance like retention may not mimic the larger disturbances typical of this type. Using a dispersed retention treatment may initially provide for an increase in growth for the retained trees (Aubry et al., 2009), which is also good for operations, but can influence the growth of the understory causing a shift toward species which are more tolerant of competition (Palik et al., 2003). For instance, shade tolerant species will likely be favored, which could impact the ecosystem and possibly reduce the stand’s collective growth (Hansen et al., 1995; Kimmins, 2004; North et al., 1996). If the objective of forest management is to  123 ensure an early- or mid-seral stand the appropriate disturbance must be created to ensure favorable growth of this species, in some cases this may mean clearcutting (Kimmins, 2004). However, depending upon the seral stage of the forest, clearcutting may alter the normal succession of the stand, causing unwanted impacts. The overarching message in the literature about variable retention, is that there is no single configuration that works across all ecosystems and locations to attain the ecological benefits of variable retention (Aubry et al., 2004). This means that in many cases, planners must consider how the particular implementation of variable retention, and subsequent effects, will influence the ecological health and the aesthetics of the forest while also ensuring the company’s bottom line .Specifically, landscape planners need to be aware that even if the design minimizes the visual impacts in their models initially, there might be longer-term implications if the retained trees in a stand die and reduce the visual quality of the original design. Alternatively, for ecologists, it is important to note that even if a harvest is ecologically sound, it may not accomplish the visual quality requirements and social needs for tourists, residents and communities. 5.1.2 Implications for Emulating Natural Forest Disturbance on Aesthetics Variable retention represents one kind of management where foresters try to emulate natural patterns and maintain ecological structures important to the forest. Recently, ecologists have expanded this notion to identify regionally-specific silvicultural practices based on natural disturbances. This is referred to as emulating natural forest disturbance (ENFD) (Crow and Perera, 2004). The discussion of how ENFD may interact with planning in visually sensitive areas is presented here. ENFD argues that the ecological health of forests may be best maintained by following the patterns of the kinds of disturbances that would naturally occur. For instance, late- successional undisturbed forests in Northeastern US forests are patterned by numerous small patches of late-seral species (Seymour et al., 2002).  So, natural emulation of these disturbances suggest that harvesting be done in relatively small patches ~.2 ha, whereas the practice of harvesting in 1-3ha was not necessarily ecologically sound except when foresters left a diversity of age or vertical structure (Seymour et al., 2002). In a black spruce forest  124 ENFD looks different. Bergeron et al. (1999) proposes that there is a three stage natural cycle that foresters can emulate: first, a fire can be emulated by a clearcut; then, understory development can be emulated by partial cutting; and finally, selection harvesting can emulate natural gaps of old growth stands. And, in long-leaf pine forests, harvesting using group selection can simulate small frequent fires (Franklin et al., 2002). In the Pacific Northwest US (as in BC), natural disasters come in a variety of frequencies, scales and types (North and Keeton, 2008). Large scale fires can happen more frequently in the drier forests (~200 years) compared the more rare large-scale coastal forest fire (~1000 years) (Agee, 1993), while wind can disrupt huge amounts of coastal forests (see North and Keeton, 2008). Aside from the large-scale disturbances, wind can be a particularly common event and can creating many gaps of smaller sizes throughout coastal forests (Spies et al., 1990). Whether through wind, disease or other kinds of smaller disturbances, the changes in forest cover tend to be limited in scale. Over the long-term, these kinds of processes can create a patchwork of numerous small “gaps” in the forest cover, and in BC these tend to favor late-seral species (Kimmins, 2004).  From an ecological perspective using ENFD as a management foundation works well to ensure ecological integrity that keeps the more natural successional processes going. Yet, there may need to be some changes in silvicultural practices. For example, in Douglas-fir forests, traditional clearcutting does not emulate natural disturbance because it leaves little or no above-ground structure, whereas large disturbance windthrow and fire do leave these elements (Franklin et al., 2002). ENFD might suffice as an appropriate means for ecological management, but there may be challenges of using ENFD within visually sensitive areas. Figure 5.1 is presented in order to provide a general framework which can help clarify the interaction between ENFD and visual quality. The examples used for this figure are oversimplified and mean only to demonstrate the kinds of questions planners must consider if using ENFD as a design criteria. As stated in the example in the previous paragraph, , ENFD would suggest leaving aboveground structure and debris within clearcuts of Douglas-fir forest. Although leaving aboveground debris is beneficial for ecological purposes, these practices are not consistent  125 with the prevailing scenic aesthetic (Sheppard, 2000).This relationship is depicted as C in Figure 5.1, where there may be good ecological reasons for clearcuts and aboveground structure, but visually this may not be preferred. This situation might exist in forests naturally dominated by lodgepole pine. In these forests larger disturbances such as fire are more frequent, so ENFD would suggest mimicking the frequency and size of this disturbance. Yet, the prevalence of large and frequent disturbances, like clearcuts, are contrary to what researchers understand about preferences. In this scenario it might be possible to reduce the visual impacts by reducing the size of harvest and/or frequency as shown in the arrow extending from C, pointing toward the middle of the figure. It may also be possible to strategically place retention blocks or distributed retention throughout the stand as long as it provides the best gain in aesthetics for the lowest ecological effect.  Figure 5.1 Relationship between visual quality and sustainability with examples (adapted from Sheppard, 2000). The flipside of this scenario is in a lodgepole pine forest where aesthetics are carefully managed and the amount of visible alteration significantly restricted. These restrictions are good for visual quality, as shown as B in Figure 5.1, but may not be consistent with natural disturbance of these forests. Bell (2001) expands on this notion, stating that in lodgepole pine  126 dominant forests the species rarely survives beyond 100 years and that in an ecologically- driven practice, ENFD would suggest that larger and more frequents harvests are appropriate. This might be appropriate for ENFD, but the frequency and size of harvests may likely be severe for visual quality. Bell (2001) points out that should foresters devise plans that restrict the amount of harvest in these existing forests (possible for aesthetic reasons), the structure and age class of this species would extend beyond their natural state. From an ecological perspective this would result in an increased likelihood for a disturbance such as fire or pest, and of course from a visual perspective, these would not be favorable either. An alternative outcome would be a premature succession whereby Douglas-fir would begin to occupy the canopy space (Kimmins ch 17). This may be a viable solution visually, but not necessarily ecologically. However, from an operational perspective, both of these cases represent increased risk. First, because of the increased likelihood of disturbance and second, because during the successional stage, the growth of lodgepole pine would be significantly diminished and it would take some time for Douglas-fir to reach its peak growing condition. In this scenario, a possible increase in harvest size and/or frequency, shown as the arrow extending from B in Figure 5.1, might help reduce some of the ecological and operational risk, but would negatively affect the visual quality. This situation, and that shown in C, represent a “Catch 22” for forest management. At B, the public might appreciate the aesthetic quality of a reduction in harvesting, but there is a high risk for natural ecological disturbance (e.g. fire), whereas in C, the forest may maintain a natural disturbance regime, but there may be the social risk of public outcry from frequent and large harvests. However, there may be cases when ENFD in visually sensitive areas may be able to function simultaneously with good aesthetic design as depicted. This solution space is depicted as A in Figure 5.1. An example of this scenario is when the natural succession of a forest promotes gaps from smaller-scale disturbances. These kinds of disturbances are more prevalent in coastal forests. A would exist at VQO levels of preservation or retention by using ENFD of small gaps or openings simulating wind, disease or other kinds of death. Harvest amounts that extend beyond these VQOs are represented by the arrow extending from A. However, it is well known that these gaps are relatively small, and as a result may increase operational costs (Palik et al., 2002). Partial cutting and variable retention have been put forth as possible alternative practices as they can increase timber availability in these areas. However, even  127 these harvest practices may not effectively emulate the natural cycle of disturbance if the openings are larger than the natural cycle would produce. Where ENFD and aesthetics function well together, the ecological and visual effects over time and space are well matched. In this example, the amount of forest harvested would be small enough to enable substantial re-growth before the next harvest. This re-growth mitigates the visual effects through time, at a rate that causes minimal visual disturbance. However, when the scale of harvest is too large, the scenic quality will most-likely suffer if large clearcuts are easily visible. This negative effect is referred to as the “aesthetic dip” (Ribe, 2005, 2009). Fortunately, the dip in aesthetic quality quickly decreases as time passes. The return to a similar level of scenic quality occurs when a stand reaches a state planners refer to as “green-up” (e.g. BC Ministry of Forests, 1994a; Ribe, 2005; Sheppard and Picard, 2006), but this can take a few decades depending upon numerous environmental (BC Ministry of Forests, 1994a). So, just as the size of harvest is important, the potential influence of time can plan an important role in planning operations, whether for ecological or aesthetic purposes. Just as the size of harvest might need to be reduced or spatially distributed over the landscape, so too can time be used to distribute when harvests are completed in order to reduce the potential visual impacts. 5.1.3 Ecological Considerations of Using Visual Magnitude to Guide Planning Another consideration of the ecological-aesthetic interface pertains to site selection for areas of harvesting. Gobster (1999) stated that the management of visual resources in the past has served its original purpose of mitigating visual impacts by simply reducing the visible impacts; essentially, hiding the harvests.  It is possible that the HCV could be used to further this practice by identifying areas where planers should avoid due the high amount of Visual Magnitude. One of the key elements of Visual Magnitude is the influence of slope.  Places where the relative difference in slope between the observer and the surface reduce Visual Magnitude. If calculated from highways or water bodies, places with lower slopes, such as valleys and plateaus, would generally have lower Visual Magnitude values. This makes these areas more enticing to harvest when managing for aesthetic purposes and may also be more ideal locations for harvesting because of the reduction in possible landslide risk.  128 However, it is crucial that planners understand the potential implications of harvesting in forested valleys and lowlands as they can represent important ecosystems. According to Jeo et al. (1999) valley-bottoms and mid-slope lands along the coastal regions of British Columbia represent some of the most critical areas for biodiversity.  According to Young (2000), riparian areas along the West-Coast of North America represent unique biological and physical processes within the forest. For instance, Wells et al. (2003) found that along the Central Coast of BC, two important and underrepresented ecosystems within riparian zones exist: Yellow-cedar bog and black cottonwood ecosystems. Yet, riparian zones like these are still part of the Timber Harvesting Landbase in BC (Moola et al., 2004). It is likely that some of these areas would also fall within the protection of a visually sensitive area, although this cannot be confirmed until an analysis is made to determine this overlap. However, for argument sake, if this assumption is carried forward, it would likely mean that portions of the riparian zones specified in Wells et al. (2003) would fall under the umbrella of visual protection in BC. According to BC’s VLI, the establishment of a Visually Sensitive Unit (VSU) is usually delineated by identifying the valley bottom and the ridgeline from bottom to top of a mountain or hillside, and is typically bounded on each side by valleys on the sides of the mountain  (BC Ministry of Forests, 1997b). One of the main criteria of visual protection is to reduce the visible amount of modification within each VSU (Marc, 2008). This is often easily done in areas of low slopes because the visible effects of harvests are reduced (Anderson et al., 1979). However, in the case of riparian zones, harvesting primarily in valleys and lowlands should involve careful consideration of the trade-offs between aesthetic quality and ecological integrity. One proposal that may allow for both the benefit of ecological and aesthetic purposes would be in the protection of steep slopes where limitations of technology, inaccessibility or factors have so far limited the management of these areas (Wells et al., 2003). According to Wells et al. (2003) areas that have been inaccessible or infeasible for management in the past, could serve as conservation sites, allowing for a more intensive use of the managed areas, but as long as the ecosystems in the unmanaged areas adequately represent the ecosystems of the managed areas. In these specific circumstances, the areas of high Visual Magnitude or visibility would be protected, while simultaneously conserving important ecosystems. If this  129 proposal were to be ecologically sufficient, it might be also be aesthetically sufficient, but it is important to understand that the proposal infers an important element landscape planners must consider. One of the assumptions of this proposal is that these unmanaged areas are meant to provide room for error when good intentions to protect species in the managed areas fail (Bunnell et al., 2003). What this means is that these areas should allow for natural ecosystem functions to exist (e.g. such as fire and disease), where these would normally be suppressed in managed stands (Wells et al., 2003). This proposal may present a bit of a dichotomy for planners designing harvest for aesthetic purposes. While it might enable for more intensive management of forests in lowland areas, these areas might also be ecologically important. Also, while the proposal limits harvesting in steeper slopes, which are more likely to be more visually significant, those areas may incur natural disturbance, such as fire, which might impact the visual quality. In situations where planners must deal with multiple-objectives, like the ecological-aesthetic challenges, the HCV can be used to aid in harvest planning. While the HCV does reveal areas of no visibility or low visibility, it is not intended to merely allow planners to merely hide harvests. Rather, in areas like uplands or steeper slopes, it can aid in the strategically positioning harvests to reduce the overall visual impact. This could mean that in areas like the Central Coast of BC, harvesting activities could exist in the uplands, where a greater abundance of similar forested types exist (Wells et al., 2003), while reducing the intensity level of harvesting in riparian zones for conservation purposes. 5.1.4 Conclusions This space between ecology and aesthetics has numerous potential conflicts (Sheppard, 2000). From a purely ecological point of view, aesthetics can seem irrelevant. From an aesthetics view, the ecology of a stand is not of primary importance. However, in visually sensitive areas, both perspectives are limited in some way by the other. Planning only with a forest ecology perspective could void many of the important social components that aesthetics provide (reduction in stress, health benefits and quality of life, pleasure, etc.). For instance, perhaps a large clearcut is the best ecological option for a pine forest near a park or suburb where people live and recreate by. This harvest practice is likely to cause negative  130 reactions. Alternatively, planning for purely aesthetic reasons may impact the long-term natural succession of a forest or overall health of the ecological system which could have implications for the environment; possibly affecting the visual quality if it leads to disease, fire, or windthrow in stands that may otherwise not typically have these problems if managed ecologically. There is no one-size-fits-all, win-win, solution that can solve both problems simultaneously. The idea of the ‘ecological aesthetic’ (Gobster, 1995, 1999) cannot apply in all situations, especially where the ‘scenic aesthetic’ is important for social reasons alone (Parsons and Daniel, 2002). There is also evidence that education or information may not change an individual’s perceptions about the environment, even if it is ecologically justified (Hill and Daniel, 2008). This serves as an important reminder that, as humans, stimuli that cause an emotional reaction still leave an impression on us (Zajonc, 1980). It would be satisfying to say that the ‘holy grail’ of an ecological aesthetic is attainable, but that is not a reasonable conclusion. While the tug-of-war between ecologists and aestheticians continues, it will be vitally important to try to align human values and ecological goals where possible (Gobster et al., 2007).  131 Chapter 6:  Conclusion This thesis has developed several new insights to foster more effective management of forests in visually sensitive areas.  The research explored ways that geospatial information and visualization technologies can be used to more objectively define and assess visual quality. Several practical examples have been demonstrated, and many important considerations for future development have been elucidated. Additionally, trade-offs existing between aesthetics and other forest management objectives have been discussed, particularly relating to new ecologically-driven forest management practices. The overarching goal of this research was to develop new ways for technology to aid forest planners to more effectively design harvests in visually sensitive areas. The two major contributions of this research have been: a more objective description of how harvest shapes can influence human preferences and the development of the HCV. The first contribution demonstrated how the HCV can be used to better delineate visually sensitive areas, measure the landform as it would be seen along a route, and to automate the calculation of the amount of visible alteration.. The second contribution demonstrated the effects of harvest shape on individual preferences and explored the use of spatial metrics for VRM, particularly to see if specific metrics could serve as indicators for visual quality based on its correlation with design preferences. The remainder of this chapter summarizes the most important conclusions from previous chapters, and synthesizes how these findings can improve VRM. Then, ideas for the continuation of this research into the future are presented. Finally, the overall significance and contribution of this research for academic and VRM practitioners will be broadly discussed. 6.1 Summary of Conclusions Chapter 1 provides an argument for the management of aesthetics in forestry by emphasizing how the paradigm shift in forestry, from the management of traditional forest products to a more holistic approach (Mery et al., 2005), may include the management of forest aesthetics for social benefit. For instance, nature, and its aesthetic quality, provides numerous services  132 to people, including the positive effects on human health (e.g. Kaplan, 1995; Kaplan and Talbot, 1983; Parsons, 1991; Rossman and Ulehla, 1977; Ulrich et al., 1991). If the management of forests can embrace alternatives practices which benefit human health and well-being, it is may be possible that subsequent economic opportunities present themselves (Mather, 2001). Forest management in BC poses a unique opportunity to explore these alternative benefits through the management of aesthetics. Like many places around the world, forests in BC play a major role in the economic health of its citizens. Traditional economic derivatives, such as wood, pulp and paper, constitute a large portion of the province’s economic foundations (Schrier, 2011). Yet, of the total amount of economic output from the forest, over 30% of the bottom-line derives from forest recreation opportunities (BC Ministry of Forests, 2010). With such a significant portion of the economic activity derived from recreation, it is important to ensure that the aesthetic quality of the landscape is maintained so that the opportunity of these activities are not lost to traditional timber management (e.g. BC Ministry of Forests, 2003). As a result The BC Ministry of Forests has established the protection of visual quality as one of its eleven resource values (BC Ministry of Forests, 2004). A VRM protocol has been created in order to help planners more effectively preserve the visual integrity of the forests they manage (Marc, 2008). However, this system can be improved. First, the current system uses only a few viewpoints from which harvest plans must be assessed for their visual quality (see Marc, 2008). It would be more appropriate to instead, find ways to objectively measure visual quality as it would be experienced by an individual moving within the landscape as demonstrated by the HCV. One of the possible improvements would be to use GIS to measure important visual criteria, such as the percent of visible alteration (shown in section 4.2.2), from a number of locations in the landscape. Another improvement was suggested; wherein methods that could more objectively measure visual quality would be used (Chamberlain and Meitner, 2009; Ode et al., 2009; Ode et al., 2010; Ode and Miller, 2011; Ode et al., 2008; Sang et al., 2008; Tveit et al., 2006). Many of the visual criteria used in the Effectiveness Evaluation are based on subjective judgments by experts. Not that these are inherently incorrect, but by developing tools, such as the HCV,  133 and mathematical expressions to measures the designs, as shown in section 2.5, planners would be able to better gauge the potential visual impacts of their plans before harvest operations begin. Chapter 2 presents new findings on how harvest block clearcut shapes in forest management can influence individual preferences. Using three characteristics of shape: geometric primitive, complexity and aspect ratio, the quantitative effects of these characteristics are explored. This study reveals that geometric primitive and complexity are nearly equally important as an influencing factor in perceptual ratings. As a general principle in landscape design, shapes which fit to the landscape and appear more natural are usually portrayed as better designs (Bell, 1999, 2004; Carlson, 1977; Crowe and Mitchell, 1988; Shafer, 1967). This study revealed the quantitative differences of the preferences, confirming that circular shapes with smooth undulating edges were far more preferred than more blocky harvest designs. Another intriguing finding was that a mid-level of complexity provide large preference increases, and that across geometric primitives, the jump from a mid-level to high-level of complexity does not increase preferences as much as the initial jump to a mid- level of complexity. Yet, when the geometric primitives are separated into their four categories, what was found is that at high levels of complexity, trapezoid and triangle primitives do differ from the preference ratings of squares. Also, that at the highest level of complexity, the preference for circular and rounded-edge shapes always increased, but marginally so at the high level of complexity. To landscape designers, these finding confirm the influence of shape on perception, but does so in a new way. The finding represents first efforts to quantitatively describe how harvest block shape can influence preference. The exploration of the quantitative measures of these different shape criteria have laid the groundwork for possibly developing shape indicators that can be used to mathematically associate values from these indicators. However, this idea was explored with mixed results as discussed in section 2.5. In regards to the shape indicators used for the analysis, there seems to be a range of low to medium amount of correlation with the image ratings of all shapes. However, when shapes are separated by primitive, there seems to be a high degree of correlation between several of the shape indicators and image ratings. At this point, the research does not provide conclusive evidence for using shape  134 indicators as a proxy for preferences, but it does pave the way for continued research in this area. In Chapter 3, the development of the HCV was presented. This algorithm is capable of portraying visibility in terms of how much a space on the terrain occupies within the viewer’s field of view and considers the degree of visibility as it may be seen from along a rount. A critical component of this algorithm is the calculation of Visual Magnitude, which combines the perceived surface of the earth with the limitation of the human eye. By integrating Visual Magnitude with the efficient viewshed algorithm, XDraw, this method provides a more appropriate measure of the way in which people ‘see’ the landscape compared to a binary form of a typical viewshed analysis. Two applications of the HCV were presented in Chapter 4 and discussed with regard for its potential to improve VRM in BC. First, the HCV can be used to more precisely locate visually sensitive areas, and can be used to understand the kinds of landform that an individual would see along a highway drive. Second, the HCV can be used to help planners more efficiently design harvest by: automating the calculation of the amount of visible alteration, saving planners time by quickly depicting areas on the landform which are less likely to be obvious, and providing a way to make these assessments from numerous viewpoints, such as along a highway route, hike or boat trip. The outcome is a better accounting for the full range of perspectives and visual criteria for which assessments using 3D simulations are simply impractical to produce. Chapter 4 also outlines considerations for landscape planners when designing harvests in visually sensitive areas and using geospatial technology for analysis. From a technical perspective planners must be aware that not all viewsheds are created equal, and that there are inherent errors (Fisher, 1991, 1992). A method for improving the current importance weighting (see Marc, 2008) of viewpoints within VRM is suggested. Frequency data, such as traffic counts, can be combined with a ratio of time spent at a viewpoint along a road corridor to derive a more objective measure of importance. These values provide spatial input which can be used to generate a weighted Visual Magnitude measure for all places seen from any of the locations along a corridor. Another suggestion proposes an alternative to using a single viewpoint location by instead creating a spatial distribution of the likelihood of viewing  135 locations within an area of interest. Finally, a discussion about the potential limitations of viewshed analyses is given. In Chapter 5, an overview regarding the trade-offs between aesthetics and ecology, related to forest operations, is given. Emphasis is placed on new forest management practices modeled after variable retention and emulation of natural forest disturbances. These two management methods have been proposed to reduce the negative ecological impacts sometimes inflicted by traditional forestry. Some of these practices, including variable retention, whether grouped or dispersed, can also increase the visual quality of a harvest. However, planners would be wise to understand how the distribution and amount of retention may cause unintended effects that reduce the aesthetic quality of a harvest (e.g. additional mortality from disease, root rot, and windthrow). In the same respect, managing a forest for ecological interests may result in a decline of aesthetic quality in some situations. In cases where the alterations are extreme there could be implications for public outcry, especially if the harvest produces large visible clearcuts (Boerner, 1986; Williams and Tolle, 2001). Of course, in visually sensitive areas, the primary concern is with the aesthetic quality of the landscape. However, this does not mean that forest practices should simply hide harvests from view. Rather, planners should find ways to try to align human values and ecological goals where possible (Gobster et al., 2007). 6.2 Future Directions The research presented in this dissertation, lays the foundation for the development of additional ways to more objectively assess aesthetics in the visually sensitive areas in BC. However, there is much research to be done before automating the assessment of visual quality. A few of the logical next steps for this research are presented next. First, one of the key characteristics of aesthetics is in the measure of percent of visible alteration (BC Ministry of Forests, 1996; Marc, 2008; Shang and Bishop, 2000). A first attempt at using the HCV to quantitatively measure this has been developed and tested. However, in order for this kind of analysis to be considered an alternative to the 3D simulations, additional validation experiments much be done. However, the accuracy of the 3D simulations should also be questioned. Therefore, it would be best to sample a number of  136 real-world, harvested landscapes, where the percent of visible alteration has been verified by the BC Ministry of Forests or forest companies. It will also be important to carefully deal with problems of varying resolutions. It may very well likely that with low resolution data, such as in the example with 30 m cell size, the use of Visual Magnitude may be irrelevant. However, it is also likely that the accuracy of the assessment is a function of both resolution, distance to the harvests and angle of view. All these should be explored in further research. Second, the empirical research which suggests that shape can play a role in preferences opens a number of new possible research areas. The literature has shown the perspective amount of alteration is a primary influence on preference ratings (BC Ministry of Forests, 1996; Shang and Bishop, 2000). However, as shown in Chapter 2, when the percent of visible alteration is controlled, shape does influence ratings. So, it would be interesting to explore the trade-off between the amount of visible alteration and the design of the shape. For instance, it should be possible to show how much more visible alteration would be allowed if a harvest design was created with curved edges rather than hard angular edges. Another interesting study would be to explore the influence of the shapes, sizes and pattern of retention blocks relative to the shape of the external boundary of a block. With positive feedback regarding group retention for both visual and ecological aspects of forestry, conducting research into the preference effects of varying these retention block shapes may provide beneficial and relevant information to planners. Third, with additional empirical research on the perceptual effects of shape design, it may be possible to further the exploration of using spatial metrics, such as those provided by McGarigal and Marks (1995), as indicators for visual quality. Indicators like these have been used in previous studies about landscape complexity pertaining to perception (Ode and Miller, 2011; Ode et al., 2008; Sang et al., 2008), but no studies have explicitly investigated how these indicators pertain the preference of harvest block shape design. However, this work will require a substantial amount of additional empirical research that can help bridge the gap between what a metric is reporting and what is actually preferred. With the lack of empirical research in the literature pertaining to shape design in forestry, it would be unwise to establish a metric for shape design that pertains to visual quality. Just as landscape ecologists are concerned with the connection between the measurement of a pattern and the  137 actual ecological process, landscape planners need to be concerned with the process of human perception and how it relates to an indicator. 6.3 Concluding Remarks The tools and insights presented in this dissertation open new doors to better understanding how humans interact with, experience and see the landscape. These new discoveries, build upon decades of research in VRM, and further upon psychology, computer science and landscape design. This research has resulted in the development of new information which may lead to more effective management of forest in visually sensitive areas.  This research has explored ways in which geospatial information and visualization technologies can be used to more objectively define and assess visual quality. As long as these objective measures relate to human perceptions, they can provide helpful information to foresters working in areas of high scenic value to the public. As foresters continue to develop plans to meet multiple objectives, including those set for in sustainable forest management, it will be imperative to provide them with the necessary tools and knowledge to make wise decisions. The balance of these objectives plays a crucial role in ensuring that the services and products derived from forests that we are so dependent upon, including aesthetics, are maintained for future generations.  138 References Agee, J. K., 1993, Fire ecology of Pacific Northwest forests, Island Press, Washingont, D.C. Amir, S., Sarig, G., 1977, Planning the multiple use of maqui land, Landscape Planning 4:359-373. Anderson, L., Mosier, J., Chandler, G., 1979, Visual Absorption Capability, in: Proceedings of our national landscape: A conference on applied techniques for anlysis and management of the visual resource (G. Elsner, R. Smardon, eds.), USDA Forest Service Pacific Southwest Forest and Range Experiment Station, Berkeley, CA, pp. 164-171. Arthur, L. M., 1977, Predicting scenic beauty of forest environments: Some empirical tests, Forest Science 23(2):151-160. Attneave, F., 1957, Physical determinants of the judged complexity of shapes, Journal of Experimental Psychology 53(4):221-227. Attneave, F., Arnoult, M. D., 1956, The quantitative study of shape and pattern perception, Psychological Bulletin 53(6):452. Aubry, K. B., Halpern, C. B., Maguire, D. A., 2004, Ecological effects of variable-retention harvests in the northwestern United States: the DEMO study, Forest Snow and Landscape Research 78(1/2):119-137. Aubry, K. B., Halpern, C. B., Peterson, C. E., 2009, Variable-retention harvests in the Pacific Northwest: A review of short-term findings from the DEMO study, Forest Ecology and Management 258(4):398-408. B. C. Ministry of Forests, 1974, Visual Landscape Inventory: Procedures and Standards Manual. Bacon, W. R., 1979, The visual management system of the Forest Service: Gen. Tech. Rep. PSW-GTR-35., in: A conference on applied techniques for analysis and management of the visual resource (G. H. Elsner, R. C. Smardon, eds.), Pacific Southwest Forest and Range Exp. Stn., Forest Service, U.S. Department of Agriculture, Berkeley, CA, pp. 660-665. BC Integrated Land Management Bureau, 2011, Land and Resource Data Warehouse, BC Integrated Land Management Bureau, Victoria, BC. BC Ministry of Forests, 1981, Forest Landscape Handbook, Ministry of Forests, Victoria, BC. BC Ministry of Forests, 1994a, A First Look at Visually Effective Green-up in British Columbia: A public perception study, Forest Practices Branch and Recreation Branch, Victoria, BC. BC Ministry of Forests, 1994b, Visual Impact Assessment Guidebook: Completing visual force and land feature analyses (Appendix 4), Forest Practices Branch, Victoria, BC. BC Ministry of Forests, 1994c, Visual Impact Assessment Guidebook: Visual Impact Assessment Summary Form (Appendix 2), Forest Practices Branch, Victoria, BC. BC Ministry of Forests, 1994d, Visual Landscape Design Training Manual, Forest Practices Branch and Recreation Branch, Victoria, BC. BC Ministry of Forests, 1996, Clearcutting and Visual Quality: A Public Perception Study, Forest Practices Branch and Recreation Branch, Victoria, BC. BC Ministry of Forests, 1997a, Visual impacts of Partial Cutting, Forest Practices Branch and Forest Development Section, Victoria, BC.  139 BC Ministry of Forests, 1997b, Visual Landscape Inventory Procedures and Standard Manual, Forest Practices Branch and Forest Recreation, Victoria, BC. BC Ministry of Forests, 1998, Procedures for factoring visual resources into timber supply analyses, Forest Practices Branch, Victoria, B.C. BC Ministry of Forests, 2001a, Bear Lake Integrated Visual Design Plan, Penticton Forest District Forest Practices Branch, Victoria, BC. BC Ministry of Forests, 2001b, Visual Impact Assessment Guidebook, Forest Practices Branch, Victoria, BC. BC Ministry of Forests, 2002, Predicting the Visual impacts of Retention Cutting, in: Forest Practices Branch, Forest Practices Branch, Victoria, BC. BC Ministry of Forests, 2003, Economic Benefits of Managing Forestry and Tourism at Nimmo Bay: a public perception study and economic analysis., Forest Practices Branch and Economics and Trade Branch, Victoria, BC. BC Ministry of Forests, 2004, Forest and Range Practices Act: Forest Planning and Pracitces Regulation (B. M. o. Forests, ed.), Queen's Printer, Victoria, BC. BC Ministry of Forests and Range, 2010, Visual Impact of Mountain Pine Beetle Attack and Resulting Salvage Operations in British Columbia: a public perception study., Forest Practices Branch, Victoria, BC. BC Ministry of Forests and Range, 2011, 2009/2010 Annual Service Plan Report, Ministry of Forests and Range, Victoria, BC. BC Ministry of Forests, M. a. L., 2010, The State of British Columbia's Forests, Ministry of Forests, Mines and Lands, Victoria, BC. Beese, W., Dunsworth, B., Zielke, K., Bancroft, B., 2003, Maintaining attributes of old- growth forests in coastal BC through variable retention, Forestry Chronicle 79(3):570-578. Beese, W. J., Bryant, A. A., 1999, Effect of alternative silvicultural systems on vegetation and bird communities in coastal montane forests of British Columbia, Canada, Forest Ecology and Management 115(2-3):231-242. Bell, M., Evans, D. M., 1997, Greening" the heart of England"-redemptive science, citizenship, and" symbol of hope for the nation", Environment and Planning D 15:257-280. Bell, S., 1999, Landscape: pattern, perception, and process, Taylor & Francis. Bell, S., 2001, Landscape pattern, perception and visualisation in the visual management of forests, Landscape and Urban Planning 54(1-4):201-211. Bell, S., 2004, Elements of Visual Design in the Landscape, Spon Press (UK). Bergeron, Y., Harvey, B., Ledue, A., Gauthier, S., 1999, Forest management guidelines based on natural disturbance dynamics: stand and forest-level considerations, Forestry Chronicle 75:49-54. Bishop, I., 2003, Assessment of visual qualities, impacts, and behaviours, in the landscape, by using measures of visibility, Environment and Planning B 30(5):677-688. Bishop, I., Wherrett, J., Miller, D., 2000, Using image depth variables as predictors of visual quality, Environment and Planning B 27(6):865-876. Bishop, I. D., Hull, R. B., 1991, Integrating Technologies for Visual Resource-Management, Journal of Environmental Management 32(4):295-312. Bishop, I. D., Hulse, D. W., 1994, Prediction of scenic beauty using mapped data and geographic information systems, Landscape and Urban Planning 30(1-2):59-70.  140 Bliss, J. C., 2000, Public perceptions of clearcutting, Journal of Forestry 98(12):4-9. Boerner, D., 1986, Monongahela revisited, American forests (USA) 92(2). Brown, T. C., Daniel, T. C., 1990, Scaling of Ratings: Concepts and Methods, Rocky Mountain Forest and Range Experiment Station, Forest Service, US Dept. of Agriculture, Fort Collins, CO. Bunnell, F., Dunsworth, G., Huggard, D., Kremsater, L., 2003, Learning to sustain biological diversity on Weyerhaeuser’s coastal tenure, The Forest Project, Weyerhaeuser, Nanaimo, BC. Burrough, P., McDonnell, R., McDonnell, R., 1998, Principles of geographical information systems, Oxford University Press Oxford, Oxford, UK. Caldwell, D., Mineter, M., Dowers, S., Gittings, B., 2003, Analysis and visualization of visibility surfaces, in: Proceedings of the 7th International Conference on GeoComputation, University of Southampton, UK. Cale, W. G., Henebry, G. M., Yeakley, J. A., 1989, Inferring process from pattern in natural communities, Bioscience 39(9):600-605. Carlson, A. A., 1977, On the possibility of quantifying scenic beauty, Landscape Planning 4(2):131–172. Chamberlain, B. C., Meitner, M. J., 2009, Automating the visual resource management and harvest design process, Landscape and Urban Planning 90(1-2):86-94. Cloke, P., Milbourne, P., Thomas, C., 1996, From wasteland to wonderland: opencast mining, regeneration and the English National Forest, Geoforum 27(2):159-174. Corripio, J. G., 2003, Vectorial algebra algorithms for calculating terrain parameters from DEMs and solar radiation modelling in mountainous terrain, International Journal of Geographical Information Science 17(1):1-24. Cronbach, L. J., 1951, Coefficient alpha and the internal structure of tests, Psychometrika 16(3):297-334. Cronbach, L. J., Shavelson, R. J., 2004, My current thoughts on coefficient alpha and successor procedures, Educational and Psychological Measurement 64(3):391. Crow, T. R., Perera, A. H., 2004, Emulating natural landscape disturbance in forest management–an introduction, Landscape Ecology 19(3):231-233. Crowe, S., Johnston, D., Cripps, J., Boyce, S., Richardson, S., 1975, Forests in Relation to Landscape and Amenity [and Discussion], Philosophical Transactions of the Royal Society of London. B, Biological Sciences 271(911):199. Crowe, S., Mitchell, M., 1988, The pattern of landscape, Packard Pub. Cumming, S., Vervier, P., 2002, Statistical models of landscape pattern metrics, with applications to regional scale dynamic forest simulations, Landscape Ecology 17(5):433-444. Dakin, S., 2003, There's more to landscape than meets the eye: towards inclusive landscape assessment in resource and environmental management, Canadian Geographer/Le Géographe canadien 47(2):185-200. Daniel, T. C., 1990, Measuring the quality of the natural environment: A psychophysical approach, American Psychologist 45(5):633-637. Daniel, T. C., 2001, Whither scenic beauty? Visual landscape quality assessment in the 21st century, Landscape and Urban Planning 54(1-4):267-281.  141 Daniel, T. C., Boster, R. S., 1976, Measuring Landscape Esthetics: The Scenic Beauty Estimation Method, Rocky Mountain Forest and Range Experiment Station, Ft. Collins, CO. Daniel, T. C., Vining, J., 1983, Methodological issues in the assessment of landscape quality, Human Behaviour and Environment.Advances in Theory and Research (USA) 6(Journal Article):39-84. Davidson, D., Watson, A., Selman, P., 1993, An evaluation of GIS as an aid to the planning of proposed developments in rural areas, Geographical Information Handling: Research and Applications:251–259. Day, H. Y., 1967, Evaluations of subjective complexity, pleasingness and interestingness for a series of random polygons varying in complexity, Attention, Perception, & Psychophysics 2(7):281-286. De Floriani, L., Magillo, P., 2003, Algorithms for visibility computation on terrains: a survey, Environment and Planning B 30(5):709-728. De Floriani, L., Magillo, P., Puppo, E., 1999, Applications of computational geometry to geographic information systems, Handbook of computational geometry:333-388. Devitt, T., 1988, What price beauty, Isthmus. July:7-9. Diaz, N. M., Apostol, D., 1992, Forest landscape analysis and design: a process for developing and implementing land management objectives for landscape patterns. Domingo-Santos, J. M., de Villaran, R. F., Rapp-Arraras, I., de Provens, E. C. P., 2011, The visual exposure in forest and rural landscapes: An algorithm and a GIS tool, Landscape and Urban Planning. Dramstad, W. E., Tveit, M. S., Fjellstad, W., Fry, G. L. A., 2006, Relationships between visual landscape preferences and map-based indicators of landscape structure, Landscape and Urban Planning 78(4):465-474. Ervin, S., Steinitz, C., 2003, Landscape visibility computation: necessary, but not sufficient, Environment and Planning B 30(5):757-766. Evans, W., Kirkpatrick, D., Townsend, G., 2001, Right-triangulated irregular networks, Algorithmica 30(2):264-286. Fahrig, L., 2003, Effects of Habitat Fragmentation on Biodiversity, Annual Review of Ecology, Evolution, and Systematics 34(1):487-515. Fairhurst, K., 2010, Geoptics landscape apparency : a dynamic visual resource indicator and tool for multi-functional landscape planning, in: Forest Resources Management, Facult of Forestry, University of British Columbia, Vancouver, BC, pp. 324. Fisher, P., 1991, First experiments in viewshed uncertainty: the accuracy of the viewshed area, Photogrammetric Engineering and Remote Sensing 57(10):1321-1327. Fisher, P., 1992, First experiments in viewshed uncertainty: simulating fuzzy viewsheds, Photogrammetric Engineering and Remote Sensing 58:345-345. Fisher, P., 1993, Algorithm and implementation uncertainty in viewshed analysis, International Journal of Geographical Information Science 7(4):331-347. Fisher, P., 1995, An exploration of probable viewsheds in landscape planning, Environment and Planning B 22:527-527. Fisher, P., 1996, Extending the applicability of viewsheds in landscape planning, Photogrammetric Engineering and Remote Sensing 62(11):1297-1302.  142 Fisher, P., Farrelly, C., Maddocks, A., Ruggles, C., 1997, Spatial Analysis of Visible Areas from the Bronze Age Cairns of Mull, Journal of Archaeological Science 24(7):581- 592. Food and Agriculture Organization of the United Nations (FAO), 2011, State of World's Forests, Food and Agriculture Organization of the United Nations (FAO) Rome, Italy. Forestry Commission, 1989, Forest landscape design guidelines (F. Commission, ed.), Edinburgh, Scotland. Franklin, J. F., Berg, D. R., Thornburgh, D. A., Tappeiner, J. C., 1997, Alternative silvicultural approaches to timber harvesting: variable retention harvest systems (K. A. Kohn, J. F. Franklin, eds.), Island Press, pp. 111-139. Franklin, J. F., Spies, T. A., Pelt, R. V., Carey, A. B., Thornburgh, D. A., Berg, D. R., Lindenmayer, D. B., Harmon, M. E., Keeton, W. S., Shaw, D. C., 2002, Disturbances and structural development of natural forest ecosystems with silvicultural implications, using Douglas-fir forests as an example, Forest Ecology and Management 155(1-3):399-423. Franklin, W., Ray, C., 1994, Higher isn’t necessarily better: Visibility algorithms and experiments, in: Advances in GIS: Proceedings of the 6th International Symposium on Spatial Data Handling (T. Waugh, R. Healey, eds.), Citeseer, pp. 751-770. Franklin, W., Ray, C., Mehta, S., 1994, Geometric algorithms for siting of air defense missile batteries, in: Technical Report DAAL03-86-D-0001, Battelle, Columbus Division, Columbus, OH, pp. 199. Franklin, W. R., Vogt, C., 2004, Efficient multiple observer siting on large terrain cells, in: GIScience 2004 Third International Conference on Geographic Information Science, University of Maryland Conference Center. Germino, M. J., Reiners, W. A., Blasko, B. J., McLeod, D., Bastian, C. T., 2001, Estimating visual properties of Rocky Mountain landscapes using GIS, Landscape and Urban Planning 53(1-4):71-83. Giles, J., R.H., Trani, M. K., 1999, Key elements of landscape pattern measures, Environmental Management 23(4):477-481. Gimblett, H. R., Durnota, B., Itami, R. M., 1996, Spatially-Explicit Autonomous Agents for Modelling Recreation Use in Complex Wilderness Landscapes, Complexity International 3. Giovannini, M., 2008, A primer on the physics of the cosmic microwave background, World Scientific Pub. Co. Inc. Gobster, P., Nassauer, J., Daniel, T., Fry, G., 2007, The shared landscape: what does aesthetics have to do with ecology?, Landscape Ecology 22(7):959-972. Gobster, P. H., 1995, Aldo Leopold’s ecological esthetic: Integrating esthetic and biodiversity values, Journal of Forestry 93(2):6-10. Gobster, P. H., 1999, An ecological aesthetic for forest landscape management, Landscape journal 18:54-64. Goodchild, M. F., Kemp, K. K., 1990, Technical issues in GIS: NCGIA core curriculum, National Center for Geographic Information and Analysis, University of Santa Barbara, Santa Barbara, CA. Grêt-Regamey, A., Bebi, P., Bishop, I., Schmid, W., 2008, Linking GIS-based models to value ecosystem services in an Alpine region, Journal of Environmental Management 89(3):197-208.  143 Grêt-Regamey, A., Bishop, I., Bebi, P., 2007, Predicting the scenic beauty value of mapped landscape changes in a mountainous region through the use of GIS, Environment and Planning B: Planning and Design 34:50-67. Gulinck, H., Múgica, M., de Lucio, J. V., Atauri, J. A., 2001, A framework for comparative landscape analysis and evaluation based on land cover data, with an application in the Madrid region (Spain), Landscape and Urban Planning 55(4):257-270. Hagerhall, C. M., Purcell, T., Taylor, R., 2004, Fractal dimension of landscape silhouette outlines as a predictor of landscape preference, Journal of Environmental Psychology 24(2):247-255. Haines-Young, R., Chopping, M., 1996, Quantifying landscape structure: a review of landscape indices and their application to forested landscapes, Progress in Physical Geography 20(4):418. Halpern, C. B., McKenzie, D., Evans, S. A., Maguire, D. A., 2005, Initial responses of forest understories to varying levels and patterns of green-tree retention, Ecological Applications 15(1):175-195. 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(2-3):171-183. Hansen, A. J., Garman, S. L., Weigand, J. F., Urban, D. L., McComb, W. C., Raphael, M. G., 1995, Alternative silvicultural regimes in the Pacific Northwest: simulations of ecological and economic effects, Ecological Applications 5(3):535-554. Hansmann, R., Hug, S. M., Seeland, K., 2007, Restoration and stress relief through physical activities in forests and parks, Urban Forestry & Urban Greening 6(4):213-225. Harshaw, H. W., Sheppard, S. R. J., Lewis, J. L., 2005, A review and synthesis of social indicators for sustainable forest management, BC Journal of Ecosystems and Management 8(21):17. Hauldren, A., 2008, Using VNS3 (Program Help Manual), 3D Nature, LLC., Evergreen, CO, pp. 1132. Hays, S., 1987, Beauty, health, and permanence: Environmental politics in the United States, 1955-85, New York: Cambridge University Press. Hill, D., Daniel, T. C., 2008, Foundations for an Ecological Aesthetic: Can Information Alter Landscape Preferences?, Society & Natural Resources 21(1):34-49. Hull, R. B., Buhyoff, G. J., 1983, Distance and Scenic Beauty: A Nonmonotonic Relationship, Environment and Behavior 15(1):77-91. Iverson, W., 1985, And that's about the size of it: visual magnitude as a measurement of the physical landscape, Landscape Journal 4(1):14. Iverson, W., 1997, The economics of scenic quality (R. A. Strain, S. R. J. Sheppard, eds.), South Lake Tahoe: Tahoe Regional Planning Agency, pp. 49–50. Izraelevitz, D., 2003, A fast algorithm for approximate viewshed computation, Photogrammetric Engineering and Remote Sensing 69(7):767-774. Jeo, R. M., Sanjayan, M., Sizemore, D., 1999, A conservation area design for the central coast region of British Columbia, Canada, Round River Conservation Studies, Salt Lake City, Utah. Jessel, B., 2006, Elements, characteristics and character-Information functions of landscapes in terms of indicators, Ecological Indicators 6(1):153-167.  144 Kaplan, R., Kaplan, S., Ryan, R. L., 1998, With people in mind: Design and management of everyday nature, Island Pr. Kaplan, S., 1995, The restorative benefits of nature: Toward an integrative framework, Journal of Environmental Psychology 15(3):169-182. Kaplan, S., Kaplan, R., Wendt, J. S., 1972, Rated preference and complexity for natural and urban visual material, Attention, Perception, & Psychophysics 12(4):354-356. Kaplan, S., Talbot, J. F., 1983, Psychological benefits of a wilderness experience, Human Behavior & Environment: Advances in Theory & Research. Kearney, A. R., Bradley, G. A., Petrich, C. H., Kaplan, R., Kaplan, S., Simpson-Colebank, D., 2008, Public perception as support for scenic quality regulation in a nationally treasured landscape, Landscape and Urban Planning 87(2):117-128. Kim, Y. H., Rana, S., Wise, S., 2004, Exploring multiple viewshed analysis using terrain features and optimisation techniques, Computers & Geosciences 30(9-10):1019-1032. Kim, Y. S., Johnson, R. L., 2002, The impact of forests and forest management on neighboring property values, Society & Natural Resources 15(10):887-901. Kimmins, J. P., 1999, Biodiversity, Beauty and the "Beast": Are beautiful forests sustainable, are sustainable forests beautiful, and is "small" always ecologically desirable?, Forestry Chronicle 75(6):955-960. Kimmins, J. P., 2004, Forest Ecology, Benjamin Cummings, San Francisco, CA. Knight, J., 2000, From timber to tourism: recommoditizing the Japanese forest, Development and Change 31(1):341-359. Lange, E., 2002, Visualization in Landscape Architecture and Planning–Where we have been, where we are now and where we might go from here, Trends in GIS and Virtualization in Environmental Planning and Design. Proceedings at Anhalt University of Applied Sciences:8-18. Langford, W. T., Gergel, S. E., Dietterich, T. G., Cohen, W., 2006, Map misclassification can cause large errors in landscape pattern indices: examples from habitat fragmentation, Ecosystems 9(3):474-488. Lewis, J. L., Sheppard, S. R. J., Sutherland, K., 2004, Computer-based visualization of forest management: A primer for resource managers, communities, and educators, BC Journal of Ecosystems and Management 5:5-13. Li, H., Wu, J., 2004, Use and misuse of landscape indices, Landscape Ecology 19(4):389- 399. Lindhagen, A., 1996, An approach to clarifying public preferences about silvicultural systems: a case study concerning group selection and clear-cutting, Scandinavian Journal of Forest Research 11:375–387. Litton, R. B., 1968, Forest Landscape Description and Inventories: A Basis for Land Planning and Design, Pacific Southwest Forest and Range Experiment Station, US Dept. of Agriculture. Litton, R. B., 1974, Visual vulnerability of forest landscapes, Journal of Forestry 72(7):392- 397. Litton, R. B., 1979, Aesthetic Values, in: Forest Resource Management: Decision-Making Principles and Cases, W.B. Saunders Company, Philadelphia, pp. 215-225. Llobera, M., 2003, Extending GIS-based visual analysis: the concept of visualscapes, International Journal of Geographical Information Science 17(1):25-48.  145 Lord, J. M., Norton, D. A., 1990, Scale and the spatial concept of fragmentation, Conservation Biology (Journal Article):197-202. Lucas, O. W. R., 1991, The design of forest landscapes, Oxford University Press. Maertens, H., 1884, Der Optische - Maasstab oder die Theorie und Praxis des Asthetischen Sehens in den bildenden Kunsten., Berlin: Baurath, Wasmuth. Magalhães, M., Magalhães, S., Andrade, M., Lisboa Filho, J., 2007, An efficient algorithm to compute the viewshed on DEM terrains stored in the external memory, in: IX Brazilian Symposium on GeoInformatices, Campos do Jordão, Brazil, pp. 183-194. Maloy, M., Dean, D., 2001, An accuracy assessment of various GIS-based viewshed delineation techniques, Photogrammetric Engineering and Remote Sensing 67(11):1293-1298. Marc, J., 2007, FRPA General Bulletin #10 - Beetle Wood Salvage and Visual Quality, BC Ministry of Forests, Victoria, BC. Marc, J., 2008, Protocol for Visual Quality Effectiveness Evaluation: Procedures and Standards, ,, Victoria, BC. Marc, J., 2010, The Effectiveness of Managing Visual Resources under the Forest Practices Code, BC MInistry of Forests,, Victoria, BC. Marc, J., 2011, Meeting with Jacques Marc, Mike Meitner and Brent Chamberlain to discuss new research in visual resource management (B. Chamberlain, ed.), Victoria, BC. Mather, A. S., 2001, Forests of consumption: postproductivism, postmaterialism, and the postindustrial forest, Environment and Planning C 19(2):249-268. McCool, S. F., Benson, R. E., Ashor, J. L., 1986, How the public perceives the visual effects of timber harvesting: An evaluation of interest group preferences, Environmental Management 10(3):385-391. McFarlane, B. L., Stumpf-Allen, R. C. G., Watson, D. O., 2006, Public perceptions of natural disturbance in Canada's national parks: The case of the mountain pine beetle (Dendroctonus ponderosae Hopkins), Biological Conservation 130(3):340-348. McGarigal, K., Marks, B. J., 1995, FRAGSTATS: Spatial Pattern Analysis Program for Quantifying Landscape Structure, Department of Agriculture, Forest Service, Pacific Northwest Research Station. Meitner, M. J., Gandy, R., D'Eon, R. G., 2005, Human perceptions of forest fragmentation: Implications for natural disturbance management, Forestry Chronicle 81(2):256. Mery, G., Alfaro, R. I., Kanninen, M., Lobovikov, M., 2005, Changing paradigms in forestry: Repercussions for people and nature, in: Forests in the global balance - Changing paradigms (G. Mery, R. I. Alfaro, M. Kanninen, M. Lobovikov, eds.), International Union of Forest Research Organizations: World Forests, Society and Environment, pp. 13-20. Michels, K. M., Zusne, L., 1965, Metrics of visual form, Psychological Bulletin 63(2):74. Mitchell, A., 2001, Growth limitations for conifer regeneration under alternative silvicultural systems in a coastal montane forest in British Columbia, Canada, Forest Ecology and Management 145(1-2):129-136. Mitchell, A., Dunsworth, B., Arnott, J., Koppenaal, R., Benton, R., Goodmanson, G., Brown, T., Sanford, J., 2004, Growth limitations of planted conifers regenerating under Montane Alternative Silviculture Systems (MASS): seven-year results, Forestry Chronicle 80(2):241-250.  146 Möller, B., 2006, Changing wind-power landscapes: regional assessment of visual impact on land use and population in Northern Jutland, Denmark, Applied Energy 83(5):477- 494. Moola, F. M., Martin, D., Wareham, B., Calof, J., Burda, C., Grames, P., 2004, The coastal temperate rainforests of Canada: the need for ecosystem-based management, Biodiversity 5(3):9-15. Muir, J., 1912, The Yosemite, The Century Co., New York. Müller, M., Job, H., 2009, Managing natural disturbance in protected areas: tourists' attitude towards the bark beetle in a German national park, Biological Conservation 142(2):375-383. Newhouse, N., 1990, Implications of Attitude and Behavior Research for Environmental Conservation, Journal of Environmental Education 22(1):p26-32. North, M., Chen, J., Smith, G., Krakowiak, L., Franklin, J., 1996, Initial response of understory plant diversity and overstory tree diameter growth to a green tree retention harvest, Northwest Science 70(1):24-35. North, M. P., Keeton, W. S., 2008, Emulating natural disturbance regimes: an emerging approach for sustainable forest management, Patterns and Processes in Forest Landscapes:341-372. O Sullivan, D., Turner, A., 2001, Visibility graphs and landscape visibility analysis, International Journal of Geographical Information Science 15(3):221-237. Ode, Å., Fry, G., Tveit, M. S., Messager, P., Miller, D., 2009, Indicators of perceived naturalness as drivers of landscape preference, Journal of Environmental Management 90(1):375-383. Ode, Å., Hagerhall, C. M., Sang, N., 2010, Analysing Visual Landscape Complexity: Theory and Application, Landscape Research 35(1):111-131. Ode, Å., Miller, D., 2011, Analysing the relationship between indicators of landscape complexity and preference, Environment and Planning B: Planning and Design 38:24-40. Ode, Å., Tveit, M. S., Fry, G., 2008, Capturing landscape visual character using indicators: touching base with landscape aesthetic theory, Landscape Research 33(1):89-117. Ontario Ministry of Natural Resources, 1987, Timber management guidelines for the protection of tourism values (Ontario Ministry of Natural Resources, ed.), Queen’s Printer for Ontario, Toronto, Ont., pp. 96. Orland, B., Budthimedhee, K., Uusitalo, J., 2001, Considering virtual worlds as representations of landscape realities and as tools for landscape planning, Landscape and Urban Planning 54(1-4):139-148. Palik, B., Mitchell, R. J., Pecot, S., Battaglia, M., Pu, M., 2003, Spatial distribution of overstory retention influences resources and growth of longleaf pine seedlings, Ecological Applications 13(3):674-686. Palik, B. J., Mitchell, R. J., Hiers, J. K., 2002, Modeling silviculture after natural disturbance to sustain biodiversity in the longleaf pine (Pinus palustris) ecosystem: balancing complexity and implementation, Forest Ecology and Management 155(1-3):347-356. Palmer, J. F., 1998, Clearcutting in the White Mountains: Perceptions of Citizens, Opinion Leaders and US Forest Service Employees, New York Center for Forestry Research and Development, Syracuse, NY.  147 Palmer, J. F., 2004, Using spatial metrics to predict scenic perception in a changing landscape: Dennis, Massachusetts, Landscape and Urban Planning 69(2-3):201-218. Palmer, J. F., 2008, The perceived scenic effects of clearcutting in the White Mountains of New Hampshire, USA, Journal of Environmental Management 89(3):167-183. Palmer, J. F., Gobster, P., Kokx, T., 1993, Long term visual effects of alternative clearcutting intensities and patterns, General technical report NE (USA). Palmer, J. F., Shannon, S., Harrilchak, M. A., Gobster, P. H., Kokx, T., 1995, Esthetics of clearcutting alternatives in the White Mountain National Forest, Journal of Forestry 93(5):37-42. Pâquet, J., 1993, Seuils d'acceptabilité de l'impact des coupes à blanc sur la qualité esthétique des paysages forestiers boréaux. Memoire de maitrise presente au programme d'amenagement do territoire et developpement regional, Universite Laval, Sainte-Foy, Québec, pp. 61. Pâquet, J., Belanger, L., 1997, Public acceptability thresholds of clearcutting to maintain visual quality of boreal balsam fir landscapes, Forest Science 43(1):46-55. Parsons, R., 1991, The potential influences of environmental perception on human health, Journal of Environmental Psychology 11(1):1-23. Parsons, R., Daniel, T. C., 2002, Good looking: in defense of scenic landscape aesthetics, Landscape and Urban Planning 60(1):43-56. Parsons, R., Tassinary, L. G., Bontempo, D., Vanman, E. J., 1997, Psychophysiology and judgments of landscape aesthetics: on the trail of awe, fascination and the sublime, ASPRS/ACSM Annual Convention Technical Papers 4:288–301. Picard, P., Sheppard, S. R. J., 2002a, Visual Resource Management in British Columbia - Part 1: The effects of visual resource management on timber availability: a review of case studies and BC legislation, BC Journal of Ecosystems and Management 1(2):177-198. Picard, P., Sheppard, S. R. J., 2002b, Visual Resource Management in British Columbia - Part II: Partial cutting in the frontcountry: A win-win solution for short-term timber availability and aesthetics?, B.C. Journal of Ecosystems and Management 1(2). Rana, S., 2003, Fast approximation of visibility dominance using topographic features as targets and the associated uncertainty, Photogrammetic Engineering and Remote Sensing 69(8):881-888. Ribe, R. G., 1989, The aesthetics of forestry: What has empirical preference research taught us?, Environmental Management 13(1):55-74. Ribe, R. G., 1999, Regeneration Harvests Versus Clearcuts: Public Views of the Acceptability and Aesthetics of Northwest Forest Plan Harvests, Northwest Science 73:102-117. Ribe, R. G., 2005, Aesthetic perceptions of green-tree retention harvests in vista views. The interaction of cut level, retention pattern and harvest shape, Landscape and Urban Planning 73(Journal Article):277-293. Ribe, R. G., 2006, Perceptions of forestry alternatives in the US Pacific Northwest: Information effects and acceptability distribution analysis, Journal of Environmental Psychology 26(2):100-115. Ribe, R. G., 2009, In-stand scenic beauty of variable retention harvests and mature forests in the US Pacific Northwest: The effects of basal area, density, retention pattern and down wood, Journal of Environmental Management 91(1):245-260.  148 Ribe, R. G., Armstrong, E. T., Gobster, P. H., 2002, Scenic vistas and the changing policy landscape: visualizing and testing the role of visual resources in ecosystem management, Landscape Journal 21(1):42. Ribe, R. G., Matteson, M. Y., 2002, Views of Old Forestry and New Among Reference Groups in the Pacific Northwest, Western Journal of Applied Forestry 17(Journal Article):173-182. Riitters, K., Wickham, J., O'Neill, R., Jones, B., Smith, E., 2000, Global-scale patterns of forest fragmentation, Conservation Ecology 4(2):3. Romano, B., 1995, National Parks Policy and Mountain Depopulation: A Case Study in the Abruzzo Region of the Central Apennines, Italy, Mountain Research and Development 15(2):121-132. Rossman, B. B., Ulehla, Z. J., 1977, Psychological reward values associated with wilderness use, Environment and Behavior 9(1):41. Sander, H. A., Manson, S. M., 2007, Heights and locations of artificial structures in viewshed calculation: How close is close enough?, Landscape and Urban Planning 82(4):257- 270. Sang, N., Miller, D., Ode, A., 2008, Landscape metrics and visual topology in the analysis of landscape preference, Environment and Planning B: Planning and Design 35(3):504- 520. Saura, S., 2004, Effects of remote sensor spatial resolution and data aggregation on selected fragmentation indices, Landscape Ecology 19(2):197-209. Sawada, M., Cossette, D., Wellar, B., Kurt, T., 2006, Analysis of the urban/rural broadband divide in Canada: Using GIS in planning terrestrial wireless deployment, Government Information Quarterly 23(3-4):454-479. Schrier, D., 2011, Exports (BC Origin) 2001-2010 (B. Stats, ed.), BC Stats and Statistics Canada, Victoria, BC. Schuetz, R., 1998, An Evaluation of Available Merchantable Volume in Areas Subject to Visual Quality Objectives, Industrial Forestry Service Ltd., Prince George, B.C. Seely, B., Nelson, J., Wells, R., Peter, B., Meitner, M., Anderson, A., Harshaw, H. W., Sheppard, S. R. J., Bunnell, F. L., Kimmins, J. P. H., 2004, The application of a hierarchical, decision-support system to evaluate multi-objective forest management strategies: a case study in northeastern British Columbia, Canada, Forest Ecology and Management 199(2/3):283-305. Seila, A. F., Anderson, L., 1984, Estimation of tree preservation costs on urban residential lots in metropolitan Atlanta, Research Division, Georgia Forestry Commission. Selin, S., Chavez, D., 1995, Developing an evolutionary tourism partnership model, Annals of Tourism Research 22(4):844-856. Seymour, R. S., White, A. S., Demaynadier, P. G., 2002, Natural disturbance regimes in northeastern North America--evaluating silvicultural systems using natural scales and frequencies, Forest Ecology and Management 155(1-3):357-367. Shafer, E. L., 1967, Forest Aesthetics-A Focal Point in Multiple Use Management and Research, pp. 47-71. Shang, H., Bishop, I. D., 2000, Visual Thresholds for Detection, Recognition and Visual Impact in Landscape Settings, Journal of Environmental Psychology 20(2):125.  149 Sheppard, S., Picard, P., 2006, Visual-quality impacts of forest pest activity at the landscape level: A synthesis of published knowledge and research needs, Landscape and Urban Planning 77(4):321-342. Sheppard, S. R. J., 2000, Beyond visual resource management: emerging theories of an ecological aesthetic and visible stewardship, in: Forest and landscapes: Linking ecology, sustainability and aesthetics (S. R. J. Sheppard, H. W. Harshaw, eds.), CABI Publishing, Wallingford, UK, pp. 149–172. Sheppard, S. R. J., Achiam, C., D'Eon, R. G., 2004, Aesthetics: are we neglecting a critical issue in certification for sustainable forest management?, Journal of Forestry 102(5):6-11. Shi, X., Zhu, A., Burt, J., Choi, W., Wang, R., Pei, T., Li, B., Qin, C., 2007, An Experiment Using a Circular Neighborhood to Calculate Slope Gradient from a DEM, Photogrammetric Engineering and Remote Sensing 73(2):143. Smardon, R., Palmer, J., Felleman, J., 1986, Foundations for visual project analysis, Wiley- Interscience. Sorensen, P., Lanter, D., 1993, Two algorithms for determining partial visibility and reducing data structure induced error in viewshed analysis, Photogrammetric Engineering and Remote Sensing 59(7):1149-1160. Sougavinski, S., Doyon, F., 2002, Variable Retention: Research Findings, Trial Implementation and Operational Issues (Synthesis Report). Spies, T. A., Franklin, J. F., Klopsch, M., Denslow, J., 1990, Canopy gaps in Douglas-fir forests of the Cascade Mountains, pp. 649-658. Stenström, S., 1964, Optics and the Eye, Butterworths, Toronto. Taylor, J. G., Zube, E. H., Sell, J. L., 1987, Landscape assessment and perception research methods, in: Methods in environmental and behavioral research (R. B. Betchel, R. Marens, W. Michelson, eds.), Van Nostrand Reinhold, New York, pp. 361-393. Thorsnes, P., 2002, The value of a suburban forest preserve: Estimates from sales of vacant residential building lots, Land Economics 78(3):426. Travis, M., Elsner, G., Iverson, W., Johnson, C., 1975, VIEWIT: computation of seen areas, slope, and aspect for land-use planning, USDA Forest Service Berkeley, California, USA. Turner, M. G., 1989, Landscape ecology: the effect of pattern on process, Annual Review of Ecology and Systematics 20(1):171-197. Turner, M. G., O'Neill, R. V., Gardner, R. H., Milne, B. T., 1989, Effects of changing spatial scale on the analysis of landscape pattern, Landscape Ecology 3(3):153-162. Tveit, M., Ode, Å., Fry, G., 2006, Key concepts in a framework for analysing visual landscape character, Landscape Research 31(3):229-255. Tveit, M. S., 2009, Indicators of visual scale as predictors of landscape preference; a comparison between groups, Journal of Environmental Management 90(9):2882- 2888. Tyrväinen, L., 1997, The amenity value of the urban forest: an application of the hedonic pricing method, Landscape and Urban Planning 37(3-4):211-222. Tyrvainen, L., Gustavsson, R., Konijnendijk, C., Ode, A., 2006, Visualization and landscape laboratories in planning, design and management of urban woodlands, Forest Policy and Economics 8(8):811-823.  150 U. S. Congress, 1960, Multiple-Use Sustained-Yield Act, in: PL 86-517; 74 Stat.  215 (Senate and House of Representatives of the United States of America in Congress, ed.). U. S. Congress, 1969, The National Environmental Policy Act, in: PL 91-190; 83 Stat.  852 (Senate and House of Representatives of the United States of America in Congress, ed.). U. S. Congress, 1974, Forest and Rangeland Renewable Resources Planning Act, in: PL 93- 378; 88 Stat. 476 (Senate and House of Representatives of the United States of America in Congress, ed.). Ulrich, R. S., 1981, Natural versus urban scenes, Environment and Behavior 13(5):523. Ulrich, R. S., 1986, Human responses to vegetation and landscapes, Landscape and Urban Planning 13:29-44. Ulrich, R. S., Simons, R. F., Losito, B. D., Fiorito, E., Miles, M. A., Zelson, M., 1991, Stress recovery during exposure to natural and urban environments, Journal of Environmental Psychology 11(3):201-230. USDA Forest Service, 1974, National Forest Landscape Management. Chapter 1: The Visual Management System, USDA Forest Service, Washington, DC. USDA Forest Service, 1995, Landscape Aesthetics: A Handbook for Scenery Management., USDA Forest Service, Washington, DC. USDI Bureau of Land Management, 1998, Visual Resource Management Policy Restatement Information Bulletin No. 98-135, Washington, D.C. van Kreveld, M., 1996, Variations on sweep algorithms: efficient computation of extended viewsheds and classifications, in: 7th International Symposium on Spatial Data Handling, Citeseer, Delft: Faculty of Geodetic Engineering, Delft University of Technology. Vogt, P., Riitters, K. H., Iwanowski, M., Estreguil, C., Kozak, J., Soille, P., 2007, Mapping landscape corridors, Ecological Indicators 7(2):481-488. Wang, J., Robinson, G., White, K., 1996, A fast solution to local viewshed computation using grid-based digital elevation models, Photogrammetric Engineering and Remote Sensing 62(10):1157-1164. Wang, S., 2004, One hundred faces of sustainable forest management, Forest Policy and Economics 6(3-4):205-213. Wells, R. W., Bunnell, F. L., Haag, D., Sutherland, G., 2003, Evaluating ecological representation within differing planning objectives for the central coast of British Columbia, Canadian Journal of Forest Research 33(11):2141-2150. Wheatley, D., 1995, Cumulative viewshed analysis: a GIS-based method for investigating intervisibility, and its archaeological application, in: Archaeology and geographical information systems: a European perspective (G. Lock, Z. Stancic, eds.), Taylor and Francis, London, pp. 171-185. Williams, G. W., Tolle, T., 2001, USDA Forest Service - Long-time Leader in Federal Forest Planning, Society Amer Foresters, pp. 403. Wilson, J., Lindsey, G., Liu, G., 2008, Viewshed characteristics of urban pedestrian trails, Indianapolis, Indiana, USA, Journal of Maps 108:118. Wisconsin Department of Natural Resources, 2003, Wisconsin Forest Management Guidelines PUB-FR-226 (D. Forestry, ed.), Wisconsin Department of Natural Resources,, Madison, WI.  151 Young, K. A., 2000, Riparian zone management in the Pacific Northwest: who's cutting what?, Environmental Management 26(2):131-144. Zajonc, R. B., 1980, Feeling and thinking: Preferences need no inferences, American Psychologist 35(2):151-175. Zakšek, K., Oštir, K., Kokalj, Ž., 2011, Sky-View Factor as a Relief Visualization Technique, Remote Sensing 3(2):398-415. Zube, E. H., Sell, J. L., Taylor, J. G., 1982, Landscape perception: research, application and theory, Landscape Planning 9(1):1-33.    152 Appendices Appendix A: Survey Questionnaire Below are the questions used in the questionnaire for the survey conducted as part of Chapter 2. The questions were broken down into four separate pages. Here, each page is presented along with the questions separated by a numbering scheme and the possible values given by a bulleted scheme. Page 1 1. In what year were you born? • Year 2. What is your gender? • Male/Female 3. What is the highest level of education you have earned (you can type your answer as well)? • High School • Some College • Technical or trade School • University • Post Graduate 4. What is your area of study? • Open-ended 5. My area of study is affiliated with a design or aesthetic emphasis • Yes/No 6. My area of study is affiliated with an environmental emphasis (forestry, conservation, ecology, etc.)? • Yes/No Page 2 1. Now that you have seen images showing many different harvest design possibilities, what are your thoughts about forest design?  153 • Open-ended Page 3 1. You identified the image to the right (actual harvest image shown) as one of your least preferred harvest designs. Please describe why you (would) rate this image low: • Open-ended 2. You identified the image to the right (actual harvest image shown) as one of your most preferred harvest designs. Please describe why you (would) rate this image high: • Open-ended Page 4 1. In this study you were exposed to a variety of different harvest designs that varied across three different variables: complexity, shape, and “skinniness”. Please rate how important each of these variables is to your preferences for a particular design.  If none of these variables are important to consider when designing a harvest, then rate them all low. If you think all of them are important to consider when designing a harvest, please rate them favourably. If one of the variables is more important than another, please reflect that in your rating as well. The images below illustrate each of the variables considered in this study. In terms of ratings: 1= the lowest importance, and 10 = the highest importance. a. Complexity: • 1 – 10 b. Shape • 1 – 10 c. Skininess • 1 - 10


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items