ASSESSING THE VISUAL QUALITY OF THE MAXTON PLAINS ALVARS  By  Emily Prieskorn                          A THESIS  Submitted to Michigan State University in partial fulfillment of the requirements for the degree of  Environmental Design  Master of Arts  2016       ABSTRACT  ASSESSING THE VISUAL QUALITY OF THE MAXTON PLAINS ALVARS  By  Emily Prieskorn  This study assesses and documents the visual quality of the Maxton Plains alvar/alvar grassland plant communities found on Drummond Island, Michigan, USA.  These small, rare landscape types are not addressed by large-scale visual quality mapping efforts and may differ from predicted scores.  Visual quality was assessed using two evolutions purpose, which was to compare their performance. Equation (1) produced a score set ranging from 52.70-57.17, with an average of 54.32, a variance of 0.72, and a standard deviation of 0.85.  Equation (2) produced a set of scores (hereafter referred to as Set 2) ranging from 47.12-52.67, with an average of 50.55, a variance of 0.84, and a standard deviation of 0.90. The results of the visual quality assessment reveal that the Maxton Plains alvars and alvar grasslands have consistently high-to-moderate visual quality and are visually equivalent.  The results also indicated that there are slight differences between evolutions of the model that could play a role in future studies.  iii  TABLE OF CONTENTS   . v  LIST OF FIGURES vi   1  What is Visual Quality?................................................................................................... 1  Visual Quality Assessment 2 .... 2  3  LITERATURE REVIEW. 5  .. 5   ...5  . 5  The expert approach 5  The perception assessment approach.. 6  ... 6   6  .... 7   8   9   9  Visual quality mapping10  Large-scale mapping efforts 10  . 11  ... 11  Alvars and alvar grasslands. 11  Comparisons and i. 12  . 12  METHODS.. 14  .14  Procedures: Equations . 16  . 20  Equation (1)20  Equation (2).. 20  RESULTS 21  DISCUSSION. 22  Understanding Visual Quality Scores.. 22  .. 22  Expectations and reason 22  Consistency within the plant community.. 23  .. 24  . 24   25  Contributio 26 iv   .. 26  . 27  CONCLUSION.. 28  .. 28  Purpose o.. 28  ............. 28  ... 29  ... 29  .... 29  .... 29  APPENDICES  31  APPENDIX A: DATA AND RESULTS.. 32  APPENDIX B: IMAGES. 40  REFERENCES 70   v  LIST OF TABLES   Table 1: Variables...... 18  Table 2: Smyser Index. 19  Table 3: Results by Location 32  Table 4: Data.... .33  Table 5: Visual Quality ...37  Table 6: Descriptive Statistics by Location.. 39       vi  LIST OF FIGURES   Figure 1: Map of the United States of America, showing Drummond Island, Michigan.  Made using Google Maps  14  Figure 2: Map of Michigan, showing Drummond Island.  Made using Google Maps  15  Figure 3: Map of Drummond Island, showing Maxton Plains. Made using Google Maps (2016). 16  Figure 4: Locations of Points Within Maxton Plains. Made using Google Maps     . 17  Figure 5: Image 5 24  Figure 6: Image 280..25  Figure 7: Image 197 25  Figure 8: Image  40  Figure 9: Image  40  Figure 10: Image . 41  Figure 11: Image  41  Figure 12: Image 15. 42  Figure 13: Image 29 42  Figure 14: Image 33 43  Figure 15: Image 35 43  Figure 16: Image 36 44  Figure 17: Image 50. 44  Figure 18: Image 51.. 45  Figure 19: Image 55 45  Figure 20: Image 58 46 vii  Figure 21: Image 60 46  Figure 22: Image 61 47  Figure 23: Image 66 47  Figure 24: Image 67 48  Figure 25: Image 73 48  Figure 26: Image 74 49  Figure 27: Image 86 49  Figure 28: Image 88 50  Figure 29: Image 100 50  Figure 30: Image 103 51  Figure 31: Image 109 51  Figure 32: Image 110 52  Figure 33: Image 129 52  Figure 34: Image 131 53  Figure 35: Image 135 53  Figure 36: Image 136 54  Figure 37: Image 145 54  Figure 38: Image 146 55  Figure 39: Image 155 55  Figure 40: Image 156 56  Figure 41: Image 163 56  Figure 42: Image 164 57  Figure 43: Image 197 57  viii  Figure 44: Image 199 58  Figure 45: Image 22158  Figure 46: Image 22259  Figure 47: Image 234 59  Figure 48: Image 238 60  Figure 49: Image 242 60  Figure 50: Image 243 61  Figure 51: Image 257 61  Figure 52: Image 259 62  Figure 53: Image 266 62  Figure 54: Image 270 63  Figure 55: Image 272 63  Figure 56: Image 276 64  Figure 57: Image 278 64  Figure 58: Image 280 65  Figure 59: Image 284 65  Figure 60: Image 286 66  Figure 61: Image 290 66  Figure 62: Image 29167  Figure 63: Image 299 67  Figure 64: Image 303 68  Figure 65: Image 306 68  Figure 66: Image 310 69   1  INTRODUCTION  What is Visual Quality?  How do people perceive landscape?  What does good landscape look like?  What makes it good or bad, and who decides which is which?  These questions, and the concepts they represent, are at the heart of professions that seek to change the land for the better.   The assessment of visual quality has traditionally been handled with subjective approaches, but great strides have been made over the last several decades toward quantitative alternatives.  A validated model has been developed that can predict the preference level (used visual quality) that the general public would be expected to express toward a given landscape (Burley, 1997; Burley & Yilmaz, 2014; Kaplan, 1985; Liu & Burley, 2013; Mo, Le Cleach, Sales, Deyoung, & Burley, 2011; Schafer, Hamilton, & Schmidt, 1969; Schafer & Tooby, 1973).   This model has many applications that are currently being explored, one of which is the construction of validated visual quality maps (Burley, Deyoung, Partin & Rokos, 2011; Jin, Burley, Machemer, & Crawford, 2016; Lu, Burley, Crawford, Schutzki, & Loures, 2012; Shafer & Brush, 1977; Yilmaz, Liu, & Burley, 2016).  The opportunity exists to improve current maps by documenting the visual quality of uncommon landscape types that large-scale mapping efforts do not capture (Yilmaz, Liu, & Burley, 2016).     2  Visual Quality Assessment Visual quality assessment is a quantitative approach to a seemingly qualitative problem.  Prior to the development of a quantitative model, many relied on the expert approach to determine visual quality (Daniel, 2001; Ulrich, 1986; Zube, Sell, & Taylor, 1982).  This approach, while useful, lacks the empirical data and statistic validation that the visual quality assessment model offers quality (Daniel, 2001; Ulrich, 1986; Zube, Sell, & Taylor, 1982). The model uses an equation to generate a visual quality score based on the physical attributes of a landscape (Schafer, Hamilton, & Schmidt, 1969).  This score represents the level of preference that the average member of the general public would be expected to express in response to the landscape.  This model, and the equation that comprises it, have evolved considerably over the last several decades, and several versions with varying levels of accuracy exist. The possible applications of the model as a professional tool are numerous.  One such application is the validated mapping of visual quality data, which has recently made great strides in the state of Michigan (Burley, Deyoung, Partin & Rokos, 2011; Jin, Burley, Machemer, & Crawford, 2016; Lu, Burley, Crawford, Schutzki, & Loures, 2012; Yilmaz, Liu, & Burley, 2016).    Visual Quality Mapping: An Opportunity While there has been significant progress in visual quality mapping (Burley, Deyoung, Partin & Rokos, 2011; Jin, Burley, Machemer, & Crawford, 2016; Lu, Burley, Crawford, Schutzki, & Loures, 2012; Shafer & Brush, 1977; Yilmaz, Liu, & Burley,  3  2016), the nature of large-scale mapping efforts creates an opportunity for further research.  The broad landscape type categories and large cell size necessary to large-scale efforts do not capture small or specialized landscape types.  There is very little data about the visual quality of these landscape types (Yilmaz, Liu, & Burley, 2016).  This data would not only contribute to the existing body of work regarding visual quality mapping, but to conservation and awareness efforts.  There is also very little data comparing the results of different versions of the model when applied to the same images.  The equation used by the validated mapping efforts draws on the same set of variables as more recent, more predictive iterations. A comparison of the results produced by different equations could yield useful information about the equations themselves, which could guide the selection of equation for future studies and assess the feasibility of generating more accurate future maps.  Purpose of Study  The purpose of this study is to assess and document the visual quality of ars.  Due to its small size and rarity (Albert, 2006; Albert, Cohen, Kost, & Slaughter, 2008), the alvar plant community does not conform to broad landscape type categories and has not yet been assessed. A secondary purpose of this study is to compare the performance of two equation evolutions in order to gain insight that might guide future studies.  This study will use two different versions of the visual quality assessment model: the older version used by visual quality mapping efforts, and a more current version with higher predictive ability.    4  .  Past work in visual quality has focused on validation and prediction, but the purpose of this study is to describe an area in terms of average visual quality and record the results.        5  LITERATURE REVIEW  Introduction The field of landscape visual quality assessment. Over the past 50 years, people have been interested in discovering what makes one landscape visually preferable to another.  There are vast amounts of literature available on the topic of visual quality; several excellent endeavors have been made to summarize the body of work available (Daniel, 2001; Zube, Sell, & Taylor, 1982).    Dominant Paradigms The expert approach.  Two major approaches dominate recent work in the field of visual quality: the expert approach and the perception assessment approach.  The expert approach is perhaps the more familiar of the two, and relies on the opinion of someone considered an expert to pass judgement based on that expertise.  This approach has a number of applications and is especially useful in the design and planning professions, which predicate on the notion that the designer or planner is more qualified than the average person.  In historic visual quality assessment, however, the d may be largely determined by the person physical sciences, and depending on the method used to collect and evaluate landscape, the information they provide may contain some amount of subjective coloring.  The viewpoints that are represented by expert opinion are limited by the  6  number of experts, which may vary significantly from one situation to the next. Expert opinion is difficult to verify, reproduce, or generalize. For the field of landscape visual quality assessment, the expert approach lacks the levels of validity, precision, and reliability necessary for scientific research (Daniel, 2001; Ulrich, 1986; Zube, Sell, & Taylor, 1982).   The perception assessment approach.  As the topic of environmental and resource management became more relevant, the perception assessment approach gained ground.  This approach focuses on the perceptions of ordinary people, collected and analyzed in objective ways that can be reliably reproduced and verified (Daniel, 2001; (Daniel, 2001; Zube, Sell, & Taylor, 1982).    The perception assessment approach offers empirical alternatives to the expert approach, relying on math rather than opinion.  One tool that has evolved out of this approach is the visual quality assessment model, which contains an equation that predicts public opinion about visual quality based on the physical attributes of a landscape.    The History of the Model Origins: Elwood Schafer.  The visual quality assessment model is a predictive  the perspective of the general public. The model involves both a methodology and an equation that generate a visual quality score.  The first version of this model was produced by Schafer, Hamilton, and Schmidt (1969).The initial equation had only six variables and explained about 66% of the variance in responses, although others claim that the actual figure is about half  7  that (Burley & Yilmaz, 2014). Their methodology, however, provided groundbreaking insight into how a qualitative assessment method could be developed and effectively applied.  A grid system was overlaid onto black and white photographs of landscape (later verified as acceptable substitution for actual landscape). The visual elements present in each landscape were then broken down into 46 variables that could be identified and recorded using the grid system.  Once the variables were recorded, preference data was collected for each image using the Q-sort method and two user groups (campers and laypeople). Lastly, statistical methods were applied to determine which, if any, variables were significant predicators of preference.  This information was then used to develop a predictive equation, which can be found below (Schafer, Hamilton, & Schmidt, 1969). Y= 184.8 - 0.5436 X1  0.09298 X2 + 0.002069 (X1 * X3) + 0.0005538 (X1 * X4)  0.002596 (X3 * X5) + 0.001634 (X2 * X6)  0.008441 (X4 * X6)  0.0004131 (X4 * X5) + 0.0006666 (X1)2 + 0.0001327 (X5)2                                                                            (1)  X1= perimeter of immediate vegetation X2= perimeter of intermediate non-vegetation X3= perimeter of distant vegetation X4= area of intermediate vegetation X5= area of any kind of water X6= area of distant non-vegetation  Schafer went on to apply this equation in a number of subsequent studies (Shafer & Brush, 1977; Schafer & Tooby, 1973). This methodology can still be found in recent work.   largely disregarded in his own time due to a lack of theoretical framework and the prevalence and ease of expert opinion (Burley, 1997; Palmer, 2004).  However, there  8  was still considerable interest in the field, and others continued to work on different aspects of visual quality assessment, generating new possible variables and versions of (Kaplan, 1985).  Aesthetics: a paradigm shift.  The successful development of a significantly more predictive equation required a paradigm shift regarding the nature of preference.  - physical attributes of the landscape, such as water or vegetation, which could be observed and recorded.  But the original that there was more involved in the way people view landscape, and why they prefer one to another.  Burley (1997) credits an Nature's Design: A Practical Guide to Natural Landscaping as the inspiration for a shift away from aesthetics.  The book offers practical advice for creating a beautiful, functional home landscape and is geared toward the average homeowner.  Its value in terms of visual quality assessment lies in the holistic approach to landscape value that it takes.  The book urges the homeowner to consider more than the aesthetic value of the landscape, and offers a fourteen-point  (Smyser, 1982; Burley, 1997).  This checklist became the base for the Smyser index, a scoring system that incorporates environmental, cultural, biological, and economic concerns into one variable of the visual quality assessment equation.  The Smyser index first appears in Burley (1997) as one of 27 total variables.  This 2=66.6%) and relies on the methodological precedents set forth by Schafer, Hamilton, and Schmidt (1969).  The  9  Smyser Index has since been dissected and re-evaluated (Liu & Burley, 2013), but the contributions that it made to the understanding of the model remain invaluable.  The present day.  The model has been refined repeatedly through subsequent studies.  The most current version explains 98.45% of the variance in viewer preference with cultural values that are similar to those of the North American respondents it was primarily developed with, which is likely a result of cultural differences the equation does not account for (Mo, Le Cleach, Sales, Deyoung, & Burley, 2011).    Applications   the first step toward the development of a truly useful assessment tool; the next step is to explore how the equation performs in a variety of settings, situations, and purposes.  This process is already in motion; variations of the model have appeared in a number of studies with diverse characteristics.  A community in Massachusetts has used the core concepts of visual quality assessment to track visual landscape change and resident perceptions over 20 years.  The resulting information was used to make vital management decisions (Palmer, 2004). A similar study applied the basic principles of visual quality assessment to rural landscapes to understand how their community viewed not only the land, but some of the visually significant agricultural techniques in use (Arriaza, Canas-Ortega, Canas-Madueno, & Ruiz-Aviles, 2004). The validity and  10  possible uses of remotely sensed data, such as ArcGIS data, has been verified (Crawford, 1994), providing even more tools to expand the field of visual quality.   Visual quality mapping.  The application that is most relevant to this study is the mapping of visual quality. The concept of visual quality mapping is nearly as old as the visual quality assessment model (Schafer & Brush, 1977), and involves the generation of visual quality scores based on general land use types.  The scores are represented graphically and validated statistically through sampling.  The resulting visual quality maps have myriad potential uses.  Changes can be tracked in space as well as time, as demonstrated by Jin, Burley, Machemer and Crawford (2016) in their Partin and Rokos (2011) demonstrated the planning applications of visual quality maps in a comparison of future and former Detroit to Frank Lloyd WCity.  Large-scale mapping efforts.  Recently, great strides have been made toward large-scale mapping.  Validated visual quality maps have been generated for areas of various sizes, such as watersheds (Lu, Burley, Crawford, Schutzki, & Loures, 2012) and cities (Burley, Deyoung, Partin, & Rokos, 2011; Jin, Burley, Machemer, & Crawford, 2016).  A validated visual quality map of the entire state of Michigan has been produced, confirming the feasibility of mapping efforts on a very large scale as well as the suitability of land uses as markers of visual quality (Yilmaz, Liu, & Burley, 2016).     11  Opportunities Visual quality mapping.  Despite the extensive body of existing work, there are abundant opportunities for further contribution.  The application of the visual quality model to an entire state is certainly impressive and useful.   However, there are inherent limitations to large-scale studies.  Studies that rely on existing cover type maps are limited by their level of detail, range of classification, and original purpose, but the production of new cover type maps is often beyond the feasible limits of a study.  Large cell sizes and broad, general landscape type classifications are not suited to capturing small, unusual landscapes, and may exclude them entirely out of necessity.  Yilmaz, Liu, and Burley (2016) note these limitations in their study alongside a call for detailed assessment of these landscape types in order to create a truly comprehensive visual quality prediction map. Alvars and alvar grasslands.  There are many different landscape types not covered by large-scale efforts.  This study focuses on the alvar plant community and its sub-community, the alvar grassland. A plant community is a type of ecosystem characterized by its dominant plants (Cohen, Kost, Slaughter, & Albert, 2014).   Alvars occur on a thin (<10 inches) layer of soil perched over flat, calcareous bedrock, such as limestone or dolomite.  Glacial action during the Ice Ages removed significant portions of existing topsoil from this soft bedrock, depositing foreign rocks and occasionally scoring deeper gouges (called grykes) directly into the bedrock.  This resulted in bare patches of exposed rock and a very thin soil profile, which is characteristic of the community.  Alvars typically occur near water sources and experience significant seasonal disturbances, such as flooding, drought, scouring winds, and occasionally fire.   12  Due to the high soil pH, punishing disturbance regimes and thin soil, very few woody plants are able to survive.  These harsh conditions create a unique plant community found in only three regions worldwide:  Northwest Ireland, the Baltic region, and the Great Lakes region.   An alvar grassland is a subcategory of alvar characterized by slightly larger soil margins and predominately gramanoid (grasses and sedges) cover as opposed to bare patches of ground or exposed substrate (Albert, 2006; Albert, Cohen, Kost, & Slaughter, 2008).  alvar grasslands unless otherwise specified. Comparisons and insights.  Another opportunity for contribution can be found most predictive model has not rendered its predecessors obsolete.  Many current studies choose to use an older version, which may be due in part to the cumbersome nature of the 99-accuracy that the final equation offers, while others may find the length and difficulty prohibitive.  However, several older versions exist, and there is little to no research available comparing older models to one another.    Conclusions The field of visual quality assessment has developed substantially in recent decades.  The evolution of a highly predictive model has produced several less  13  predictive but still useful versions of the central equation, many of which are still in use today.  The applications of this model are vast and growing as new uses are realized and explored.  One of these applications, visual quality mapping, is making considerable progress in increasingly large-scale endeavors at the expense of examining small, rare landscape types. This study will explore the visual quality of an alvar plant community, a landscape cover type found in very small, specific areas of Michigan.  The visual quality of the alvar will be assessed using two different versions of the equation, which will then be compared to one another as well as to the validated visual quality map of Michigan. This comparison could yield valuable information about the equations themselves that could influence their future use.    14   METHODS  Area of Study Drummond Peninsula, near the Canadian-American border.  The island is accessible by ferry or boat.    Figure 1: Map of the United States of America, showing Drummond Island, Michigan.  Made using Google Maps (2016).  15   Figure 2: Map of Michigan, showing Drummond Island.  Made using Google Maps (2016).  The northern part of the island is home to Maxton Plains, part of The Nature Conservancy. Maxton Plains contains up to 8 square miles of scattered alvar and alvar grassland, and is accessible by car or bicycle (Bailey, 2009).    16   Figure 3: Map of Drummond Island, showing Maxton Plains. Made using Google Maps (2016).   Procedures: Equations and Variables A set of 60 images, gathered from 31 points, were collected for analysis.  All images were gathered on July 31, 2015.  Images can be found in Appendix B.  These points are spread out over five locations as depicted in Map B.  For the purposes of this study, a location consists of a single, contiguous alvar or alvar grassland.  17   Figure 4: Locations of Points Within Maxton Plains. Made using Google Maps (2016).   The images were then analyzed according to the methodology set forth by other studies.  For a complete and exhaustive review of the process, refer to Schafer and Brush (1977) or Burley (1997).  The base set of variables that both equations use can be found below.      18  Table 1: Variables Variable Name Description HEALTH/CVQ Smyser Index Number derived from the application of the Smyser Index (Table 2) X1/V1 Perimeter of Immediate Vegetation Number of boundary edges in which individual leaves, needles, bark, or stems of trees/shrubs are easily distinguishable  X2/V2 Perimeter of Intermediate Non-Vegetation Number of boundary edges in which individual rocks, snow, or patches of bare ground are distinguishable but lack fine detail  X3/V3 Perimeter of Distant Vegetation Number of boundary edges in which vegetation is present but individual trees/shrubs are indistinguishable  X4/V4 Area of Intermediate Vegetation Number of squares in which the outlines of individual trees/shrubs are recognizable, but lack fine detail  X5/V5 Area of Any Kind of Water Number of squares containing any kind of water  X6/V6 Area of Distant Non-Vegetation Number of squares in which rocks, bare soil, and snow occur but lack recognizable individual detail  X7/V7 Area of Pavement Number of squares containing man-made pavement X8/V8 Area of Buildings Number of squares containing man-made structures X9/V9 Area of Vehicles Number of squares containing any kind of man-made vehicle X10/V10 Area of Humans Number of squares containing humans X11/V11 Area of Smoke Number of squares containing smoke X12/V12 Area of Fire Number of squares containing fire X13/V13 Area of Herbaceous Foreground Material Number of squares containing non-woody plant material in the foreground X14/V14 Area of Wildflowers in Foreground Number of squares containing wildflowers in bloom in the foreground X15/V15 Area of Utilities Number of squares containing utility poles, wires, pipes, etc X16/V16 Area of Boats Number of squares containing boats X17/V17 Area of Dead Foreground Vegetation Number of squares containing dead woody vegetation in the foreground X18/V18 Area of Exposed Foreground Substrate Number of squares containing patches of bare ground in the foreground X19/V19 Area of Wildlife Number of squares containing animals, excluding humans X20/V20 Smoothness (Scale 1-5) Uniformity and height of ground texture  X30/V30 Open Landscapes (X2 + X4 + 2(X3 + X6))  X31/V31 Closed Landscapes (X2 + X4 + 2(X1 + X17))  X32/V32 Openness (X30  X31) Amount of space perceivable to viewer  X34/V34 Mystery (X30 * X31 * X7 / 1140) Promise of new but related information  X44/V44 Complexity (Variables X1-X19 squared, them summed) Richness or intricacy; number of different elements  X46/V46 Sum of Variables X1-X19  X51/V51 Wetness (X5/X46)  X52/V52 Noosphericness (X7 + X8 + X9 + X15 + X16) Man-made or otherwise non-natural elements X53/V53 Greenness  X63/V63 Schafer Index (3, 4, 5, 6)  X80/V80 X63 * X52  (Burley, 1997; Burley, Deyoung, Partin, & Rokos, 2011; Kaplan, Kaplan, & Brown, 1989; Schafer, Hamilton, & Schmidt, 1969)                 19  Table 2: Smyser Index A Purifies Air 1 0 -1 B Purifies Water 1 0 -1 C Builds Soil Resources 1 0 -1 D Promotes Human Cultural Diversity 1 0 -1 E Preserves Natural Resources 1 0 -1 F Limits Use of Fossil Fuels 1 0 -1 G Minimizes Radioactive Contamination 1 0 -1 H Promotes Biological Diversity 1 0 -1 I Provides Food 1 0 -1 J Ameliorates Wind 1 0 -1 K Prevents Soil Erosion 1 0 -1 L Provides Shade 1 0 -1 M Presents Pleasant Smells 1 0 -1 N Presents Pleasant Sounds 1 0 -1 O Does not Contribute to Global Warming 1 0 -1 P Contributes to the World Economy 1 0 -1 Q Accommodates Recycling 1 0 -1 R Accommodates Multiple Use 1 0 -1 S Accommodates Low Maintenance 1 0 -1 T Visually Pleasing 1 0 -1 (Burley, 1997).    20  Equations  Equation (1).  The first equation (hereafter referred to as Equation (1)) is found in Burley (1997).  This equation explains about 67% of the variance in responses, and contains 18 terms.  This equation has been utilized in a number of existing visual quality studies, making it a useful baseline for comparisons across studies.   Y= 68.30  1.878 HEALTH + 0.131 X1  0.064 X6 + 0.020 X9 + 0.036 X10 + 0.129 X15  0.129 X19 - .006 X32 + 0.00003 X34 + 0.032 X52 + 0.008 X1X1 + 0.00006 X6X6  0.0003 X15X15 + 0.0002 X19X19  0.0009 X2X14  0.00003 X52X52  0.0000001 X52X34                                     (2)   Equation (2).  The second equation (hereafter referred to as Equation (2)) is found in Burley, Deyoung, Partin, & Rokos (2011), and is an updated version based on Equation (1). It has 19 terms and explains about 75% of the variance in responses.    Y= 58.98827 + 0.07725 V2 + 0.03775 V10  1.18505 CVQ  0.01074 V32 + 0.01161 V52  0.00181 V1V2  0.00026 V1V5 + 0.00134 V1V10  0.00071 V2V14 + 0.00018 V5V9  0.00092 V7V18 + 0.00025 V8V14 + 0.00425 V8V15 + 0.00023 V15V18  0.00012 V2V32 + 0.000000613388 V6V34  0.000000783802 V8V34 + 0.00117 V11V52                             (3)  21  RESULTS  Equation (1) produced a set of scores (hereafter referred to as Set 1) ranging from 52.70-57.17, with an average of 54.32, a variance of 0.72, and a standard deviation of 0.85.  Equation (2) produced a set of scores (hereafter referred to as Set 2) ranging from 47.12-52.67, with an average of 50.55, a variance of 0.84, and a standard deviation of 0.90.       22  DISCUSSION  Understanding Visual Quality Scores  Overview.  The Maxton Plains alvars produced two sets of consistently moderate to low scores that reflect the anticipated levels of visual quality and uniformity.  While the score sets are numerically close, there are differences between them that allow for meaningful comparison and offer insight into each equation. Expectations and reasoning.  The expectations of this study are based on the application of key concepts found in previous visual quality studies.  These concepts are particularly important to the interpretation and comprehension of visual quality data.  The first and most important is that visual quality scores are inversely related to the level of preference they represent.  High visual quality scores indicate low levels of preference, while low scores indicate high levels of preference.  Scores of 30 or lower are highly preferred, 50-60 moderately preferred, 70 less preferred, and scores of 100+ are not preferred (Burley, 2006).  It is also helpful to consider the normative theories developed by Burley (1997), which offer insights into why certain landscapes produce the scores that they do.  The first is the biospheric preference theory, which states that people tend to prefer natural, nonhuman landscapes that include elements such as vegetation, water, and sky.  These landscapes produce mid-to-low scores.  Conversely, human or built (noospheric) elements, such as roads, cars, boats, or other humans, tend to lower visual quality and produce higher scores.  Finally, temporary natural elements, such as wildflowers or wildlife, raise visual quality and produce lower scores  23  (Burley, 1997).   Alvars are plant communities; by definition, they contain almost no noospheric elements to lower their visual quality, and may even contain quality boosters in the form of wildflowers or wildlife. Therefore, an alvar would be expected to produce neutral to low scores and high visual quality, which is in line with both sets of average results.    Consistency within the plant community.  Both score sets also have low standard deviations, which indicate the high levels of visual consistency that are expected in a visually unified plant community.  This corroborates the findings of Lu, Burley, Carwford, Schutzki, and Loures (2012), who concluded that landscape type is predictor of visual quality, and therefore an appropriate way to generate visual quality maps.  This consistency is also present within each location- even the least consistent location (Location 3) still has a very low standard deviation.    Location 3 also contains the highest score for both data sets and the lowest score of Set 2, resulting in the highest ranges of any location and its slightly elevated standard deviation.  Interestingly, there is not a most consistent location.  All five locations produced very close averages, indicating that the locations are visually similar to one another.  This indicates that, at least to the general public, an alvar grassland is visually interchangeable with a true alvar. Initially, this conclusion may sound contradictory to Lu, Burley, Carwford, Schutzki, and Loures (2012) findings regarding land use categories.  However, it is important to remember that there are major differences in scale and purpose between a plant community and a land use category.  A plant community is delineated according to the plants it contains and how they relate to one another.  The distinctions that necessitate the creation of different plant communities may have little impact on their  24  overall visual character.  Land use categories are determined by much broader criteria for a variety of purposes.  The distinction between an alvar and an alvar grassland may be of great use and interest to professionals in the natural sciences, but it is too minute to be visually significant.  Comparing Equations Score sets.  A comparison of the score sets to one another yields some  higher lower and were produced by a newer, more predictive version of Equation (1).  The differences in score sets per image are relatively consistent (Table 5).  Each image does not necessarily occupy the same relative position within both sets: Image 103 generates the maximum value for both sets, but Image 280   25   Figure 5: Image 5    Figure 6: Image 280                                           Figure 7: Image 197   Interpretation.  These results suggest that Equation (1) and Equation (2) are relatively equivalent in terms of performance.  Equation (2) generated slightly lower scores than Equation (1).  Further comparison of the equations to one another reveals some interesting differences.   Both equations draw from the same list of measured variables, but not all of the variables make it into both equations.  For instance,  26  Equation (2) relies heavily on the variable X2 (perimeter of intermediate non-vegetation,) which appears once by itself and three times as part of a larger term.  However, this variable only appears once in Equation (1), and is multiplied by X14 (area of wildflowers in foreground,) a variable that is frequently 0.  In fact, out of 60 images, only one image has nonzero integers for both variables.  Altering this value would have significant effects on Equation (2), but almost no effect on Equation (1).  Some variables only appear in one equation, such as X11 (area of smoke) and X16 (area of boats.) , but could play more significant roles in other landscape types.    Contributions  Visual quality mapping.  Our findings contribute to the knowledge base of visual quality mapping efforts by assessing the visual quality of a previously undocumented landscape type that falls outside the purview of previous studies. While the land use classifications mapping studies utilize are far less specific than the plant communities this study focused on, their results provide some context.  Yilmaz, Liu, and Burley (2016) found that the average scores of highly noospheric land uses were higher (92 Industrial, 74 Downtown, 68 Commercial, 62 Residential and Farmland) than average scores for more biospheric land uses (55 Forested/Woodland, 57 Water, 58 Savanna, 62 Grassland Dunes.)  The Maxton Plains alvars produced an average score that is slightly lower than any of these biospheric land use categories, but the difference is very small. This indicates that the visual quality of the study area is slightly higher  27  than other biospheric categories.  The difference in scores is not surprising when the grain and scale of the studies are considered, but it does support Yilmaz, Liu and Burley (2016)that there is value in the continued assessment of smaller, rarer landscape types.  Conservation.  These findings may also be of interest to the conservation efforts in part to its lack of noospheric elements when viewed from the central road.  The quality, and an understanding of this may help conservation efforts plan to retain them.  This insight may have value to conservation efforts taking place in other plant communities as well.    28  CONCLUSION   - -54.32. 50.55     29  Limitations of Study While this study came about as a response to the limitations of large-scale mapping endeavors, it invariably has limitations of its own.   Some of these limitations result from the design of this specific study, while others are inherent to the nature of the model. Generalization and applicability. The average visual quality scores generated herein can only be confidently applied to the Drummond Island alvars, and have not yet been replicated and corroborated.  Since this study was only interested in the visual quality of the plant community, special care was taken to avoid noospheric elements and other plant communities.  It would therefore be inappropriate to apply these findings to all of Maxton Plains or Drummond Island.  The process of image collection requires larger alvars that allow for a full image; smaller alvars were excluded out of necessity.  Very little data is available regarding the impact of season or time of day on visual quality; for the purposes of this study, it is assumed to be negligible in order to allow comparison between this and other studies.   Further Studies   Visual quality.  The field of visual quality is growing and developing rapidly as future work.  Michigan is home to many different rare landscape types that could be examined to contribute toward the creation of a truly comprehensive map. While this study documents one example of one such landscape, future studies are needed to  30  increase sample size and corroborate our findings in other alvars.  The impacts of season, weather, and time of day have not yet been studied, but could be relevant in areas with significantly different seasonal views.       quality assessment models.  While this study was able to make some general conclusions, a more extensive study could yield further information.  There are numerous evolutions of the model that have not yet been examined.       31   APPENDICES  32  APPENDIX A: DATA AND RESULTS Table 3: Results by Location Location Points Images Results Equation 1  Equation 2 1 5 9 Highest Score 54.54 51.08 Lowest Score 53.10 48.70 Range 1.44 2.38 Average 54.13 50.56 Variance 0.28 0.67 Standard Deviation 0.53 0.82 2 6 12 Highest Score 54.44 51.06 Lowest Score 53.16 49.40 Range 1.28 1.66 Average 54.10 50.58 Variance 0.26 0.40 Standard Deviation 0.51 0.64 3 8 16 Highest Score 57.17 52.67 Lowest Score 53.18 47.14 Range 3.99 5.53 Average 54.74 50.59 Variance 1.31 1.46 Standard Deviation 1.14 1.21 4 5 10 Highest Score 55.22 51.29 Lowest Score 53.19 49.03 Range 2.03 2.27 Average 54.08 50.32 Variance 0.47 0.78 Standard Deviation 0.68 0.89 5 6 12 Highest Score 55.84 51.79 Lowest Score 52.70 48.58 Range 3.13 3.20 Average 54.30 50.65 Variance 0.74 0.71 Standard Deviation 0.86 0.84    33  Table 4: Data   Location 1 Location 2 Variables Point 2 Point 3 Point 4 Point 7 Point 8 Point 11 Point 12 Point 13 Point 14 Point 15  Point 17 Name Description 5 8 10 12 15 29 33 35 36 50 51 55 58 60 61 66 67 73 74 86 88 HEALTH Environmental Quality Index 7 7 8 7 7 7 8 7 7 8 8 7 7 7 7 7 7 7 7 7 8 X1 Perimeter of Immediate Vegetation 100 100 96 96 96 96 96 96 96 96 96 96 96 96 96 98 96 96 96 96 96 X2 Perimeter of Intermediate Non-Vegetation 0 0 0 0 0 0 6 0 0 0 0 0 0 6 0 0 0 0 0 0 0 X3 Perimeter of Distant Vegetation 82 94 88 88 82 86 106 82 80 96 82 82 83 82 79 84 82 85 86 82 101 X4 Area of Intermediate Vegetation 345 190 129 197 238 228 112 126 114 190 176 177 142 188 173 114 107 144 131 152 100 X6 Area of Distant Non-Vegetation 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 X7 Area of Pavement 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 X8 Area of Buildings 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 X9 Area of Vehicles 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 X10 Area of Humans 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 X11 Area of Smoke 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 X14 Area of Wildflowers in Foreground 0 0 0 0 0 0 0 1 0 0 2 0 0 0 0 0 0 2 0 0 0 X15 Area of Utilities 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 X16 Area of Boats 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 X17 Area of Dead Foreground Vegetation 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 X18 Area of Exposed Foreground Substrate 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 X19 Area of Wildlife 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 X30 X2 + X4 + 2(X3+X6) 1163 1330 1371 1393 1298 1340 1354 1420 1474 1210 1374 1393 1448 1398 1399 1456 1463 1414 1385 1424 1138 X31 X2 + X4 + 2(X1 + X17) 545 390 321 389 430 420 310 318 306 382 368 369 334 386 365 310 299 336 323 344 292 X32 X30-X31 618 940 1050 1004 868 920 1044 1102 1168 828 1006 1024 1114 1012 1034 1146 1164 1078 1062 1080 846 X34 (X30 * X1*X7)/1140 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 X52 X7+X8+X9+X15+X16 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0      34   Location 3 Point 20 Point 21 Point 25 Point 26 Point 27 Point 29 Point 30 Point 35 100 103 109 110 129 131 135 136 145 146 155 156 163 164 197 199 8 7 7 7 7 7 7 7 8 7 7 7 7 7 8 7 96 132 96 112 113 112 136 116 100 96 96 110 96 96 96 96 0 0 0 0 8 7 0 12 0 0 0 0 0 0 24 0 99 82 92 84 80 82 86 84 90 84 102 86 74 86 98 84 131 106 114 152 152 111 228 142 163 184 190 121 125 168 177 92 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 42 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 42 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1253 1450 1408 1360 1454 1346 1376 1420 1333 1332 1398 1393 1501 1398 1147 1398 323 454 306 380 386 342 500 386 363 376 382 341 317 360 393 284 930 996 1102 980 1068 1004 876 1034 970 956 1016 1052 1184 1038 754 1114 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0                          35   Location 4 Point 40 Point 43 Point 44 Point 46 Point 48 221 222 234 238 242 243 257 259 266 270 7 7 7 8 8 8 7 7 8 7 114 107 96 96 96 107 96 96 96 96 0 0 0 0 6 0 0 0 6 0 86 92 91 98 89 96 84 86 91 84 182 211 201 190 197 237 228 228 235 114 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 2 5 0 0 0 3 2 4 10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1302 1313 1337 1188 1203 1179 1332 1294 1153 1382 410 425 393 382 395 451 420 420 433 306 892 888 944 806 808 728 912 874 720 1076 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0                    36    Location 5 Point 49 Point 50 Point 51 Point 52 Point 54 Point 55 272 276 278 280 284 286 290 291 299 303 306 310 7 8 8 8 7 8 6 7 7 7 7 7 121 100 96 96 104 96 106 96 96 96 96 96 0 4 0 0 0 0 6 0 0 0 0 0 88 52 86 139 92 30 80 86 106 84 80 88 114 254 298 169 227 248 203 190 190 266 266 91 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 1 2 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1396 1284 1300 1191 1269 1092 1357 1306 1426 1304 1324 1185 356 458 490 361 435 440 421 382 382 458 458 283 1040 826 810 830 834 652 936 924 1044 846 866 902 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0          37  Table 5: Visual Quality Score Sets Location Point Picture number  Visual Quality Score Set 1 Visual Quality Score Set 2 GPS Coordinate (Point) Difference in Scores 1 2 5 54.54 51.08 46 04 .217 3.46 8 54.39 50.82 83 35 .451 3.57 3 10 53.31 49.68 46 04 .231 3.64 12 54.31 50.86 083 35 .495 3.45 4 15 54.39 50.99 46 04 .235 3.39    83 35 .539  7 29 54.34 50.91 46 04 .168 3.43 33 53.10 48.70 83 35 .599 4.40 8 35 54.39 50.99 46 04 .238 3.39 36 54.41 51.04 83 35 .583 3.37 2 11 50 53.22 49.51 46 04 .341 3.71 51 53.39 49.81 83 36 .729 3.58 12 55 54.39 50.99 46 04 .344 3.39 58 54.37 50.97 83 36 .823 3.40 13 60 54.39 50.43 46 04 .382 3.95 61 54.42 51.06 83 36 .806 3.36 14 66 54.44 50.99 46 04 .403 3.44 67 54.39 50.99 83 36 .862 3.39 15 73 54.35 50.93 46 04 .403 3.42 74 54.34 50.91 83 36 .903 3.43 17 86 54.39 50.99 46 04 .347 3.39 88 53.16 49.40 83 37 .026 3.76 3 20 100 53.18 49.44 46 04 .459 3.74 103 57.17 52.67 83 39 .260 4.50 21 109 54.27 50.78 46 04 .467 3.49 110 55.15 51.34 83 39 .352 3.81 25 129 55.23 50.45 46 04 .517 4.78 131 55.15 50.51 83 39 .492 4.64 26 135 57.00 51.77 46 04 .471 5.24 136 55.37 49.88 83 39 .506 5.49 27 145 53.44 49.72 46 04 .475 3.72 146 54.36 50.95 83 39 .552 3.41 29 155 54.15 50.56 46 04 .528 3.58 156 54.98 51.21 83 39 .717 3.77 30 163 54.48 51.17 46 04 .567 3.32 164 54.34 50.91 83 39 .775 3.43 35 197 53.19 47.14 46 04 .520 6.06 199 54.36 50.95 83 39 .813 3.41 38   4 41 221 55.22 51.29 46 05 .220 3.93 222 54.74 51.02 83 42 .143 3.73 43 234 54.28 50.80 46 05 .181 3.48 238 53.19 49.46 83 42 .103 3.73 44 242 53.30 49.09 46 05 .201 4.21 243 53.70 49.74 83 41 .978 3.95 46 257 54.36 50.95 46 05 .161 3.41 259 54.34 50.91 83 41 .782 3.43 48 266 53.26 49.03 46 05 .145 4.23 270 54.36 50.95 83 41 .593 3.41 5 49 272 55.68 51.40 46 05 .074 4.28 276 53.90 50.17 83 41 .593 3.73 50 278 53.34 49.72 46 05 .072 3.62 280 52.70 48.58 83 41 .175 4.12 51 284 54.59 50.95 46 05 .114 3.64 286 54.01 50.93 83 41 .182 3.09 52 290 55.84 51.79 46 05 .064 4.05 291 54.34 50.91 83 41 .103 3.43 54 299 54.10 50.48 46 05 .062 3.62 303 54.36 50.95 83 40 .964 3.41 55 306 54.41 51.04 46 05 .094 3.37 310 54.31 50.86 83 40 .936 3.45             39  Table 6: Descriptive Statistics by Location  Location 1 2 3 4 5 Number of pictures 9 12 16 10 12 Maximum Set 1 54.54 54.44 57.17 55.22 55.84 Maximum Set 2 51.08 51.06 52.67 51.29 51.79 Minimum Set 1 53.10 48.70 53.18 53.19 52.70 Minimum Set 2 48.70 53.16 47.14 49.03 48.58 Range Set 1 1.44 1.28 3.99 2.03 3.13 Range Set 2 2.38 1.66 5.53 2.27 3.20 Average Set 1 54.13 50.56 54.74 54.08 54.30 Average Set 2 50.56 54.10 50.59 50.32 50.65 Variance Set 1 0.28 0.67 1.31 0.47 0.74 Variance Set 2 0.67 0.26 1.46 0.78 0.71 Standard Deviation Set 1 0.53 0.82 1.14 0.68 0.86 Standard Deviation Set 2 0.82 0.51 1.21 0.89 0.84  40  APPENDIX B: IMAGES  Figure 8: Image 5  Figure 9: Image 8 41   Figure 10: Image 10  Figure 11: Image 12 42   Figure 12: Image 15  Figure 13: Image 29 43   Figure 14: Image 33  Figure 15: Image 35 44   Figure 16: Image 36  Figure 17: Image 50 45   Figure 18: Image 51  Figure 19: Image 55 46   Figure 20: Image 58  Figure 21: Image 60 47   Figure 22: Image 61  Figure 23: Image 66 48   Figure 24: Image 67  Figure 25: Image 73  49   Figure 26: Image 74  Figure 27: Image 86  50   Figure 28: Image 88  Figure 29: Image 100 51   Figure 30: Image 103  Figure 31: Image 109 52   Figure 32: Image 110  Figure 33: Image 129 53   Figure 34: Image 131  Figure 35: Image 135 54   Figure 36: Image 136  Figure 37: Image 145 55   Figure 38: Image 146  Figure 39: Image 155 56   Figure 40: Image 156  Figure 41: Image 163 57   Figure 42: Image 164  Figure 43: Image 197 58   Figure 44: Image 199  Figure 45: Image 221 59   Figure 46: Image 222  Figure 47: Image 234 60   Figure 48: Image 238  Figure 49: Image 242 61   Figure 50: Image 243  Figure 51: Image 257 62   Figure 52: Image 259  Figure 53: Image 266 63   Figure 54: Image 270  Figure 55: Image 272 64   Figure 56: Image 276  Figure 57: Image 278 65   Figure 58: Image 280  Figure 59: Image 284 66   Figure 60: Image 286  Figure 61: Image 290 67   Figure 62: Image 291  Figure 63: Image 299 68   Figure 64: Image 303  Figure 65: Image 306 69   Figure 66: Image 310  70  REFERENCES 71  REFERENCES  Albert, D. (2006). Natural community abstract for alvar. Michigan Natural Features Inventory, Lansing, MI. 10 pp. Arriaza, M., Canas-Ortega, J., Canas-Madueno, J., & Ruiz-Aviles, P. (2004). Assessing the visual quality of rural landscapes. Landscape and Urban Planning, 69(1), 149-161. Bailey, B. (2009). Drummond Island Alvar.  Drummond Island Tourism Association, n.d. Web. 20 Mar. 2016. <http://www.drummondislandchamber.com/index.php?page=Alvar>. Burley, J. (1997). Visual and Ecological Environmental Quality Model for Transportation Planning and Design. Transportation Research Record, 1549, 54-60. Burley, J. (2006).  The science of design: green vegetation and flowering plants do make a difference: quantifying visual quality.  The Michigan Landscape, 49(8), 27-30. Burley, J., DeYoung, G., Partin, S., & Rokos, J. (2011). Reinventing Detroit: reclaiming grayfields- new metrics in evaluating urban environments. Challenges, 2, 45-54. Burley, J., & Yilmaz, R. (2014). Visual quality preference: the Smyser index variables. International Journal of Energy and Environment, 8, 147-253. Cohen, J., Kost, M., Slaughter, B., Albert, D. (2014). A field guide to the natural communities of Michigan. Michigan State University Press. Crawford, D. (1994). Using remotely sensed data in landscape visual quality assessment. Landscape and Urban Planning, 30(1), 71-81. Daniel, T. (2001). Whither scenic beauty? Visual landscape quality assessment in the 21st century. Landscape and Urban Planning, 54(1), 267-281. Jin, Y., Burley, J., Machemer, P., & Crawford, P. (2016). Validating a visual quality prediction map of southern Michigan: assessing environmental quality across 200 years for the Detroit Michigan metropolitan area. Unpublished. Kaplan, R. (1985). The analysis of perception via preference: a strategy for studying how the environment is experienced. Landscape planning, 12(2), 161-176 Kaplan, R., Kaplan, S., & Brown, T., (1989). Environmental preference a comparison of four domains of predictors. Environment and Behavior, 21, 509-530 Liu, C., & Burley, J. (2013). Landscape evaluation criteria: dispersed expectations. Modern Landscape Architecture. Proceedings of the 6th WSEAS 72  International Conference on Landscape Architecture (LAÕ13), Nanjing China, November 17. Lu, D., Burley, J., Crawford, P., Schutzki, R., & Loures, L. (2012). Quantitative methods in environmental and visual quality mapping and assessment: a Muskegon, Michigan watershed case study with urban planning implications. Advances in Spatial Planning, 7, 127-142 Mo, F., Le Cleach, G., Sales, M., Deyoung , G., & Burley, J. (2011). Visual and environmental quality perception ad preference in the people's republic of China, France, and Portugal. International Journal of Energy and Environment, 4(5), 549-557. Palmer, J. (2004). Using spatial metrics to predict scenic perception in a changing landscape: Dennis, Massachusetts. Landscape and Urban Planning, 69(2), 201-218. Shafer, E., & Brush, R. (1977). How to measure preferences for photographs of natural landscapes. Landscape planning, 4, 237-256. Shafer, E., Hamilton, J., & Schmidt, E., (1969). Natural landscape preferences: a predictive model. Journal of Leisure Research, 1(1), 1-19. Schafer, E., & Tooby, M. (1973). Landscape preferences: an internal replication. Journal of Leisure Research, 5, 60-65. Rodale Press. Yilmaz, R., Liu C., & Burley, J. (2016). Validating a visual quality prediction map for Michigan. Unpublished. Zube, E., Sell, J., & Taylor, J. (1982). Landscape perception: research, application and theory. Landscape Planning, 9, 1-33.