VALIDATING A VISUAL QUALITY PREDICTION MAP OF SOUTHERN MICHIGAN By Yuemin Jin A THESIS Submitted to Michigan State University In partial fulfillment of the requirements for the degree of MASTER OF ARTS Environmental Design 2012 ABSTRACT VALIDATING A VISUAL QUALITY PREDICTION MAP OF SOUTHERN MICHIGAN By Yuemin Jin Investigators are seeking methods to predict and assess the visual quality of environments. In this study I employ an equation (Burley, 1997) which predicts visual quality to create a visual quality map of southern Michigan and then text to determine the map’s reliability. Through Kendall’s coefficient of concordance statistical test, I determined that the map is significantly reliable (p<0.01) and conclude that it is possible to construct reliable visual quality maps. Key Words: Environmental Psychology, Landscape Architecture, Land-use Planning, Landscape Planning ACKNOWLEDGEMENT First of all, I want to thank Dr. Jon B. Burley who is my academic advisor and my main committee member for his support and suggestion for my research. He suggested me to the major research direction for my thesis and helped me to collect the data and conducted the statistic work. Secondly, I would like to thank my two committee members: Dr. Machemer and Dr. Crawford who offer me help in my study. Finally, I appreciate to my classmates, Di Lu who offered his help to me collecting the data and giving me suggestion for the statistic process; and Shawn Partin and Dustin Corr who also helped me to collect the data. iii Table of Contents List of Tables…………………………………………………………………………………....vii List of Figures………………………………………………………………………………...vii List of Equations……………………………………………………………………………..xiv Chapter 1…………………………………………………………………………………………1 1.1 Introduction…………………………………………………………………………….........1 1.2 Basic concepts contain in the thesis………………………………........…..........….............1 1.2.1 Landscape…………………………………………………………………………………..1 1.2.2 Landscape Quality………………………………………………………………………....2 1,2.3 Visual Quality……………………………………………………………………………....2 1.2.4 Assessment VS. Valuation…………………………………………...................................2 1.2.5 Systematic Assessment of Visual Quality………………………………………………...4 1.2.6 Challenges of Visual Quality Assessments……………………………………………….6 1.2.6.1 Representing ecosystems………………………………………………………………...6 1.2.6.2 Geo-spatial variations……………………………………………………………………6 1.2.6.3 Temporal variations……………………………………………………………………..7 1.2.6.4 On the horizon: the bio-social chasm ……………………….........................................7 1.3 Literature Review of Visual Quality……………………………………………………......8 1.3.1 Elwood L. Shafer, Jr., the Father of Visual Quality Research………………………....8 1.3.2 The Critical Opinions………………………………………………………………….....11 1.3.3 Deeper Research on Assessing Visual Quality………………………………………….15 1.3.3.1 Assessment based on professional background………………………………………15 iv 1.3.3.2 Researching by new techniques……………………………………………………......17 1.3.4. Other Applications of Visual Quality Assessment……………………………………..19 1.3.4.1 The case of atmospheric environment…………………………………………………21 1.3.4.2 The 1998 Norwegian Monitoring Program…………………………………………...21 1.3.4.3 The case of a universal equation……………………………………………………….22 1.3.4.4 The visual quality study in Other Countries………………………………………….23 1.3.4.5 Case of making use of visual quality assessment……………………………………..25 1.3.4.6 Visual quality assessment at Lower Muskegon Watershed………………………….26 1.3.5 Conclusion………………………………………………………………………………...27 Chapter 2: Study Area and Methodology……………………………………………….…...28 2.1 Purposes of Study………………………………………………………………………......28 2.2 Method………………………………………………………………………………………29 2.2.1 Research Area………………………………………………………………………….....29 2.2.2 Data Collection…………………………………………………………………………....31 2. 3 Analysis Techniques……………………………………………………………………….35 2.4 Sample Selection…………………………………………………………………………….40 Chapter 3: Results…………………………………………………………………………...…44 3.1 Building the predictive map……………………………………………………………..…44 3.2 Validating the Map………………………………………………………………………....46 Chapter 4 Discussion…………………………………………………………………………...50 4.1 The ranking of six types……………………………………………………………………50 4.2 The deviations of the ranking…………………………………………………………..…55 v 4.3 The main points of the research…………………………………………………………59 5. Limitation…………………………………………………………………………………….60 6. Conclusion……………………………………………………………………………………63 Appendix………………………………………………………………………………………...65 Bibliography…………………………………………………………………………………….76 vi List of Tables Table 1.2: Comparison of two approaches of visual quality assessment……………………..5 Table 2.2.2.1: The schedule of collecting data………………………………………………...31 Table 2.2.2.2: Reclassification of the land-cover types………………………………………33 Table 2.3: The environmental health index in the equation.………………………………..37 Table 2.4: The selection of the samples in two sets…………………………………………...43 Table 3.1: The scores of set 1…………………………………………………………………..44 Table 3.2.1: The scores, predictive rankings and real rankings of set 2……………………47 Table 3.2.2: The W value and calculating process……………………………………………48 vii List of Figures Figure 2.2.1: Geographic location of Southern Michigan and the relationship between the Southern Michigan and Lower Muskegon Watershed………………………………………30 Figure 2.2.2.1: Original land-cover map of Southern Michigan…………………………….34 Figure 2.2.2.2: Reclassified land-cover map of Southern Michigan…………………………35 Figure 2.4.1: Location of set 1 of commercial………………………………………………...41 Figure 2.4.2: Location of set 2 of commercial………………………………………………...41 Figure 2.4.3: Location of set 1 of residential………………………………………………….41 Figure 2.4.4: Location of set 2 of residential………………………………………………….41 Figure 2.4.5: Location of set 1 of downtown………………………………………………….41 Figure 2.4.6: Location of set 2 of downtown…………………………………………………41 Figure 2.4.7: Location of set 1 of farmland…………………………………………………..42 Figure 2.4.8: Location of set 2 of farmland………………………………………………….42 Figure 2.4.9: Location of set 1 of industrial…………………………………………………..42 Figure 2.4.10: Location of set 2 of industrial…………………………………………………42 Figure 2.4.11: Location of set 1 of forestry…………………………………………………..42 Figure 2.4.12: Location of set 2 of forestry…………………………………………………..42 Figure 3.1: Predictive visual quality map of Southern Michigan………………………….46 Figure 4.1.1: Sample No.26 of Forested (54.40120) …………………………………………50 viii Figure 4.1.2: Sample No.10 of Residential (61.58117) ………………………………………51 Figure 4.1.3: Sample No.16 of Farmland (59.12320) ……………………………………….52 Figure 4.1.4: Sample No.3 of Commercial (68.00592) ………………………………………53 Figure 4.1.5: Sample No.12 of Downtown (69.28333) ………………………………………53 Figure 4.1.6: Sample No.22 of Industrial (86.56068) ……………………………………….54 Figure 4.2.1: Sample No.53 of Industrial (67.9152) …………………………………….…..55 Figure 4.2.2: Sample No.55 of Industrial (107.80500) ……………………………………..55 Figure 4.2.3: Sample No.44 of Downtown (74.04022) ……………………………………..56 Figure 4.2.4: Sample No.42 of Downtown (89.67798) ………………………………….… 56 Figure 4.2.5: Sample No.34 of Commercial (66.76252) ……………………………………57 Figure 4.2.6: Sample No.31 of Commercial (82.12213) ……………………………………57 Figure 4.2.7: Sample No.50 of Farmland (51.93535) ……………………………………….57 Figure 4.2.8: Sample No.46 of Farmland (66.15512) ……………………………………….57 Figure 4.2.9: Sample No.40 of Residential (61.63773) ……………………………………..58 Figure 4.2.10: Sample No.39 of Residential (73.41253) ……………………………………58 Figure 4.2.11: Sample No.57 of Forested (53.74060) ……………………………………….59 Figure 4.2.12: Sample No.58 of Forested (55.19320) ……………………………………….59 Figure A.1: Set 1 NO.1……………………………………………………………………..…66 ix Figure A.2: Set 1 NO.2……………………………………………………………………..…66 Figure A.3: Set 1 NO.3………………………………………………………………………..66 Figure A.4: Set 1 NO.4……………………………………………………………………..…66 Figure A.5: Set 1 NO.5……………………………………………………………………..…66 Figure A.6: Set 1 NO.6………………………………………………………………………..66 Figure A. 7: Set 1 NO.7……………………………………………………………………….67 Figure A.8: Set 1 NO.8………………………………………………………………………..67 Figure A.9: Set 1 NO.9………………………………………………………………………..67 Figure A.10: Set 1 NO.10……………………………………………………………………..67 Figure A.11: Set 1 NO.11…………………………………………………………………..…67 Figure A.12: Set 1 NO.12…………………………………………………………………..…67 Figure A.13: Set 1 NO.13…………………………………………………………………..…68 Figure A.14: Set 1 NO.14………………………………………………………………..……68 Figure A.15: Set 1 NO.15…………………………………………………………………..…68 Figure A.16: Set 1 NO.16………………………………………………………………..……68 Figure A.17: Set 1 NO.17…………………………………………………………………..…68 Figure A.18: Set 1 NO.18…………………………………………………………………..…68 x Figure A.19: Set 1 NO.19…………………………………………………………………..…69 Figure A.20: Set 1 NO.20…………………………………………………………………..…69 Figure A.21: Set 1 NO.21…………………………………………………………………..…69 Figure A.22: Set 1 NO.22…………………………………………………………………..…69 Figure A.23: Set 1 NO.23………………………………………………………………..……69 Figure A.24: Set 1 NO.24……………………………………………………………………..69 Figure A.25: Set 1 NO.25…………………………………………………………………..…70 Figure A.26: Set 1 NO.26…………………………………………………………………..…70 Figure A.27: Set 1 NO.27…………………………………………………………………..…70 Figure A.28: Set 1 NO.28…………………………………………………………………..…70 Figure A.29: Set1 NO.29……………………………………………………………….……..70 Figure A.30: Set 1 NO.30……………………………………………………………….…..…70 Figure A.31: Set 2 NO.1…………………………………………………………………….…71 Figure A.32: Set 2 NO.2………………………………………………………………….……71 Figure A. 33: Set 2 NO.3………………………………………………………………………71 Figure A.34: Set 2 NO.4………………………………………………………………………71 Figure A.35: Set 2 NO.5………………………………………………………………………71 xi Figure A.36: Set 2 NO.6………………………………………………………………………71 Figure A.37: Set 2 NO.7………………………………………………………………………72 Figure A.38: Set 2 NO.8………………………………………………………………………72 Figure A.39: Set 2 NO.9………………………………………………………………………72 Figure A.40: Set 2 NO.10……………………………………………………………………..72 Figure A.41: Set 2 NO.11………………………………………………………………..……72 Figure A.42: Set 2 NO.12…………………………………………………………………..…72 Figure A.43: Set 2 NO.13…………………………………………………………………...…73 Figure A.44: Set 2 NO.14………………………………………………………………………73 Figure A.45: Set 2 NO.15………………………………………………………………………73 Figure A.46: Set 2 NO.16………………………………………………………………………73 Figure A.47: Set 2 NO.17………………………………………………………………………73 Figure A.48: Set 2NO.18………………………………………………………………………73 Figure A.49: Set 2 NO.19………………………………………………………………………74 Figure A.50: Set 2 NO.20………………………………………………………………………74 Figure A.51: Set 2 NO.21………………………………………………………………………74 Figure A.52: Set 2 NO.22………………………………………………………………………74 xii Figure A.53: Set 2 NO.23………………………………………………………………………74 Figure A.54: Set 2 NO.24………………………………………………………………………74 Figure A.55: Set 2 NO.25………………………………………………………………………75 Figure A.56: Set 2 NO.26………………………………………………………………………75 Figure A. 57: Set 2 NO.27………………………………………………………………………75 Figure A.58: Set 2 NO.28………………………………………………………………………75 Figure A.59: Set2 NO.29………………………………………………………………….……75 Figure A.60: Set 2 NO.30………………………………………………………………………75 xiii List of Equations Equation 1……………………………………………………………………………………..…9 Equation 2………………………………………………………………………………..….…..10 Equation 3………………………………………………………………………………………38 Equation 4………………………………………………………………………………………38 Equation 5………………………………………………………………………………………38 Equation 6………………………………………………………………………………………39 Equation 7………………………………………………………………………………………39 Equation 7………………………………………………………………………………………39 xiv Chapter 1 1.1 Introduction When describing the environment of a certain area, different people have different opinions. These opinions can be summarized as to whether they like the environment or not. Experts have discovered that there are some common preferences by respondents concerning environmental visual quality. Some investigators have studied the responds of citizens. Dr. Jon Burley is one investigator who has studied respondents in North America, including Michigan (Burley et.al. 2011). This thesis will make use of his efforts, to study the visual quality in the Southern Michigan. In this thesis, I will attempt to produce and validate a visual quality map of Southern Michigan. 1.2 Basic concepts contain in the thesis 1.2.1 Landscape The researcher Daniel states that a landscape is: A, picture/view of natural inland scenery; B, the landforms of a region in the aggregate; C, A portion of a territory that the eye can comprehend in as single view (Daniel, 2000, p269). All focus on “a limited area of land surface” and “views/scenes of the land surface”. However, in the view of landscape practitioners and researchers, the meaning of landscape can be extended to include biological and ecological processes (Daniel, 2001, p269). In this thesis, when discussing landscape, I will focus upon work that is based upon the perceptions (aesthetic, ecological, economic, cultural, and functional) the comprise landscape quality. 1 1.2.2 Landscape Quality Landscape quality is much more than the “quality”, which contains the basic needs of human beings, for example, food, water, shelter or recreation opportunities, etc. and the needs in spirit like oneness with nature or sense of a higher power; or original natural values. Instead, landscape quality consists of the concepts such as landscape meaning and sense of place— memories, inferences, and the extraction of symbolic meaning, historical cultural/social significances, ethical/moral obligations and spiritual values. (Daniel, 2001) In another words, exceptional landscape quality is the degree of landscape excellence, comprised of superior characteristics. 1.2.3 Visual Quality When defining the landscape quality of certain areas, observers will often rely on “features of the landscape” which is objective and within the “perceptual/experiential processes”. That means, viewers judge the landscape by the visible features of the landscape as well as the “psychological (perceptual, cognitive and emotional) process” related to the visible features (Brown and Daniel, 1987 and 1990, quoted by Daniel, 2001, p271). Although the viewers sometimes simply regard these visible features as the quality they prefer or not, the visible features in fact represent many other features such as the ecological value, the culture background, the chances of ecology, etc. All the above are sensed by the respondents in the shape of “visual features”. In this research, “visual quality” is used to imply the sense of the people viewing or experiencing landscape. 1.2.4 Assessment VS. Valuation 2 In fact, in order to compare the landscapes, people need to seek a particular criterion or convert them as specific values by offering them a quantitative score. Here I call the process of the above as “assessment” and “valuation”. Assessment usually is a process in which people make a judgment about a person or situation. When facing with the landscape, respondents make their opinions by judging the relative aesthetic excellence of one landscape area as compared to others. So it is a qualitative process. On the other hand, valuation is a judgment about how much something is worth or how useful or effective a particular idea or plan will be act (Shafer, 1969). A certain equation or standard will be used to test the exact quantity of the distance between two objects. Actually, assessment and valuation are always connected with each other. For assessment depends on the unique value or score from the valuation while the evaluative criteria of the valuation is determined by the general assessment. Generally, the distinction between assessment and valuation procedures is clear for most traditional resources. However, it is less clear for landscape aesthetic quality because there is almost no traditional procedure for determining the value of landscape aesthetic quality. Many experts are now devoting themselves to establishing reasonable valuations and testing the validity of them. They are trying to answer the question raised by Elwood L. Shafer Jr. that “What quantitative variables in a landscape are significantly related to public preferences for those landscapes?” (Shafer,1969). It is believed that with the development of the visual quality valuation, the assessment of it will become more accurate. Thus, the topic of this research could be called “visual quality assessment” which in fact making use of the equation to get the exact scores of the images then testing the preference of the different landscape. It is subjective as well as quantitative. 3 1.2.5 Systematic Assessment of Visual Quality Sometimes compared with the most basic survival needs, such as food, water, air and shelter which one must always have for one’s very existence, the visual quality of the landscape might be considered less important. However, when people select the areas to build their houses, rent their apartments, and go for vacation or have trip, the truth is that they are not seeking the places with more survival needs, but finding the areas with more “pleasing views” with means better visual quality. Researchers have tried to develop the systematic assessment of visual quality which would determine not only which landscape condition is aesthetically better, but also how much better than the other ones. Generally, there are two branches of the systematic assessment: expert/design approach and perception-based approach. According to Daniel (Daniel, 2001), if the assessment was based on experts, the classical spatial design features would be stressed while if based on perception, assessment would emphasize on preferences among landscape of human observers. Table 1.2 presents characteristics of both branches. 4 Table 1.2 Comparison of two approaches of visual quality assessment Expert/design approach Perception-based approach Who assessments A single person. rely on The advantage of Environmental the approach practice. Formal feature The selection, grades of landscapes offered by the observers what is made up with a group of human viewers. management The interaction of the landscape quality which agree to human viewers. The biophysical features of the The visual characteristics of landscape, landscape (mountains, lakes, sometimes contains sensory/perceptual, trees, etc.) emotional or cognitive factors. The contents of 1. Rank areas from low to high 1. Introduce derived perceptual factors. quality (visual quality or scenic the assessments 2. Mix emotional responses and the class) relationship between landscapes and 2. Make recommendations human aesthetic responses. about environmental management activities, and visual quality (scenic class) which should be protected in multiple-resource decision contexts. The criticize by Inadequate levels of precision, Without enough agreement by the public reliability and validity. viewers. The limits which The assessments of landscape cause the criticize aesthetic quality are based on single expert, which makes the result unreliable. Because different experts always have different opinions when making assessment. Observers judge the landscape aesthetic quality by their personal understanding and experiences of what visual quality is, which may not be correct. That situation makes the assessments less consensus. For a time, there was a controversy between philosophers, artist/designers and environmental managers/policy makers concerning: the objective (expert based predictive models) and subjective models (perception based). The possible reason for the objectivesubjective controversy is that landscape aesthetic quality assessment depends on both features of 5 the landscape which is objective and the perceptual/experiential processes which is subjective and changes based upon the human viewer. Although many institutions (especially governmental) still use the expert model because it is necessary for them to obtain a professional and fast plan or direction. The subjective model is popular for the investigators due to the fact it represent the general people’s point of view. And that is the approach that my thesis will focus on-- subjective respondent perception. 1.2.6 Challenges of Visual Quality Assessments There are several problems here in methodology, which will be faced with when making the perception-based assessment approach. The reason is due to the fact that the biological value and visual quality of the site are not acknowledged with people at the same time, usually the latter one is newer, and the high ecological quality is not always the same as the high visual landscape aesthetic quality. Thus, as Daniel noted, the relationship between ecological quality and visual quality became a complex problem which relied on the kind of ecosystem contains in it (Daniel, 2001). 1.2.6.1 Representing ecosystems If the visual quality assessment is focused on ecosystem management, what cannot be ignored are the geographic and temporal complexities (Daniel, 2001). The geographic scales range from micro-site to global landscape while the temporal scales range from much shorter to much longer intervals and sometimes become infinite or indefinite. All the above perplex the condition of ecosystem management. 1.2.6.2 Geo-spatial variations 6 On one hand, what human viewers consider as high quality in aesthetic values may not have many geographic variations. On the other hand, the real geo-spatial variations which are important of aesthetic quality should be displayed as two methods. One is the “bird’s eye views” which shows the range of spatial variations, the other is “walk-through” interaction which shows the true levels of landscape conditions (Daniel, 2001). If the observers can have both kinds of “views” to the one site, it will be a great help for the researchers getting the accuracy result. 1.2.6.3 Temporal variations There are temporal changes in ecological quality which makes it difficult for people to determine the relationships between ecological quality and visual quality. This kind of research is being conducted by investigators today. That problem increases the complexity of the assessments. 1.2.6.4 On the horizon: the bio-social chasm There is a huge conflict in visual quality assessments. One group support the bio-centric environmental management, which agrees with the fact that human viewer may have insight and perceptions concerning the quality of landscapes with good ecology (Daniel, 2001). Another group stands on the side of socio-cultural approach that environmental managements should be social problems and human values are more important than natural values. That approach is expressed by the expert/design assessments group (Daniel, 2001). The result of the controversy is that some viewers will change their opinions in order to meet with those in the other group, which makes the assessments less reliable. When examining landscapes, every viewer has his or her own opinions. In other words, when identifying their preference for landscapes, different people will offer different answers. 7 But there is always some agreement in this preference, which means there are always some images which are liked by most of the observers. In order to find the properties where people agree with each other, many experts in science, landscape architecture, forestry, etc pursue research in this field, resulting in the assessment of visual quality. 1.3 Literature Review of Visual Quality As a research in this topic began over fifty years ago, there are many experts who have contributed to the research of visual quality assessment. Their conclusions are at time different with each other. 1.3.1 Elwood L. Shafer, Jr., the Father of Visual Quality Research The first person who should be remembered is Elwood L. Shafer, Jr., as a professor of forestry currently at the New York State College of forestry, who studied perception of the rural exterior environment. The study was based upon citizen perceptions. Before his study, predicting environmental perception was thought to be accomplished with extremely difficult. At that time, he decided to find the relationship between the natural environment, trees, water, mountains, etc., and what people felt about the outdoor natural environment (Shafer, 1969) . Shafer did many surveys. One type relied on the opinions from the group of members who decided to join the investigation. They needed to explain if they like the environment and why they have their viewpoint (Shafer, 1969, p 75). In the survey in 1966, people responded which parts were the most important ones in environment: campsites near water, swimming and water sport facilities, landscape variability surrounding the campground, campground designcampsite spacing, and vegetative screening and tourist attractions nearby. The answer was 8 related to recreation planning issues (Shafer, 1969, p 75). This facilitated Shafer’s ideas about what landscape properties might be important. The approach in 1968 was seeking the relationship between the physical properties of the image and personal behavior in the environment. The resulting predictive equation illustrated the answer (Equation1). Y=3409-[0.0183 (X1+X2)]+[0.157 (X3² +[0.0002(X4· 3² )] X )] (1) Where: Y = Annual total visitor-days per campground. X1 = Area (square feet) of land at the developed swimming beach. X2 = Area (square feet) of water at the developed swimming beach. X3 = Total number of campsites. X4 = Number of islands accessible by outboard motorboat (Shafer, 1969, p76). According to Shafer, the equation above was obviously not complete, because it just contained four characteristics of environment though they were regarded as the important features in environment (Shafer, 1969, p76). However, this experiment indicated that creating equations could be an effective method in assessing environmental quality. With the inclusion, environmental variables such as “mountains, streams, lakes, waterfalls, and various kinds of timber growth supposedly form a heterogeneous environment”, 9 Shafer’s goal shifted from “how do people feel about an environment?” to “how is the total variation of that environment distributed among its various elements” (Shafer, 1969, p76).One solution is to classify each kind of the elements in the environment. In Shafer’s previous study in 1968, he divided the physical features in the environment into 9 different factors: : “Waterwilderness facilities; Swimming area facility; Campground-lake accessibility; White birch predominance; Supplementary lake expansiveness; Campsite picturesqueness; Tourist magnetism; Remoteness; Campsite-lake accessibility” (Shafer, 1969, p77).He thought these 9 factors were important for recreational planning(Shafer, 1969, p78). This is also the beginning of his explanation of the environment by statistic methods. In 1969, the opinions of viewer were surveyed. They were shown 8*10-inch black and white pictures of environments. Thus, “each of 100 landscapes received a preference score(Y)”, and thus revealed the equation (2): Y=184.8-[0.5436X1-0.009298X2] +[0.002069(X1· 3)² X ]+[0.0005538(X1· 4)] X (2) -[0.06(X3· 5)]+[0.001634(X3· 6)]-[0.008441(X4· 6)]-[0.0004131(X4· 5)]+[0.0006666 X X X X X1² ]+[0.0001327 X5² ] Where: Y = Landscape preference. X1 = Perimeter of immediate vegetation-section of the photo where characteristics of individual leaves and bark are easily distinguishable. X2 = Perimeter of intermediate nonvegetation-section of the photo where nonvegetation is visible, but not in fine detail. 10 X3 = Perimeter of distant vegetation-section of photo where shapes of v vegetation cannot be distinguished. X4 = Area of intermediate vegetation-section of photo where vegetation is visible, but not in fine detail. X5 = Area of any kind of water-section of photo that includes water. X6 = Area of distant nonvegetation-section of photo where shapes of nonvegetation cannot be distinguished (Shafer, 1969, p79). Shafer thought, his equation can be widely used when studying the environmental perception. The equation illustrated the relationship between people’s perception of photos of environment and the landscape (Shafer, 1969, p79). It was the first major contribution in visual quality assessment. However, for the reason that his research was made more than 54 years ago, there were many shortcomings in his research which made his results become less accepted. Just as he himself suggested that there were a lot of unconscious characteristics in this perception study, which need to be improved by researchers after him. For example, he suggested: the different perception of pictures in the same place but in different qualities; the changeable view of people who live locally; the influence to the preference of pictures of personal experiences; whether people will change their mind if researcher change the shape of samples, etc. These suggestions brought insight to the other researchers (Shafer, 1969, p80 and 81). Even though Shafer’s study was far from perfect, he made a major advancement in studying visual quality. 11 1.3.2 The Critical Opinions Several experts found aspects in Shafer’s research incomplete. Neil David Weinstein was one of them. He investigated psychological stress and environmental problems. In Weinstein’s article, “The Statistical Prediction of Environmental Preferences: ‘Problems of Validity and Application’”, he pointed out that there was a statistical problem in Shafer’s study which lead to the misunderstanding of the results in it (Weinstein, 1976, p611). This problem was described as “chance associations”. It revealed that there were deviations in Shafer’s results. Moreover, a variance between one-sixth and one-half was caused by the irreproducible connection between the “measured attributes” and the view of the campers who took part in this research. Because of these variables which had no association with each other, the regression analysis was problematic. The solution to the problem could have been identifying the viewers as a regressor in the study. In Shafter’s case, different campers would be needed. According to Weinstein’s opinion, the more data collected, the stronger relationship will be detected (Weinstein, 1976, p611). In Shafer’s time, he did his best to get an accurate result, which was not perfect but leaded other investigators to explore the topic. A.A. Carlson, in his article “On the Possibility of Quantifying Scenic Beauty”, examined philosophical aesthetics. His work was based on two important aspects about the environment, one was that more people know the importance of aesthetic quality, and the other was the demands there were placed the environment (Carlson, 1977, p133). These two aspects seriously influenced the result. 12 In the author’s view, the nature of working on environmental aesthetics should be divided into four separate types of themes, which were “the objectivity, the quantification, the egalitarian, and the formalist” (Carlson, 1977, p134). The objectivity theme referred to the objective achievement when viewing aesthetic quality; the quantification theme meant the aesthetic quality should be developed in a systematic method; the egalitarian themes qualified the viewers who make assessments of the aesthetic quality should own the “general values of the public” (Carlson, 1977, p135); the formalist theme focused on the formal aspects, “shapes, lines, colors and their patterns an combinations”, for example, along with the formal aesthetic qualities, like “balance, proportion, unity and diversity” (Carlson, 1977, p136) . Based on the four themes, aesthetic value of the environment could be evaluated accurately (Carlson, 1977). Carlson’s idea classified the view of the scenic into different aspects. When reviewing Shafer’s study in 1969, Carlson argued that the landscape preference model built by Shafer in his research was not totally correct. He concentrated on three assumptions in Shafer’s work, which reflected the four themes he indicated before. The first assumption was that the certain preferences of landscape were associated with the aesthetic quality in that place. The second assumption was related to the third theme. The assumption was that if the personal preferences agreed with the aesthetic quality, this viewpoint would represent the public view. The third assumption was associated with the forth theme, formalist ideas. He suggested that since Shafer’s work included his methodology and viewpoint were formal, the result could be valuable in only formal aspects of photographs (Carlson 1977, p140 and 141). Moreover, the photographs Shafer used in his study exchanged “‘three-dimensional landscape to something much like the “‘two-dimensional’ photograph.” And they were not considered equal 13 to the real landscape. Carlson believed that the method by Shafer which use quantification to render the quantitative term of landscape quality was too complex (Carlson, 1977, p144). Carlson argued that the first and second assumption which were closely related were not wrong when considered separately. But there would be some problems when they connected to each other because “that is to say it is no more than another case of going to considerable trouble and expense to achieve quantified results, when the same results are already known to us in terms of qualitative assessment.”(Carlson, 1977, p146). The another problem was that not every environment consisted of the same features, such as mountains, water, and what’s more, the study should focus on the aesthetic perceptions instead of nature appreciation (Carlson, 1977,p 147) . It is true that many times the elements of natural structures may not be one hundred percent equal to their aesthetic values. Both authors’ critiques of Shafer’s work offered their opinions about Shafer’s study concerning visual quality. Some of the errors may be due to the limits of the period and working conditions which could not be avoided. Some investigations and critiques really lead to deeper thinking concerning this field of study. Also these critical ideas gave readers more suggestions when judging the assessment of visual quality. Steven C. Bourassa made his critique in 1991. He considered Shafer’s approach as “beating a dead horse” (Bourassa, 1991 p123) which means it was not a new topic for the public. Bourassa defined Shafer’s result of his study as the one “completely without justification” (Bourassa, 1991 p124). The reason was that Shafer did his research which based on people’s preferences were not supported by any valuable theories. In other words, Shafer did not offer any background information of the landscape quality to the observers. The information Bourassa 14 referred to was the “biological mode of aesthetic experience” only by which the “formal approaches” of landscape quality could be assessed truly (Bourassa, 1991 p124). What Bourassa disagreed with Shafer’s research was that Shafer only focused on individual preferences which seemed to lack basic knowledge of the landscape value in ecosystem. That led to incorrect results concerning Shafer’s approach. Bourassa’s critique might pay too much attention one investigator’s ideas. Lacking of information in ecological values may have truly affected the fair judgment of landscape quality by human experiences. However, in Shafer’s study, people’s preferences of the landscapes definitely reflected their perceptions about landscape. Whether Bourassa was right or not, it showed another issue in visual quality assessment. 1.3.3 Deeper Research on Assessing Visual Quality 1.3.3.1 Assessment based on professional background While Shafer’s study was criticized by some researchers, more and more experts did their own research concerning methods and results. One of them named Jonathan G Taylor, wrote “Landscape Assessment and Perception Research Methods” with other co-workers in order to show his insight into the range of visual quality assessment (Taylor et. al. 1987, p361). Taylor chose to assess the landscape quality based on his background (recreation resources) (Taylor et. al. 1987, p361). This approach was somewhat different than Shafer’s. There were four paradigms indicated in this article: “expert, psychophysical, cognitive, and experiential”. Each of them could be used to solve various kinds of problems although each 15 of them may focus on one aspect of visual quality research (Taylor et. al., 1987, p362). All the above are based on the interaction between humans and landscape. The expert paradigm was the principle that the assessment of visual quality should be made by experts, who have more knowledge of a discipline background such as landscape architecture or ecology. (Taylor et. al., 1987,p 362) There were two parts contained in this paradigm. One was that experts ought to be “better judges of landscape aesthetic quality” which was more than others who had only common knowledge of landscape since experts had more concepts on identifying what was beauty as well as what was ugly. The other was that nature and ecosystems had an inherent value of aesthetic quality which should be understood by experts (Taylor et. al., 1987, 363). In other words, experts were the best group of people to make the assessment of landscape. The psychophysical paradigm, according to Taylor, was deriving interpretation through observer’s opinions. Human viewer’s response to certain landscape would be stimulated by the elements of it which agreed with the psychological or especially behaviorism perspectives. The viewers would see images and rate them. Taylor stated: “For example, a chair ‘affords’ sitting down.” (Taylor et. al., 1987, 371) Also, the cognitive paradigm meant that personal information from social condition of the viewers would influence their aesthetic value of the landscape (Taylor et. al., 1987, 363) that different people had different backgrounds and living condition which changed their mind, and what the observers chose to be the landscape with high values were those what were meaningful to them. Knowing this situation, experts could pay more attention on why the landscape was thought of as valuable than others and what kinds of landscape were preferred by viewers. It is even possible for them to predict which group of people with particular backgrounds will highly like or dislike some landscapes. 16 Finally, Taylor suggested the experiential paradigm was where human viewers should be considered as “active participants in landscape”, whose preferences of landscape would be decided by their experiences (Taylor et. al., 1987, 362). That meant, as Taylor suggested: “human qualities such as intentions, needs, knowledge, abilities, and culture, affect judgments.” which made people’s own experiences a part of their preference in the responses. And thus, researchers could make full use of this paradigm to consider more of “the nature of the interaction and its outcomes” instead of wasting too much time on “identifying particular scenic landscape features.” (Taylor et. al., 1987, 382) As a result, the landscape having interaction with viewers should be considered during the assessment. 1.3.3.2 Researching by new techniques Some investigators suggest that making visual quality assessment requires testing the preference by the viewers instead of judging by a group of experts who work on the relative discipline. However, in order to obtain the precise result, they are willing to do work more extensively to obtain information. These investigations not only contain more detail elements in the landscape being considered, but also assist with developing new methods. One of the central ideas is called Q Methodology. It was first used as a theory based on Q sorting technique and then Q Methodology became a research tool (Partin, 2011). Personal assessment was the important part in early uses of Q sort method which played a significant role in Q methodology. As currently popularly used, the Q methodology is a way of “quantifying what would otherwise be considered qualitative data” (Partin, 2011). The Q samples could be images, sound clips, video clips, etc. that represents the research topic and thus help people give their own opinions. 17 The Q sort processing provided a range of responses from least like to most like. In words, they have the sample Q sorted. After that, the results would be compared, averaged, and analyzed. The final result could be emerged then. (Pitt and Zube, 1979, quoted by Partin, 2011). Some other experts believe in modern techniques. The researcher, Doug Crawford, suggested that after getting ideas of the usage of traditional methods such as “manual mapping, extensive fieldwork and teams of trained observers” and modern techniques, for example “geographic information systems”, the assessments of visual quality could be more effective and accurate. (Crawford, 1994, p72) That meant with the development of the technique, the research would be improved. In order to demonstrate his ideas, the author made a study of Cooks River, which would be compared with the previously research on the same place in 1979. And thus he could test the “feasibility of using remotely sensed data in the assessment of landscape visual quality”. (Crawford, 1994, 72) It is a good way to confirm how the new method works for the permanent site. The project was composed of forty landscape units which were defined and classified as “residential”, “open space” or “industrial” according to the features of them The characteristics of the place, which consisted of landform, structures, tree cover, water bodies-extent, water bodies-edge condition, activity, unit outlook, diversity, harmony and contrast were re-assessed in this research with the help of the new technique (Crawford, 1994, p73-p76). After the reassessment, Crawford checked the agreement between the results with those in former survey and found that the disagreement between the two study was due to the fact that the score of the fundamental nature varies from subjectivity to objectivity as well as fundamental errors in the 18 old study (Crawford, 1994, p80). As Crawford even indicated that “it is feasible to use remotely sensed data together with a digital elevation model to delineate homogeneous landscape units and assess their relative visual quality” (Crawford, 1994, p80). That means, as a new method, the geographic information systems did help to support the project, and it could be used when combining with the old data. It offers great help for the investigators. 1.3.4. Other Applications of Visual Quality Assessment Further more, many experts chose their study in areas which had special features. Their study might concentrate on specific aspects because of the character of the areas they picked. However, their studies were still valuable. 1.3.4.1 The case of atmospheric environment Douglas A. Latimer did this research with his coworkers in a wild place in western U.S.A. (Grand Canyon National Park and Mt. Lemmon near Tucson, Arizona) and in the eastern United States (Great Smoky Mountains and Shenandoah national parks). He studied air quality with a team of people. Their goal was to describe the personal preferences of the viewers which were subjective and aesthetic, although they could do not actually alter the pollution levels in this area (Latimer, 1981, p1865). This study was focused on the air quality condition, which was important to the residents in that area. In this study, the first step was building a connection between physical measures of the environment and the perception of air quality as well as aesthetic values that were judged from viewers according to the psychophysical experiment. The following elements were important when judging the visual air quality and aesthetic value: 19 (1) The physical environment (specific features such as hills and mountains, water bodies, vegetation, buildings, and clouds); (2) The atmosphere through which the environment is viewed; (3) The human eye-brain system; (4) The characteristics of the observer; various values, expectations, and attitudes of different people and social groups may affect their perceptions and the significance they ascribe to their perceptions (Latimer, 1981, p1865). The researchers took the colored pictures of landscape with varying light and atmospheric conditions. Assessments were made by a group of viewers from both public side, such as “church, civic and social or business organizations” and individuals from “air pollution consulting firm, air pollution professionals at an annual society meeting, a major oil company, an oil industry group” and “from the Environmental Protection Agency” with high interests in the air quality and professional knowledge of air pollution. They gave their opinions about the photographs which showed the different atmospheric and weather conditions in the place. (Latimer, 1981, p1867) In this research, observers were selected according to their professional background, occupation, interest, position etc.. After collecting the data, an equation was derived illustrating the relationship between human perception and the physical condition. Though the study and result might not be a perfect one, there were several aspects that could be considered as successful at that time. (Latimer, 1981, p1873).Firstly, in this study of air quality, observers, whether with professional background or not, had similar preferences of the visual quality. Moreover, there was a strong relationship between the aesthetic values and the visual air quality. The former was influenced 20 deeply by the latter. What’s more, the further the visual distance was, generally the more preference people would show according to the increase of visual air quality and scenic beauty (Latimer, 1981, p1874). 1.3.4.2 The 1998 Norwegian Monitoring Program, The Norwegian program which was for agricultural landscapes, and just like many other monitoring programs, was based on data collected through aerial photography. The goal of Dramstad and his co-workers was to understand the connection between “the map-based land cover data” and “perceived aesthetic quality” (Dramstad et. al., 2006, p466), and what he used was a study based on the preference through photographs. Preference study was a big part of this program. 53 local experts and 38 students from the Norwegian University of Life Sciences were chosen to be the viewers to participate into the study. They were asked to offer scores from 1 which was least preferred and 5 which was most preferred for 30 photographs. These pictures showed the varied open landscape. After calculating the preference scores, Dramstad and the co-workers found a great positive correlation between preferences and the numbers which present the land types and patches while there was no correlation between the preferences and landscape openness or spatial heterogeneity (Dramstad et. al., 2006, p470). These investigators also mentioned that the preferences might depend on the personal experiences of the viewers, for example, the education of ecology to the students, the fact that those who are familiar with certain areas more would give higher scores for the same pictures than the others, or the scale of the place which might influence the scores. These can not be ignored when having preference study. What Dramstad and other researchers did was also a 21 study and compare of expert views and public ideas. The result showed they had different preferences of landscape which might still a complex problem of that discipline. Nevertheless, he encouraged that more extensive research should be conducted. 1.3.4.3 The case of a universal equation Dr. Jon Bryan Burley illustrated his research investigation within transportation planning and design setting. He employed images from North America (Burley, 1997, p55) His results are used in my thesis study. The elements contained in the pictures were: “and environmental quality index (derived in part from Carol Smyser) buildings, automobiles, boats, people, wildlife, water, roads, fire, smoke, clouds, flowers, vegetation, and nonvegetated substrate across prairie, woodland, wetland, agricultural, urban savanna, and cliff detritus cover types.” (Burely, 1997, p55) They were the typical aspects contained in almost all of the environment. There were two kinds of regressors in his study which stood opposed to each other. The more vehicles, human beings, and utility structures or the other noospheric features exhibited in the place, the more negative the results would be. On the contrary, if the place was full of vegetation, wildlife, openness, flowers, and mountains the results would be more positive, and thus it would earn the better impression of visual quality. Dr. Burley proposed the theory of “biospheric preference theory.” (Burley, 1997, p56) It does make sense that most people will feel pleased with the environment with a lot of trees and open line of sight while little disturbances from vehicles, people or utility structures. Regressors with a “-” symbols, meaning negative indicate preferred visual quality as lower scores indicated preferred visual quality. These kinds of regressors include an 22 environmental quality index perimeter of immediate vegetation, area of distant non-vegetation, areas of wildflowers in foreground and openness. The health index suggests that the more air, water, soil resources and some other natural resources are less polluted as well as the less fossil fuel is used, the more the visual quality will be increased. The second set is opposite the first one. The regressors are with positive coefficients to the “Y” variable, while negative related to the visual quality. They are X9, X10, X15 and X52 which represent the area of vehicles, humans, utilities and noospheric features. According to these coefficients, the more vehicles, humans, utilities or noospheric features occupy in the image, the higher score and the less preference it will get from respondents. In addition, suppose all the elements of the regrossors are zero at the same time, in other words, the image is 100% comprises of natural elements such as the sky with clouds, sun or moon, water, snow etc., the score will be equal to 68.30. That means, a positive visual quality picture should basically has a score lower than 68.30 and a poor rating is above this score. Burley also suggested this study could be useful when planning and designing transportation areas (Burley, 1997). When the designers know which elements are more welcomed by the people, they can assess the general visual quality of their designs. 1.3.4.4 Visual Quality Studies in Other Countries Burley and his coworkers applied their visual quality assessment approach other countries: People's Republic of China, France, and Portugal. This research is valuable because it compared similarities and differences between the people who participated in the American preference study. 23 In this study, they found: In China, there were differences with the perceptions of Westerners (Mo et. al., 2011). For example, Chinese people preferred the images if they were taken in China; if an area had lots of foreground flowers, they would prefer it more than others. Due to the thousands of years of history, the ideal landscape according to Chinese’s philosophy was associated with the beautiful natural environment being in harmony; “implicit, indistinct beauty” and the “blank space” were thought of with unique value on aesthetics in China. In addition, owning to admiring of the wealthy lifestyle, the pictures which described the modern and rich aspects of the life would be preferred by Chinese viewers. As an eastern ancient country, people hold the view of their traditional habits while imagine the life of western and modern, and thus cause the different unique preference of the landscape. In France, people preferred the pictures with these elements: water, mountains which were far away, and buttes and hills while they did not like the pictures with a lot of urban factors including human beings. From the study, it seems that the answer from the French people and North American people are very similar (Mo et al., 2011). It is easy to understand because they share a lot in common, like the lifestyle, the culture, or even the human appearances. In Portugal, similar to the French, human beings had similar preference of the images. The more they liked the pictures, the less people were in them. The same situation happened in the number of vehicles in the image. What was the same as the Chinese was that if the picture had no evidence of utility structures, viewers in Portugal would like it. And they also preferred the ones with less intermediate ground nonvegetation. The Portuguese also gave high preference 24 to the images when they were less open (Mo et al., 2011). Even through Portugal and French are both European countries, there were still some small differences indicated by the study. This study informs us that the region is a significant aspect when assessing visual quality. The more research we do in different areas, the more accurate results and understanding we may consequently obtain. 1.3.4.5 Case of making use of visual quality assessment After talking about the differences between countries, Burley and his investigators did research on design ideas for Detroit (Burley et. al., 2011). This research was about how to assess a design proposal for the Detroit city by using visual quality assessment. The assessment was made with respect to the importance of satisfying economic, ecological, aesthetic and cultural requirements. That is the same view from Frank Lloyd Wright which was called “Broadacre City” (Burley et al., 2011, p46). Detroit, which was used to be a popular industrial city has been in decline and is now being considered for redevelopment. It is a good target for designers. The study addressed the reclamation of “gray-field” (Burley et. al., 2011). A decentralized city should be built by adding different kinds of structures in it. With the success of the study in North America, France and Portugal, Burley and the co-workers used their equation to make the predictive model. Pictures with old vision of Detroit and new digital model imagines were studied with the equation. As a result, the old pictures from Detroit had scored from 81 to 86 which meant they had lower preferences while the digital one for the new Detroit ranged from 55 to 60 which meant people preferred them a lot. Although the “Broadacre City” proposal 25 scorde from 50 to 55, which was even lower than the new Detroit, the new Detroit proposal would a lot better than the composition of Detroit from an earlier time. This study offers a new clue for the planners and designers that before they start the project which is: a visual quality assessment may tell them whether the new idea will please people and even how much pleasure they will get. 1.3.4.6 Visual quality assessment at Lower Muskegon Watershed The Lower Muskegon Watershed is located on the west side of Michigan which contains a lot of inland lakes and rivers. The research of that area is relayed to my thesis. The author Lu took 60 pictures from the 6 counties: Muskegon County, Lake County, Mecosta County, Montcalm County, Newaygo County, and Osceola County. The pictures represented the features of land-cover as: Downtown, Industry, Urban Savanna, Farmland, Water and Forest (Lu, 2011). These pictures were separated as two groups, each group contained 30 pictures. As the mean score of the first group, he obtained: farmland 53.6607, water 53.8357, forest 57.47877 and urban savanna 65.46785. They were all positively preferred landscapes. In the other side, industry with the mean score of 95.16019 and downtown had the score of 77.3249, they were not preferred by people (Lu, 2011). Lu used these scores to build a predictive visual quality map of that area and then validate the map by comparing the real scores and the predictive scores of each image in the second group. The result he obtained was “through statistical analysis, that the relationship of predictions (land-use map based scores) and the real photographs are in concordance and 26 significant to a high (95%) confidence level (Lu,2011, p38)”. That meant the equation of visual quality could be mapped and predicted in the Lower Muskegon Watershed. As an extension of this research, this thesis will focus on the southern part of Michigan. 3.5 Conclusion The scientific research of systematic visual landscape quality assessments was accepted and developed well in 20th century; both expert/design and perception-based approaches are accepted and made use by the public. And more effective technique and tools are created successfully in order to meet with better assessments. On the other side, there still be some problems people have not solved when they want to make deeper and more accurate research of the assessments in 21st century. Many challengers and controversy will be a lesson waiting for people to resolve. It seems that visual landscape quality assessments are developing continually 27 Chapter 2: Study Area and Methodology 2.1 Purposes of Study The purpose of this study is to develop a visual environmental quality map that is reliable and repeatable, similar to the Lower Muskegon Watershed map by Lu (2011). My hypothesis for the study is: Hypothesis 1: The predicted scores on a map have high concordance with the scores of the pictures taken from the same study area Null Hypothesis 1: The predicted scores have a low concordance with scores of the pictures taken in the study area. Also some other questions could be considered in a discussion after the initial study is complete include: Question 1: Which land-cover type has the highest preference in Southern Michigan? Why? Question 2: Which land-cover type get the lowest preference from people? What caused that to happen? Question 3: Besides the visual quality model, what else can the equation be used for in other applications? Question 4: What could be the next step for this line visual quality assessment research? 28 2.2 Method 2.2.1 Research Area Michigan is located in the Great Lakes Region of the United States of America. Michigan is the only state to consist of two peninsulas. These two peninsulas are linked by the Mackinac Bridge. The Upper Peninsula is separated from the Lower Peninsula by the Straits of Mackinac. The Lower Peninsula whose shape looks like a mitten is chosen as the main part of my study area (Figure 2.1). The area of this research is the Southern Michigan which is no further north than N44.2 in latitude. There are 47 counties in Southern Michigan including the 6 counties within the Lower Muskegon Watershed: Muskegon County, Lake County, Mecosta County, Montcalm County, Newaygo County, and Osceola County. The reasons why Southern Michigan has been picked for this research are: first, according to 2010 Census, the 10 largest municipalities, Detroit, Grand Rapids, Warren, Sterling Heights, Lansing, Ann Arbor, Flint, Dearborn, Livonia and Clinton Township are all located in Southern Michigan. The majority of people live in this southern portion Michigan contains many more people than the other part of Michigan and so this study might be useful to a larger group of people; second, the other important cities like Battle Creek, Benton Harbor, Jackson, Kalamazoo, Bay City, Midland, Holland, Saginaw and Muskegon are also in Southern Michigan; third, East Lansing where Michigan State is located is a central location in Southern Michigan, that leads to the convenience of collecting data from the Southern Michigan area; finally, based on the study of the Lower Muskegon Watershed, Southern Michigan is a good choice for continued research. 29 Figure 2.2.1: Geographic location of Southern Michigan and the relationship between the Southern Michigan and Lower Muskegon Watershed 30 For interpretation of the references to color in this and all other figures, the reader is referred to the electronic version of this dissertation. 2.2.2 Data Collection The data used in the research is comprised of two sources of information: pictures taken from the Southern Michigan by myself and previously the Lower Muskegon Watershed pictures taken by Di Lu (2011). It took several trips traveling around the state to collect the images of Southern Michigan. The positions of the pictures are tagged on the map by Global Positioning System (GPS). The landscape elements including vegetation, roads, flowers, buildings, people, facilities, farmland and forestry are the main content elements of the images. All the pictures were taken from the viewpoint of a person standing on the ground. The pictures were all collected during summer. Many of the important cities in the study area were visited to collect data. Table 2.2.2.1 The schedule of collecting data Number Date of times Destination (main cities visited) Number of pictures 1 5/9/2011 Lansing, Ann Arbor, Detroit 45 2 5/16/2011 Detroit 24 3 5/20/2011 Jackson, Kalamazoo, Grand Rapids, Portage 35 4 9/24/2011 Midland, Saginaw, Flint, Port Huron 30 Most of these pictures belong to 6 types which are studied in the research study: commercial, residential, downtown, farmland, industrial and forested land. There are also some 31 others types in the pictures including: campus, parking lot, airport, office park etc. They will not be counted as a part of the research. At the same time, when checking the previous data from the Lower Muskegon Watershed, Di Lu separated his image into two sets and 6 types: Downtown, Industry, Urban, Farmland, Water and Forest (2011).Comparing the two data sets, there were 5 similar types of land cover types: commercial, downtown, farmland, industrial and forested. The two data sets were combined to form the Southern Michigan data set which were regarded as the images taken from west side of Southern Michigan. Thus, the data contains 134 pictures take in Southern Michigan. The basic land-use map of Southern Michigan was obtained from “Geographic Data Library” which is at the site of “Michigan Department of Technology, Management & Budget”. The site is: http://www.mcgi.state.mi.us/mgdl/?rel=thext&action=thmname&cid=5&cat=Land+Cover%2F Use+MIRIS+1978. The land-cover map was made in 1978 and the data was gathered county by county. After merging them together, the original land-cover map of Southern Michigan was constructed. The original information of the structures of the land is defined as different types in the map’s legend at a level two classification. Land cover types were reclassified as commercial, residential, downtown, farmland, industrial, and forested. All these areas are accessible to the people. Other land-use types not used in the study are all defined as “others”(see table 2.2.2.2). Thus, this constructed map catalog for the research investigation was included to build the new reclassified land-cover map. 32 Table 2.2.2.2 Reclassification of the land-cover types Level 2 number Original classification New classification 11 Residential Residential 12 Commercial, Services, and Institutional Commercial 13 Industrial Industrial 14 Transportation, Communication, and Utilities Industrial 17 Extractive Industrial 19 Open and Other Downtown 21 Cropland, Rotation, and Permanent Pasture Farmland 22 Orchards, Vineyards, and Ornamental Farmland 23 Confined Feeding Operations Farmland 24 Permanent Pasture Farmland 29 Other Agricultural Land Farmland 31 Herbaceous Rangeland Forested 32 Shrub Rangeland Forested 33 Pine or Oak Opening (Savanna) Forested 4 Forest Land Forested 41 Broadleaved Forest (Generally Deciduous) Forested 42 Coniferous Forest Forested 51 Streams and Waterways Others 52 Lakes Others 53 Reservoirs Others 54 Great Lakes Others 6 Wetlands Others 33 Table 2.2.2.2 Reclassification of the land-cover types 62 Non-Forested (non-wooded) Wetlands Others 7 Barren Others 72 Beaches and Riverbanks Others 73 Sand Other than Beaches Others 74 Bare Exposed Rock Others Landcover Map of Southern Michigan Figure 2.2.2.1: Original land-cover map of Southern Michigan(Geographic Data Library.“Michigan Department of Technology, Management & Budget”. http://www.mcgi.state.mi.us/mgdl/?rel=thext&action=thmname&cid=5&cat=nd+Cover%2FUs e+MIRIS+1978.) 34 Figure 2.2.2.2: Reclassified land-cover map of Southern Michigan 2. 3 Analysis Techniques The equation used in the research is the same one of the one in Dr. Jon Burley’s research in 1997. It came from a respondent survey in North American, which had a high degree of reliability across many landscape settings. This equation is used to predict visual quality in his other studies. At the first step of the process to measure the variables of each picture a 38*30 grid is placed over the image, and every 1*1 square in the grid is counted as 1 unit. The numbers used 35 for each variable are obtained by counting areas or perimeters (Burley, 1997). Other variables in the equation employ an index adapted from Carol Smyser. The variables are: HEALTH= environmental quality index (Table 2.3) X1 = perimeter of immediate vegetation X2 = perimeter of intermediate non-vegetation X3 = perimeter of distant vegetation X4 = area of intermediate vegetation X6 = area of distant non-vegetation X7 = area of pavement X8 = area of building X9 = area of vehicle X10 = area of humans X11 = area of smoke X14 = area of wildflowers in foreground X15 = area of utilities X16 = area of boats X17 = area of dead foreground vegetation X19 = area of wildlife X30 = open landscapes = X2 +X4 + (2 ×(X3 +X6)) X31 = closed landscapes = X2 +X4 + (2 × (X1 +X17)) X32 = openness = X30 − X31 X34 = mystery = X30 × X1 × X7/1140 X52 = noosphericness = X7 + X8 + X9 + X15 + X16 36 The “health” index means the natural environmental quality of the site, which is shown in table 2.3: Table 2.3 The environmental health index in the equation. Variable Score 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 Total Score 37 After getting all the above data, the final score of the visual quality from Dr. Burley’s study is: Y= 68.30 - (1.878*HEALTH) - (0.131*X1) - (0.064*X6) + (0.020*X9) + (0.036*X10) + (0.129*X15) - (0.129*X19) - (0.006*X32) + (0.00003*X34) + (0.032*X52) + (0.0008*X1*X1) + (0.00006*X6*X6) - (0.0003*X15*X15) + (0.0002*X19*X19) - (0.0009*X2*X14) (0.00003*X52*X52) - (0.0000001*X52*X34) (3) The lower “Y” scores the better the visual quality. Thus, a group of predictive scores could be made. In my research study, I am going to test the validity a constructed map, which is based on the predictive scores for the area. Kendall’s Coefficient of Concordance (Daniel 1978) was employed to evaluate the map. It is the statistical method which tests the agreement between several different ranks (predicted and actual measured ) for the same object (a map). Basically, if there are “n” elements (“n” equals to 30 in this thesis) to be ranked by “m” judgers (“m” equals to 2 in the thesis), and suppose the object “i” is ranked “Rj” by the judger No. “j” , then we can get: The sum of the total ranks of object “i” which represents how much resemblance of all the rankings (Kendall, 1939, p276) is: Ri=R1+R2+R3+……+Rj+……+Rm (4) The mean value of the ranks of either object “r” is: R=1/2*m (n+1) (5) 38 As the author concerned, if S means “the observed sum of squares of the deviations of sums of ranks from the mean value” (Kendall, 1939, p276), then S will be: n S=∑ (Ri-R)2 (6) Specially, the maximum value of “S” appears at the situation while every judger has the same ranking for each element which means the resemblance is the largest just as author said: “if there is complete unanimity in the rankings, (Kendall, 1939, p276)” Thus: 2 2 2 Smax=[1*m-m(n+1)/2] +[2*m-m(n+1)/2] +[3*m-m(n+1)/2] +……+ [n*m-m(n+1)/2] 2 2 2 2 2 2 2 2 =(1 +2 +3 +……+n )m +nm (n+1) /4-m(n+1)m(1+2+3+……+n) 2 2 2 2 2 =m n(n+1)(2n+1)/6+ nm (n+1) /4-m (n+1) n/2 2 =m n(n+1)(4n+2+3n+3-6n-6)/12 2 = m n(n+1)(n-1) 2 3 =m (n -n)/12 (7) And Kendall’s Coefficient of Concordance “W” is defined as the proportion of S to Smax (Kendall, 1939, p276), which means: W=S/Smax=S/[ m2(n3-n)/12] 39 2 3 W=12S/[m (n -n)] (8) As the result, “W” is between 0 and 1. The less the “W” is, the more resemblance the rankings are. When “W” is 0, the ranking of the respondents share no agreement while when “W” is 1, all the respondents give the same ranking. In the thesis, “W” will be used to test the agreement with the predicted ranking and the actual ranking of the scores of the images’ visual quality. Chi-square distribution will also be obtained by consulting a chi-square table. If “W” is greater than the Chi-square number, the the results are in accordance with the p-values for that Chi-square number. 2.4 Sample Selection In order to get representative scores of the land-use in Southern Michigan, samples should be chosen dispersed across the whole area. In this research, the predictive model of visual quality is built on 6 types—each type contained 5 scores collected from 5 different parts of Southern Michigan. That is set 1, 30 scores to make a prediction map. There is another set, set 2. Set 2 is also composed of the same 6 types—each type also contained 5 scores from different parts of the study area. Thus, set 2 had 30 scores, too, which will be regarded as the sample testing and validating the predictive visual quality map. In this way, 60 scores are needed in total. The scores from west come from the pictures taken in the Southern Michigan in 2011 and by Di Lu. 40 Figure 2.4.1: Location of set 1 of commercial Figure 2.4.2: Location of set 2 of commercial Figure 2.4.3: Location of set 1 of residential Figure 2.4.4: Location of set 2 of residential Figure 2.4.5: Location of set 1 of downtown Figure 2.4.6: Location of set 2 of downtown 41 Figure 2.4.7: Location of set 1 of farmland Figure 2.4.8: Location of set 2 of farmland Figure 2.4.9: Location of set 1 of industrial Figure 2.4.10: Location of set 2 of industrial Figure 2.4.11: Location of set 1 of forestry Figure 2.4.12: Location of set 2 of forestry These samples need to be renamed as the number from 1-60 as the table 7. 42 Table 2.4 The selection of the samples in two sets Set 1 Source of data New Number Source of data Picture No.5 1 Picture No.38 31 Picture No.60 2 Picture No.74 32 3 Picture No.91 Picture No.123 4 Picture No.114 34 LMW set 1,No.24 5 LMW set 2,No.1 35 Picture No.19 6 Picture No.6 36 Picture No.53 7 Picture No.32 37 8 Picture No.98 Picture No.84 9 Picture No.125 39 Picture No.118 10 Picture No.129 40 Picture No.7 11 Picture No.30 41 Picture No.35 12 Picture No.81 42 13 Picture No.104 Picture No.120 14 Picture No.109 44 LMW set 1,No.11 15 LMW set 2,No.22 45 Picture No.12 16 Picture No.18 46 Picture No.25 17 Picture No.71 47 18 Picture No.116 Picture No.134 19 Picture No.122 49 LMW set 1,No.12 20 LMW set 2,No.10 50 Picture No.36 21 Picture No.40 51 22 Picture No.79 Picture No.96 Picture No.72 Picture No.99 Picture No.93 Property Set 2 Commercial Residential Downtown Farmland Picture No.69 Industrial Property Commercial Residential Downtown Farmland New Number 33 38 43 48 52 Industrial Picture No.82 23 Picture No.110 53 Picture No.95 24 Picture No.132 54 43 Table 2.4 The selection of the samples in two sets Picture No.17 26 Picture No.83 56 Picture No.68 27 Picture No.94 57 28 Picture No.119 58 Picture No.86 29 Picture No.127 59 LMW set 1,No.13 30 LMW set 2,No.18 60 Picture No.78 Forested Chapter 3: Results 3.1 Building the predictive map The scores in set one are presented in table 3.1. Table 3.1 The scores of set 1 Number Score 1 60.22472 2 79.42317 3 68.00592 4 65.89642 5 64.24578 6 63.67283 7 73.69797 8 55.94164 9 53.83222 10 61.58117 11 56.67802 12 69.28333 13 Property Mean Value Commercial 67.55920 Residential 61.74517 Downtown 73.53599 84.30117 44 Table 3.1 The scores of set 1 14 78.74230 15 78.67512 16 59.12320 17 57.74200 18 73.40120 19 66.19820 20 53.00983 21 88.84700 22 86.56068 23 81.12710 24 94.63968 25 107.25810 26 54.40120 27 59.55400 28 56.13090 29 46.98920 30 59.41400 Downtown 73.53599 Farmland 61.89489 Industrial 91.68651 Forested 55.29786 The predictive model of Southern Michigan can be set up by the set 1 and presented in Figure 3,1. 45 Figure 3.1: Predictive visual quality map of Southern Michigan 3.2 Validating the Map After building the predictive visual quality map of Southern Michigan, each value of the six types can be obtained from the mean value of a land-use from five scores in set 1. And according to the scores of set 1, the mean values are ranked as: industrial: 91.68651, downtown: 73.53599, commercial: 67.55920, farmland: 61.89489, residential: 61.74517, forested: 55.29786. Later, the scores in set 2 are also calculated equation 3. The next step is validating the map by the scores of set 2. 46 According to Kendall’s theory and based on the ranking of the six types, the average predictive ranking of them are: industrial’s predictive ranking is 3 which is the mean value of ranking 1,2,3,4,5; downtown’s is 8 which is the mean value of ranking 6,7,8,9,10; commercial’s predictive ranking is 13 which is the mean value of ranking 11,12,13,14,15; farmland’s is 18 which is the mean value of ranking 16,17,18,19,20; residential’s predictive ranking is 23 which is the mean value of ranking 21,22,23, 24,25; downtown’s is 28 which is the mean value of ranking 26,27,28,29,30. In addition, all the samples of set 2 has its real ranking which is in table 3.2.1. Table 3.2.1 The scores, predictive rankings and real rankings of set 2 Property Predicted Ranking Set 2 Number Score Real Ranking Industrial 3 55 107.80500 1 Downtown 8 42 89.67798 2 Industrial 3 52 83.11517 3 Industrial 3 51 82.92093 4 Commercial 13 31 82.12213 5 Downtown 8 45 81.84413 6 Downtown 8 41 77.90797 7 Industrial 3 54 77.11530 8 Commercial 13 33 76.05165 9 Downtown 8 43 75.23112 10 Commercial 13 32 74.14795 11 Downtown 8 44 74.04022 12 Residential 23 39 73.41253 13 Commercial 13 35 72.91732 14 Industrial 3 53 67.91152 15 Commercial 13 34 66.76252 16 47 Table 3.2.1 The scores, predictive rankings and real rankings of set 2 Farmland 18 46 66.15512 17 Residential 23 36 65.50428 18 Residential 23 38 62.65881 19 Farmland 18 47 62.13320 20 Residential 23 37 61.67768 21 Residential 23 40 61.63773 22 Forested 28 58 55.19320 23 Forested 28 59 55.08880 24 Farmland 18 49 54.73720 25 Forested 28 60 54.71145 26 Forested 28 56 53.82770 27 Forested 28 57 53.74060 28 Farmland 18 48 53.27078 29 Farmland 18 50 51.93535 30 Then “W” value can be obtained by the two different rankings: the predictive one and the real one. Table 3.2.2 shows the process of getting the “W”. In this thesis, the “n”=30, “m”=2, mean ranking=2*(30+1)/2=30. Table 3.2.2 The W value and calculating process Predicted Number Rank 3 8 3 3 13 8 8 3 55 42 52 51 31 45 41 54 Score Rank 107.80500 89.67798 83.11517 82.92093 82.12213 81.84413 77.90797 77.11530 1 2 3 4 5 6 7 8 Sum of the ranks 4 10 6 7 18 14 15 11 48 Ranks squared 16 100 36 49 324 196 225 121 Sum of rank minus mean -27 -21 -25 -24 -13 -17 -16 -20 Squared deviations 729 441 625 576 169 289 256 400 Table 3.2.2 The W value and calculating process 13 8 13 8 23 13 3 13 18 23 23 18 23 23 28 28 18 28 28 28 18 18 33 43 32 44 39 35 53 34 46 36 38 47 37 40 58 59 49 60 56 57 48 50 76.05165 75.23112 74.14795 74.04022 73.41253 72.91732 67.91152 66.76252 66.15512 65.50428 62.65881 62.13320 61.67768 61.63773 55.19320 55.08880 54.73720 54.71145 53.82770 53.74060 53.27078 51.93535 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 22 18 24 20 36 27 18 29 35 41 42 38 44 45 51 52 43 54 55 56 47 48 484 324 576 400 1296 729 324 841 1225 1681 1764 1444 1936 2025 2601 2704 1849 2916 3025 3136 2209 2304 -9 -13 -7 -11 5 -4 -13 -2 4 10 11 7 13 14 20 21 12 23 24 25 16 17 sum of deviations S= W= 2 81 169 49 121 25 16 169 4 16 100 121 49 169 196 400 441 144 529 576 625 256 289 8030 8050 0.893214683 3 As the equation, the value of W is equal to 12*8030/2 *(30 -30) which is 0.8932. The degree of freedom is 29 in this research. And the derived value at the end is 51.80645161, which comes from 2*(30-1)*0.8932. According to the Chi-Square table, when df is 29, 51.80645161 is bigger than 49.58788 which means 99% confidence level (p≤0.01). So, the hypothesis that the two sets of images have high concordance (99%) with each other is 49 successfully obtained. In another words, with the statistic work, the predictive map of visual quality in Southern Michigan have a 99% concordance and are significantly similar to the real images. Chapter 4 Discussion 4.1 The ranking of six types The first idea I would like to discuss is: why the mean scores, from low to high, of these six types are ranged in the following order: forested, residential, farmland, commercial, downtown and industrial? (Residential and farmland are actually very close) That means respondents prefer the former than the later. I selected one of each type from set 1 to make a comparison of the scores, and typically each of them is the one who is very close to the mean value. 50 Figure 4.1.1: Sample No.26 of Forested (54.40120) — Copyright © 2011, Yuemin Jin Comparing with these six samples that represent the mean values of their types, it is concluded that because of the different features in different type and due to the different quantities and percentage these features occupied, they are certain to have varied visual quality scores. The No.26 sample of Forestry (the score is 54.40120) contains almost all natural features, including a lot of immediate vegetation, wildflowers on the ground and no disturbance by humans. Also, it offers a pleased aesthetic quality, bringing pure air and soil to the environment which is thought of as being very healthy. Thus, forestry gets high preference of visual quality by people. Figure 4.1.2: Sample No.10 of Residential (61.58117) — Copyright © 2011, Yuemin Jin 51 The sample No.10 of Residential (score is 61.58117) is the one combining nature with the human living. With much vegetation around the house that gives a good sight to the residents, people feel that they are living with peaceful, pleasant and silent surroundings which make them prefer the visual quality. Figure 4.1.3: Sample No.16 of Farmland (59.12320) — Copyright © 2011, Yuemin Jin The sample No.16 of Farmland (score is 59.12320) is similar as forestry which is full of natural features. Although there is not much woody vegetation in the image, the site is thought of as a very healthy one, which not only offers soil, water resources and fresh air, but also makes viewers thinking of that it is a great place where food comes from. Moreover, high openness is another characteristic of the farmland. No human disturbance is observed and nature is strongly present, making the farmland welcomed by the viewers. 52 Figure 4.1.4: Sample No.3 of Commercial (68.00592) — Copyright © 2011, Yuemin Jin The sample No.3 of Commercial (score is 68.00592) contains only a little vegetation with some roads, utilities, buildings. Through it is still with fair openness, it makes people think of that it is a modern man-made site with more or less natural features participates in. Some other samples contain some human beings and vehicles which make it become less appreciated by all. Figure 4.1.5: Sample No.12 of Downtown (69.28333) — Copyright © 2011, Yuemin Jin 53 The sample No.12 of downtown (score is 69.28333), comparing with commercial areas, is more crowded with high buildings, busy traffic and more human beings. Although it might not pollute the air or water, it brings much less vegetation and other natural features to the people. That is why it cannot get better preference by the respondents. Figure 4.1.6: Sample No.22 of Industrial (86.56068) — Copyright © 2011, Yuemin Jin Last, the sample No.22 of Industrial (86.56068) has really poor preference. Because it not only contains none of the natural elements like vegetation, soil, flowers in the sight of the people, but also pollutes the natural environment more or less. Using so many utilities, making the sky grey while leaving no place for the vegetation, all which makes less people prefer the scenic at industrial sites. 4.2 The deviations of the ranking 54 The second question is: why different type has different deviations in the predictive ranking? The different composition of each type probably helps to understand the reason. Here in each type the highest one and lowest one from set 2 are picked and compared with each other, which will explain the phenomenon. Figure 4.2.1: Sample No.53 of Industrial Figure 4.2.2: Sample No.55 of Industrial — Copyright © 2011, Yuemin Jin — Copyright © 2010, Di Lu The industrial sites are complicated. During the process of taking pictures, it is found that some factories are small and built close to the downtown. As the sample No. 53 whose score is 67.9152, the surrounding is clean and with more trees around it. The trees also provide shade to the people. Consequently, the visual quality of this site is surely not bad. And it’s ranking is 15 instead of the predicted rank: 8. Some factories have large scale and may give off polluted air, so they are built far from the center of the city. The environment of sample No.55 who has 107.80500 is the one that only be used for industry and don’t want people to stay aside. It is believed that little people will like the visual quality of it. The far more different scale, location and function of the factory make the scores far away from each other. Thus, the real ranking may close to the predictive ranking. 55 Figure 4.2.3: Sample No.44 of Downtown Figure 4.2.4: Sample No.42 of Downtown — Copyright © 2011, Yuemin Jin — Copyright © 2011, Yuemin Jin Owning to the fact that different cities have their different history, development process and main function, their scores may not very close to each other. The sample No.44 whose score is 74.04022 shows the downtown is built as a traditional city center with historic architectures, colorful vegetation and classical utility for people having activities, taking trip in, in this way, people may like it. The sample No.42 whose score is 89.67798 looks the opposite, a mass of vehicles, utilities, busy traffic and high-rises fill into the imagine making the image disliked by the respondents. This situation will appear in many different cities if they haven’t got the same functions. And that’s why the predictive ranking of downtown not the same as the real ranking. 56 Figure 4.2.5: Sample No.34 of Commercial Figure 4.2.6: Sample No.31 of Commercial — Copyright © 2011, Yuemin Jin — Copyright © 2011, Yuemin Jin Sometimes, some differences of the samples are caused by the position of the sites, or in words, the different conditions of different places. For instance, if one city or county is well organized, the commercial area like sample NO.34 whose score is 66.76252 will at least be a clean one with appropriate quantities of utilities, vehicles and shadows to the people. If the site is developed for wholesale as sample No.31 whose score is 82.12213, then it will contain large outdoor open areas for the large amounts of vehicles. The latter one may not get a good preference score from people because it is a functional one instead of an enjoyment. Also, No. 31’s real ranking (5) is far away than the predictive ranking (13). Figure 4.2.7: Sample No.50 of Farmland Figure 4.2.8: Sample No.46 of Farmland — Copyright © 2010, Di Lu — Copyright © 2011, Yuemin Jin 57 Figure 4.2.9: Sample No.40 of Residential Figure 4.2.10: Sample No.39 of Residential — Copyright © 2011, Yuemin Jin — Copyright © 2011, Yuemin Jin The scores of farmland and residential are very closed. Still, some of these samples in set two are beyond the predictive rankings. Comparing with the images close to the predictive ranking and far from the ranking, it is found that some elements lead to the result. Some differences happen to appear because of the random in the picture taking process. For example: even in the same or near neighborhood, whether residential samples contain cars, human beings, shade or the number of the vegetation in the image will influence the scores just like sample NO.40 whose score is 61.63773 and sample NO.39 whose score is 73.41253 and as a result No.39 is far from the predicted ranking; the proportion of roads or utilities appeared in the pictures will affect the result, Sample No.46 whose score is 66.15512 contains more road and pavement, that makes it has higher score than sample No.50 whose score is 51.93535 which is full of green vegetation and shade, and of course, it gets a good preference. Usually, these kinds of distinction won’t make big unlikeness in the final scores, taking more pictures and get the mean value can balance these scores. 58 Figure 4.2.11: Sample No.57 of Forested Figure 4.2.12: Sample No.58 of Forested — Copyright © 2011, Yuemin Jin — Copyright © 2011, Yuemin Jin For the forested land, which has really high visual quality according to the predicted model, there are little differences of the features in the sample 57 and 58 through they are the highest rank and lowest rank. And their scores are quite similar. Although they were taken in different areas and have different depth and level of trees, shrubs and grass, the two images are all full of vegetation. Between these two samples, the distance is only around 1.5. That means respondents will share almost the same view of the landscape of forestry at Southern Michigan and they all like it. For forested land, the predictive map is really helpful. 4.3 The main points of the research If the research is introduced to those who have no professional background of visual quality assessment, there will be three main points which are worth discussing and getting ideas of visual quality assessment. 1. There is a new methodology of investigating the sense of the landscape from people, which depends on an exact mathematics score instead of the description of the scenic. 59 2. With the scores of a group of landscape, people can not only know which ones are better, but also get the idea of how much better than the others. 3. With the map which is constructed by the scores, there is no need for people to take pictures from all over the areas. The map will predict the visual quality of the whole area. 5. Limitation Although the predictive map of visual quality in Southern Michigan have 99% concordance and significant to the real images by the statistical work, the scores in each type still haave more or less exceptions which are beyond the predictive ranking. What’s more, thinking of the scores in set two, some of them are far from the mean value in set one. It is considered that if some more measures are taken in the process, the predictive map will be more accurate in the future. First of all, the classification of land-cover is a brief one. Even in the same type, there are several different detailed classifications: Industrial contains Commercial, Services, Institutional, Industrial, Transportation, Communication, Utilities, and Extractive, Farmland contains Cropland, Rotation, and Permanent Pasture, Orchards, Vineyards, and Ornamental, Confined Feeding Operations, Permanent Pasture and Other Agricultural Land; Forestry contains Herbaceous Rangeland, Shrub Rangeland, Pine or Oak Opening (Savanna), Forest Land, Broadleaved Forest (Generally Deciduous) and Coniferous Forest. This is just the second level of the land-cover classification from GIS data (level 2). If the third level of the GIS data (level 3) is considered into the map, there will be more delicate classification of the land-cover, for example, the residential areas are composed of single families, multiplied families, neighborhood etc.. Moreover, there are some of the types not involved in this research such as Streams and 60 Waterways, Lakes, Wetlands, Barren, Sand, Beaches, Bare Exposed Rock etc. which occupy plenty of the acreage in Southern Michigan through people may not go there. Furthermore, some places are hard to classify like the campus, parking lot and so on. If researchers want to make a deeper and detailed predictive visual quality map of Southern Michigan, all of these particular types of the land, for example the classification in level 2, might be further considered. Chances are that some of them may have very different scores even they are in the same brief land-cover type. Secondly, the basic map of the site was made at 1978, 34 years before now. This foundational data is so old that nobody can assures that certain place in the map is still the same land-cover type. It would be helpful to have more recent data. Thirdly, according to the validating result, the scores of downtown, commercial, industrial may have differences due to the different function of the cities. It may be more accurate to get the better validating result by classify the cities based on the function, for example, cities for tourism, cities for industry, cities for agriculture, cities holding history or cities for convenient transportation etc. . In this way, cities with similar function will have closer and accurate scores in their downtown, commercial and industrial areas. Another limitation in this research is that the samples were all picked in late spring and summer time. That means, the seasonal changes are ignored in the visual quality map. The seasonal changes may happen more or less on natural environment such as vegetation, water, wildlife and the quality of the air. As the result, the scores of the images may not modify a lot on industrial, downtown, commercial areas but may influence the result of the scores of the pictures 61 of forestry, residential areas or farmland. A study of the visual quality score in same position in different period which reflect the seasonal change is recommended in the future. The limitation also appears in the numbers of sample. The more samples selected and calculated by the equation, the better mean value will be to predict visual quality. In this research, each type of the visual quality predictive score is built by balancing the 5 scores of the samples as well as 5 samples in another set are selected validating the map. If the number of the samples is larger, the predictive scores may be more accurate. And thus the result of the predictive ranking might influence the final map a little. There may be some future further studies following this one. The scale of the study area can extend to the whole Michigan State, then the whole United State, then the North America and even the whole world. One of the issues will be comparing the different visual quality of the areas with same characteristic but in different cities, states, countries or continents and discussing what causes the distinctions such as the environmental quality, atmosphere, geographic position, cultural background, population, human income etc.. Moreover, the zoom (proximity) of the pictures can be considered for further research. The reason is that if the picture is taken very close to the object, the viewpoint of the landscape might change a lot. The change of the zoom should also be checked with how much detail the original GIS document can offer. What’s more, making use of the predictive visual quality map, a guideline of visual quality can be created for those who want to travel, work, or even settle down to other cities, state, or countries. In the reference of the visual quality map which may own different scores in different areas, people will have a basic impression of the visual quality degree in the goal places. 62 If they need to make some choice from these places, than the predictive map can be a significant direction for them. The predictive visual quality map can also be a part of the analysis program in the planning or designing process of a project. Planners may get the basic idea for the current visual quality of the site in order to know which construction projects require planning and design sensitivity and valued areas in certain places should be protected. By the predictive map, the advantage and disadvantage of the area will be shown to them. Furthermore, a model can be built according to the planners’ and designers’ creation by digital software. The model will also be evaluated as a sample, testing if the visual quality of the expected plan can be preferred by viewers. That will give the planners and designers an effective direction, too. 6. Conclusion After realizing that people may have high coincidence in their preference of the landscape aesthetic quality, many researchers pay attention to the realm seeking the exact relationship of the preference of visual quality and the features of the site. Though some researchers insist on judging the visual quality by the professional point of view, there are many others who keep on looking for common respondents. Some equations are made consequently. The present research is based on one of the equation which came from a study in North America. The next step is building a predictive visual quality map of Southern Michigan by a group of data and then validating the map by another group of samples. The result shows that the predictive map of visual quality in Southern Michigan is successful which has 99% concordance and significant to the real images. Despite the fact that there are several limitations in the research, such as lacking of recent original land-use data, without detailed classification, the 63 research is valuable. By predicting the visual map, the condition of different areas caused by their historic background, environment, economic position, culture aesthetic standards can be found. And also it offer a guidelines and direction for both the common people who will move from one place to another and the professional planners and designers who want to improve the areas. It is hoped this research paper can provide some information for the further researchers who want to make deeper visual quality assessment study, no matter in Southern Michigan or in even larger areas. 64 APPENDIX 65 Figure A.1: Set 1, NO.1 Figure A.2: Set 1, NO.2 — Copyright © 2011, Yuemin Jin — Copyright © 2011, Yuemin Jin Figure A.3: Set 1, NO.3 Figure A.4: Set 1, NO.4 — Copyright © 2011, Yuemin Jin — Copyright © 2011, Yuemin Jin Figure A.5: Set 1, NO.5 Figure A.6: Set 1, NO.6 — Copyright © 2010, Di Lu — Copyright © 2011, Yuemin Jin 66 Figure A.7: Set 1, NO.7 Figure A.7: Set 1, NO.7 — Copyright © 2011, Yuemin Jin — Copyright © 2011, Yuemin Jin Figure A.9: Set 1, NO.9 Figure A.10: Set 1, NO.10 — Copyright © 2011, Yuemin Jin — Copyright © 2011, Yuemin Jin Figure A.11: Set 1, NO.11 Figure A.12: Set 1, NO.12 — Copyright © 2011, Yuemin Jin — Copyright © 2011, Yuemin Jin 67 Figure A.13: Set 1, NO.13 Figure A.14: Set 1, NO.14 — Copyright © 2011, Yuemin Jin — Copyright © 2011, Yuemin Jin Figure A.15: Set 1, NO.15 Figure A.16: Set 1, NO.16 — Copyright © 2010, Di Lu — Copyright © 2011, Yuemin Jin Figure A.17: Set 1, NO.17 Figure A.18: Set 1, NO.18 — Copyright © 2011, Yuemin Jin — Copyright © 2011, Yuemin Jin 68 Figure A.19: Set 1, NO.19 Figure A.20: Set 1, NO.20 — Copyright © 2011, Yuemin Jin — Copyright © 2010, Di Lu Figure A.21: Set 1, NO.21 Figure A.22: Set 1, NO.22 — Copyright © 2011, Yuemin Jin — Copyright © 2011, Yuemin Jin Figure A.23: Set 1, NO.23 Figure A.24: Set 1, NO.24 — Copyright © 2011, Yuemin Jin — Copyright © 2011, Yuemin Jin 69 Figure A.25: Set 1, NO.25 Figure A.26: Set 1, NO.26 — Copyright © 2010, Di Lu — Copyright © 2011, Yuemin Jin Figure A.27: Set 1, NO.27 Figure A.28: Set 1, NO.28 — Copyright © 2011, Yuemin Jin — Copyright © 2011, Yuemin Jin Figure A.29: Set 1, NO.29 Figure A.30: Set 1, NO.30 — Copyright © 2011, Yuemin Jin — Copyright © 2010, Di Lu 70 Figure A.31: Set 2, NO.32 Figure A.32: Set 2, NO.32 — Copyright © 2011, Yuemin Jin — Copyright © 2011, Yuemin Jin Figure A.33: Set 2, NO.33 Figure A.34: Set 2, NO.34 — Copyright © 2011, Yuemin Jin — Copyright © 2011, Yuemin Jin Figure A.35: Set 2, NO.35 Figure A.36: Set 2, NO.36 — Copyright © 2010, Di Lu — Copyright © 2011, Yuemin Jin 71 Figure A.37: Set 2, NO.37 Figure A.38: Set 2, NO.38 — Copyright © 2011, Yuemin Jin — Copyright © 2011, Yuemin Jin Figure A.39: Set 2, NO.39 Figure A.40: Set 2, NO.40 — Copyright © 2011, Yuemin Jin — Copyright © 2011, Yuemin Jin Figure A.41: Set 2, NO.41 Figure A.42: Set 2, NO.42 — Copyright © 2011, Yuemin Jin — Copyright © 2011, Yuemin Jin 72 Figure A.43: Set 2, NO.43 Figure A.44: Set 2, NO.44 — Copyright © 2011, Yuemin Jin — Copyright © 2011, Yuemin Jin Figure A.45: Set 2, NO.45 Figure A.46: Set 2, NO.46 — Copyright © 2010, Di Lu — Copyright © 2011, Yuemin Jin Figure A.47: Set 2, NO.47 Figure A.48: Set 2, NO.48 — Copyright © 2011, Yuemin Jin — Copyright © 2011, Yuemin Jin 73 Figure A.49: Set 2, NO.49 Figure A.50: Set 2, NO.50 — Copyright © 2011, Yuemin Jin — Copyright © 2010, Di Lu Figure A.51: Set 2, NO.51 Figure A.52: Set 2, NO.52 — Copyright © 2011, Yuemin Jin — Copyright © 2011, Yuemin Jin Figure A.53: Set 2, NO.53 Figure A.54: Set 2, NO.54 — Copyright © 2011, Yuemin Jin — Copyright © 2011, Yuemin Jin 74 Figure A.55: Set 2, NO.55 Figure A.56: Set 2, NO.56 — Copyright © 2010, Di Lu — Copyright © 2011, Yuemin Jin Figure A.57: Set 2, NO.57 Figure A.58: Set 2, NO.58 — Copyright © 2011, Yuemin Jin — Copyright © 2011, Yuemin Jin Figure A.59: Set 2, NO.59 Figure A.60: Set 2, NO.60 — Copyright © 2011, Yuemin Jin — Copyright © 2010, Di Lu 75 BIBLIOGRAPHY 76 BIBLIOGRAPHY Bourassa, S. C. (1991). The Aesthetic of Landscape. Belhaven Press. London and New York. 120-126 Brown, T. C., Daniel, T. C., (1987). Context effects in perceived environmental quality assessments: scene selection and landscape quality ratings. J. Environ. Psychol. 7, 233-250. Burley, J. B. (1997).Visual and Ecological Environmental Quality Model for Transportation Planning and Design. Transportation Research Record: Journal of the Transportation Research Board. Transportation Research Board of the National Academies, 1549, 54-60. Burley, J.; Deyoung, G.; Partin, S.; Rokos, J. (2011). Reinventing Detroit: Reclaiming Grayfields—New Metrics in Evaluating Urban Environments. Challenges. 2, 45-54. Carlson, A.A. (1977). On the Possibility of Quantifying Scenic Beauty. Landscape Planning. 4,131-172. Crawford, D. (1994). Using Remotely Sensed Data in Landscape Visual Quality Assessment. Landscape and Urban Planning. 30, 71-81. Daniel, T. C. (1990). Measuring the quality of the human environment: a psychophysical approach. Am. Psychol. 45, 633-637. Daniel, T. C. (2001). Whither Scenic Beauty? Visual Landscape Quality Assessment in the 21st Century. Landscape and Urban Planning. 54, 267-281 Daniel, W.W., 1978. Applied nonparametric statistics. Houghton Mifflin Company. 326–335. Dramstad, W., Tveit, M., Fjellstad, W., & Fry, G. (2006). Relationships between visual landscape preferences and map-based indicators of landscape structure. Landscape and Urban Planning.78(4), 465-474. Hadrian, D.R., Bishop, I.D., Mitcheltree, R., (1988). Automated Mapping of Visual Impacts in Utility Corridors. Landscape and Urban Planning.16,261–282. Kendall, M.G., (1939). The problem of m rankings. The annals of mathematical statistics. 10(3), 275-287. Latimer, D.A., Hogo, H., Daniel, T.C. (1981). The Effects of Atmospheric Optical Conditions on Perceived Scenic Beauty. Atmospheric Environment.15,1865-1674. Lu,D.. Visual Quality Assessment at Lower Muskegon Watershed. (2011) Mo, F., G. Le Clé ach, M. Sales, G. Deyoung, and Burley, J.B. (2011). Visual and environmental quality perception and preference in the People's Republic of China, France, and Portugal. International Journal of Energy and Environment. 4,5,549-557. 77 Partin, S.. (2011). Sketchup Validity Modeling: A Comparison between Photographs and 3D Portugal. International Journal of Energy and Environment. 4(5), 549-556. Shafer, E. L., Jr. (1969). Perception of Natural Environment. Environment and Behavior.1,17182. Taylor, , J. G., Zube, E. H., Sell., J. L., (1987). Landscape Assessment and Perception Research Methods. Methods in Environmental and Behavioral Research. Eds. Bechtel, R. B. et al. Van NOstrand Reinhold Company,361-393. Weinstein, N. D.. (1976).The Statistical Prediction of Environmental Preferences: “Problems of Validity and Application”. Environment and Behavior,8,4,611-626. 78