**1. Introduction**

For almost 50 years, investigators have been seeking quantitative methods to predict and assess the visual and environmental quality of the landscape. The literature on this subject is vast and continues as illustrated through recent investigations and contributions by psychologists, engineers, landscape architects, planners, and natural resource specialists [1–11]. One of the best summaries of the early efforts was described by Taylor et al. [12]. Despite the

© 2016 The Author(s). Licensee InTech. This chapter is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. © 2018 The Author(s). Licensee IntechOpen. This chapter is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

scientific advances, the application of equations and exploration of theories seemed somewhat impractical for many practitioners. In attempts to translate the research, two of the best practical summaries of the ideas revealed by the research were found in *The Experience of Nature: a Psychological Perspective* and in *With People in Mind: Design and Management of Everyday Nature* [13, 14]. Along with developing predictive equations, investigators were interested in constructing maps that predict visual quality. Brush and Shafer produced one of the earliest well known visual quality maps [15]. But this approach was not widely adopted. In the United States, a heuristic approach (without strong statistical evidence for the variables and coefficients but instead based upon normative theory) comprised of an index developed by Jones and Jones, becoming widely employed and over time refined [16, 17]. The success of this index may have hindered the development of more science-based equations, as stakeholders adopted this relatively easily understandable methodology and were unwilling to seek additional methods. The actual science supporting this normative theory approach has been relatively weak and not aggressively challenged by the scientific community. But to landscape architects not trained in the ways of statistical analysis, p-values, and variance explanations, the index seemed to make good logical sense. In some respects, there seemed to be a lull at the turn of the century concerning visual quality assessment, where a methodology to validate maps was not self-evident, and the production of equations to explain increased levels of the variance were at a standstill. The early attempts to predict visual quality were primarily focused upon artistic composition normative values such as foreground, mid-ground, background, geometrical harmony, and natural/urban components. By pursuing this set of esthetic and spatial variables, investigators were able to explain 30% to nearly 70% of the variance [18].

perceptions about landscape. Asians may have a different sensibility concerning environmental preference. Concurrently, on the digital visualization forefront Partin et al. studied the response of participants to computer generated images and reported that the perception of computer images was similar to the perception of photographs of landscapes [25]. In other words, it was possible for investigators to present computer-generated images to respondents and to obtain a similar response as if the respondents were examining photographs. He also demonstrated how a small study could be folded into a larger and widely studied set of images to obtain stable and reliable results. However, these equations and results are still formative and require duplication by others to refute them or support them. In addition, there are opportunities to explore the responses of various cultural groups and to refine these equations. While the literature on this topic is vast, there seems to be many more instances to

A Visual Quality Prediction Map for Michigan, USA: An Approach to Validate Spatial Content

http://dx.doi.org/10.5772/intechopen.79490

115

Also during this recent timeframe, attempts to explore mapping making potential were renewed. Lu et al. examined the Lower Muskegon watershed, located on the west side of the lower peninsula in Michigan [26]. He and his colleagues studied images of urban areas, farmland, wetlands, and forests and attempted to construct an environmental quality map of his study area. The results he obtained through statistical analysis revealed that the relationship between his predictions in the map and the real photographs are in concordance and at a reasonable (95%) confidence level. He concluded that visual/environmental quality could be mapped and reliably predicted in the Lower Muskegon Watershed. There was a strong relationship between the perception of environmental quality and land cover. Following Lu, Jin examined the southern portion of Michigan, an area much larger than the Lower Muskegon Watershed [27]. She reported similar results to Lu. She had demonstrated that it was possible to develop reliable maps for much larger areas, but still not at the scale of a province or state.

add to the body of knowledge.

**Figure 1.** The location of Michigan, the study area in North America.

At the time, some scholars were frustrated with being unable to make much new headway. Some more recent investigators entered the subject area and began to explore the importance of other types of variables that were less esthetic in character, and more ecological, cultural, economic, and functional [19]. At a conference in 2005 in Switzerland (Our Shared Landscape 2005, Integrating ecological, socio-economic and esthetic aspects in landscape planning and management´ http://www.osl.group.shef.ac.uk, at Centro Stefano Franscini in Ascona, Switzerland), scholars were coming to the realization that respondents evaluated the landscape with more than just esthetic values. When these other potential predictors are added to the study, more of the variance was explained collectively and statistically. Stronger and more reliable predictive equations evaluating landscapes could be generated from respondents, explaining over 90% of the variance when esthetic, ecological, cultural, functional, and economic predictors are combined [20–22]. Along with the equations a series of theories to explain the results evolved [23]. Investigators noted that contents where humans infringed upon other humans (buildings, roads, people) were less preferred (even high quality architecture)—human disturbance theory; the benign environment (plants, waters, sky) were neutral in preference—natural preference theory; and temporal features (wildlife, momentary views of mountains, and flowers) were in preferred environments—temporal enhancement theory. Thus, theory and models were advancing together concerning evaluating and assessing the quality of the environment.

During this period, some scholars were beginning to explore the potential for globally universal predictive equations. Mo et al. recently reviewed much of this pertinent literature and discusses the perceptions of respondents in North America, France, Portugal, and PR China [24]. One finding suggests that Europeans and North Americans may have broad similar perceptions about landscape. Asians may have a different sensibility concerning environmental preference. Concurrently, on the digital visualization forefront Partin et al. studied the response of participants to computer generated images and reported that the perception of computer images was similar to the perception of photographs of landscapes [25]. In other words, it was possible for investigators to present computer-generated images to respondents and to obtain a similar response as if the respondents were examining photographs. He also demonstrated how a small study could be folded into a larger and widely studied set of images to obtain stable and reliable results. However, these equations and results are still formative and require duplication by others to refute them or support them. In addition, there are opportunities to explore the responses of various cultural groups and to refine these equations. While the literature on this topic is vast, there seems to be many more instances to add to the body of knowledge.

scientific advances, the application of equations and exploration of theories seemed somewhat impractical for many practitioners. In attempts to translate the research, two of the best practical summaries of the ideas revealed by the research were found in *The Experience of Nature: a Psychological Perspective* and in *With People in Mind: Design and Management of Everyday Nature* [13, 14]. Along with developing predictive equations, investigators were interested in constructing maps that predict visual quality. Brush and Shafer produced one of the earliest well known visual quality maps [15]. But this approach was not widely adopted. In the United States, a heuristic approach (without strong statistical evidence for the variables and coefficients but instead based upon normative theory) comprised of an index developed by Jones and Jones, becoming widely employed and over time refined [16, 17]. The success of this index may have hindered the development of more science-based equations, as stakeholders adopted this relatively easily understandable methodology and were unwilling to seek additional methods. The actual science supporting this normative theory approach has been relatively weak and not aggressively challenged by the scientific community. But to landscape architects not trained in the ways of statistical analysis, p-values, and variance explanations, the index seemed to make good logical sense. In some respects, there seemed to be a lull at the turn of the century concerning visual quality assessment, where a methodology to validate maps was not self-evident, and the production of equations to explain increased levels of the variance were at a standstill. The early attempts to predict visual quality were primarily focused upon artistic composition normative values such as foreground, mid-ground, background, geometrical harmony, and natural/urban components. By pursuing this set of esthetic and spatial variables, investigators were able to explain 30% to nearly 70% of the variance [18]. At the time, some scholars were frustrated with being unable to make much new headway. Some more recent investigators entered the subject area and began to explore the importance of other types of variables that were less esthetic in character, and more ecological, cultural, economic, and functional [19]. At a conference in 2005 in Switzerland (Our Shared Landscape 2005, Integrating ecological, socio-economic and esthetic aspects in landscape planning and management´ http://www.osl.group.shef.ac.uk, at Centro Stefano Franscini in Ascona, Switzerland), scholars were coming to the realization that respondents evaluated the landscape with more than just esthetic values. When these other potential predictors are added to the study, more of the variance was explained collectively and statistically. Stronger and more reliable predictive equations evaluating landscapes could be generated from respondents, explaining over 90% of the variance when esthetic, ecological, cultural, functional, and economic predictors are combined [20–22]. Along with the equations a series of theories to explain the results evolved [23]. Investigators noted that contents where humans infringed upon other humans (buildings, roads, people) were less preferred (even high quality architecture)—human disturbance theory; the benign environment (plants, waters, sky) were neutral in preference—natural preference theory; and temporal features (wildlife, momentary views of mountains, and flowers) were in preferred environments—temporal enhancement theory. Thus, theory and models were advancing together concerning evaluating and assessing the quality of the environment. During this period, some scholars were beginning to explore the potential for globally universal predictive equations. Mo et al. recently reviewed much of this pertinent literature and discusses the perceptions of respondents in North America, France, Portugal, and PR China [24]. One finding suggests that Europeans and North Americans may have broad similar

114 Land Use - Assessing the Past, Envisioning the Future

Also during this recent timeframe, attempts to explore mapping making potential were renewed. Lu et al. examined the Lower Muskegon watershed, located on the west side of the lower peninsula in Michigan [26]. He and his colleagues studied images of urban areas, farmland, wetlands, and forests and attempted to construct an environmental quality map of his study area. The results he obtained through statistical analysis revealed that the relationship between his predictions in the map and the real photographs are in concordance and at a reasonable (95%) confidence level. He concluded that visual/environmental quality could be mapped and reliably predicted in the Lower Muskegon Watershed. There was a strong relationship between the perception of environmental quality and land cover. Following Lu, Jin examined the southern portion of Michigan, an area much larger than the Lower Muskegon Watershed [27]. She reported similar results to Lu. She had demonstrated that it was possible to develop reliable maps for much larger areas, but still not at the scale of a province or state.

**Figure 1.** The location of Michigan, the study area in North America.

Lothian presents the fundamentals and an overview of various approaches to constructing visual quality maps and is a substantia update to the work of Taylor et al. [12, 28]

X4 = area of intermediate vegetation X6 = area of distant non-vegetation

X14 = area of wildflowers in foreground

X17 = area of dead foreground vegetation

X30 = open landscapes = X2 + X4 + (2 × (X3 + X6))

X52 = noosphericness = X7 + X8 + X9 + X15 + X16

copies of Lu et al. and Jin for a complete explanation [26, 31].

summed and squared, to compute Kendall's W value (Eq. 2). (Rj)2

j=1 n

the rankings for a column in computing the Kendall's W value [29, 30].

Next, the second group of images was compared to predictions made by the map through the use of Kendall's Concordance, a statistical technique that examines and tests for significant agreement/similarity [29, 30]. If the scores statistically agree, it is possible to create a reliable visual quality prediction map. This step in the methodology used here is explained in great detail by Jin [31]. Investigators interested in applying this methodology are advised to obtain

A Visual Quality Prediction Map for Michigan, USA: An Approach to Validate Spatial Content

http://dx.doi.org/10.5772/intechopen.79490

117

The test statistics were determined by applying Eqs. (2) and (3) to the data. The results are based upon rankings of treatment scores across rows. In this case, the rows are pairs of images between two treatments: (1) the predicted score for a randomly chosen site in the study area and (2) the actual score from a photograph taken at that location. There are 30 rows (pairs of scores) for this study (n = 30). The treatments are the columns (m = 2). The rankings are

Kendall's W value is a number ranging between 0 and 1. When W is near 0, there is no strong overall trend of agreement among the respondents. If W is near 1, then the responses could be

is the sum of the squares of

(Rj)2 − [3 m<sup>2</sup> n (n + 1)2]/[m<sup>2</sup> n(n<sup>2</sup> − 1)] (2)

X31 = closed landscapes = X2 + X4 + (2 × (X1 + X17))

X7 = area of pavement X8 = area of building X9 = area of vehicle

X10 = area of humans X11 = area of smoke

X15 = area of utilities

X19 = area of wildlife

X32 = openness = X30 − X31

X34 = mystery = X30 × X1 × X7/1140

W = 12∑

X16 = area of boats

These recent studies provide a setting for our investigation. As an extension of Lu's et al. and Jin's research, we were interested in applying this approach to all of Michigan (**Figure 1**) [26, 27]. We wanted to make a validated map for all of Michigan.
